20 Sci Tech Law 1

DATE DOWNLOADED: Tue Jun 25 21:34:45 2024
SOURCE: Content Downloaded from HeinOnline
Citations:
Please note: citations are provided as a general guideline. Users should consult their preferred
citation format's style manual for proper citation formatting.
Bluebook 21st ed.

David Husband & Stacey Zumo 20 SciTech Law. 35 (2023).
ALWD 7th ed.

David Husband & Stacey Zumo, , 20 SciTech Law. 35 (2023).
APA 7th ed.

Husband, David, & Zumo, Stacey. (2023). SciTech Lawyer, 20(1), 35-1.
Chicago 17th ed.

David Husband; Stacey Zumo, "," SciTech Lawyer 20, no. 1 (Fall 2023): 35-1
AGLC 4th ed.

David Husband and Stacey Zumo, '' (2023) 20(1) SciTech Lawyer 35
OSCOLA 4th ed.

David Husband & Stacey Zumo, '' (2023) 20 SciTech Law 35 Please
note: citations are provided as a general guideline. Users should consult their
preferred citation format's style manual for proper citation formatting.
-- Your use of this HeinOnline PDF indicates your acceptance of HeinOnline's Terms and
Conditions of the license agreement available at
https://heinonline.org/HOL/License
-- The search text of this PDF is generated from uncorrected OCR text.
-- To obtain permission to use this article beyond the scope of your license, please use:
Copyright Information
FORENSIC POLICE BODY DEEPFAKES
VOL 20, NO 2
FACIAL AND DASH AND
WINTER 2024
IDENTIFICATION CAMERAS GENERATIVE AI
TheSciTechLawyer
A PUBLICATION OF THE AMERICAN BAR ASSOCIATION | SCIENCE & TECHNOLOGY LAW SECTION
VIDEO EVIDENCE
SANDRA RISTOVSK A
ISSUE Published
EDITOR in The SciTech Lawyer, Volume 20, Number 2, Winter 2024. © 2024 American Bar Association. Reproduced with permission. All rights reserved. This information or any portion thereof
may not be copied or disseminated in any form or by any means or stored in an electronic database or retrieval system without the express written consent of the American Bar Association.
The
SciTech
Lawyer
SCIENCE & TECHNOLOGY
EDITORIAL BOARD
EDITOR-IN-CHIEF LAW SECTION OFFICERS MESSAGE FROM THE CHAIR
LARRY W. THORPE
Springield, TN
CHAIR
LAURA POSSESSKY By Laura Possessky
larrywthorpe@comcast.net Corporation for Public Broad-
SENIOR EDITORS casting
Washington, D.C.
MICHAEL A. AISENBERG lpossessky@cpb.org
Mitre Corp. SEEING IS BELIEVING
McLean, VA CHAIR-ELECT
maisenberg@mitre.org JOAN R.M. BULLOCK he Lumiere brothers took Paris by storm on a wintery Satur-
joan@reformedlawprof.com
MATTHEW HENSHON
VICE CHAIR
day in 1896 by screening the irst moving picture of a locomotive
Henshon Klein LLP
Boston, MA LOIS D. MERMELSTEIN arriving at a station. So convinced by what they saw, the audi-
mhenshon@henshon.com The Law Oice of
HON. RODERICK KENNEDY Lois D. Mermelstein ence stampeded out of the theater to escape the oncoming train.
Austin, TX
Albuquerque, NM
brazolargo@me.com lois@loismermelstein.com While many people for many years have debated the story’s truth, it illustrates
PETER F. MCLAUGHLIN SECRETARY the power video has had from the very beginning. By the early twentieth century,
Boston, MA MATTHEW HENSHON
peterfmclaughlin@outlook.com Henshon Klein LLP video reached beyond novel entertainment as politicians recognized the power
Boston, MA
LOIS D. MERMELSTEIN
The Law Oice of mhenshon@henshon.com of video to inluence opinion. As the irst universal mass medium, video quickly
Lois D. Mermelstein BUDGET OFFICER
CHRISTOPHER A. SUAREZ
became a valuable tool for governments and organizations to promote propa-
Austin, TX
lois@loismermelstein.com Steptoe & Johnson LLP ganda. Persuasion certainly has its role in the law and the irst documented uses
CAROL WILLIAMS Washington, D.C.
Aberystwyth University csuarez@steptoe.com of video evidence arose in the 1930s. By the twenty-irst century, digital inno-
Ceredigion, Wales UK
cas55@aber.ac.uk
SECTION DELEGATES
RICHARD L. FIELD
vations made video storytelling available to anyone with a phone: the resulting
ASSISTANT EDITORS Clifside Park, NJ
field@pipeline.com
impact of digital media in court proceedings has presented myriad opportunities
PETER J. GILLESPIE
Laner Muchin, Ltd. LUCY THOMSON and challenges.
Livingston PLLC
Chicago, IL
pgillespie@lanermuchin.com Washington, D.C. With the demonstrated ability of video to captivate, persuade, and inluence
lucythomson.aba@mindspring.
ROBERT KNAIER com
people, video naturally plays a valuable role in the courtroom. Lawyers, who are
Fitzgerald Knaier LLP
San Diego, CA IMMEDIATE PAST CHAIR trained storytellers, have relied on video evidence to artfully convince both judge
rknaier@fitzgeraldknaier.com GARTH B. JACOBSON
CT Corporation and jury of their clients’ version of the facts. But video evidence also generates con-
LISA R. LIFSHITZ Seattle, WA
Torkin Manes LLP gbjacobson@hotmail.com cerns about bias, voracity, and evidentiary authentication. Evidence procedures
Toronto, ON
llifshitz@torkinmanes.com PAST CHAIR LIAISON
TO OFFICERS
relating to admissibility are muddled with confusing and inconsistent approaches
SARAH E. MCMILLAN
McGlinchey Staford PLLC
HUGH BUTLER WELLONS both within and across court systems. Technological innovations raise the stakes
Spilman Thomas & Battle PLLC
New Orleans, LA
semcmillan@mcglinchey.com
Roanoke, VA even further, as demonstrated by AI’s impact on video evidence, with increasingly
hwellons@spilmanlaw.com
BRIAN SCARPELLI sophisticated deepfakes presenting new and unique challenges for evidentiary
ACT | The App Association AMERICAN BAR
Washington, D.C. ASSOCIATION authentication and admissibility.
CONTACTS
bscarpelli@actonline.org
DIRECTOR
Covering these topics and more, this issue also illustrates the important cross-
CHRISTOPHER A. SUAREZ
Steptoe & Johnson LLP BARBARA MITCHELL disciplinary collaborations within the ABA Science & Technology Law Section. It
Washington, D.C. barbara.mitchell@americanbar.
csuarez@steptoe.com org is a collaborative outgrowth of a 2021 project by Dr. Sandra Ristovska, professor of
ABA PUBLISHING EDITOR
TOMMY TOBIN
Perkins Coie LLP LORI LYONS
Media Studies at University of Colorado Boulder, who was in residence with SciTech’s
Seattle, WA
ttobin@perkinscoie.com
lori.lyons@americanbar.org Scientiic Evidence Committee as a Mellon/ American Council of Learned Societ-
ART DIRECTOR
SARA WADFORD ies (ACLS) Scholars and Society Fellow. Her project, “hrough the Lens of the Law:
sara.wadford@americanbar.org
ADVERTISING REPRESENTATIVE
Interpreting Video Evidence in US Courts in the Digital Age,” took an in-depth schol-
CHRIS MARTIN
(410) 584-1905
arly look at the use of video evidence in court proceedings (https://www.colorado.
chris.martin@wearemci.com edu/cmcinow/2021/04/20/through-lens-law). As part of this project, Dr. Ristovska
SECTION EMAIL ADDRESS presented the excellent and comprehensive program with SciTech, the Video in
stserve@americanbar.org
MEMBERSHIP QUESTIONS OR ADDRESS CHANGES
Legal Decision Making Webinar Series (https://www.americanbar.org/groups/
1-800-285-2221 or service@americanbar.org science_technology/events_cle/video-in-legal-decision-making-webinar-series).
The SciTech Lawyer (ISSN 1550-2090) is published quarterly as a service to
hese cooperative exchanges forge new insights on how science and technol-
its members by the Science & Technology Law Section of the American Bar
Association, 321 North Clark Street, Chicago, IL 60654-7598. It endeavors to
ogy inluence our work in the legal profession. SciTech’s active involvement of our
provide information about current developments in law, science, medicine, associate members—technologists, scientists, researchers, and academics among
and technology that is of professional interest to the members of the ABA
Science & Technology Law Section. Any member of the ABA may join others—inform, challenge, and question the course of the law in a way that lawyers
the Section by paying its annual dues of $60. Subscriptions are available
to nonmembers for $75 a year by contacting the ABA Service Center, oten cannot. hese unique dialogues have led SciTech to become an important
American Bar Association, 321 North Clark Street, Chicago, IL 60654-7598;
1-800-285-2221. Digital subscription packages for ABA periodicals are voice within the ABA to inform practitioners of the evolution and developments
available through HeinOnline. If interested, former print subscribers and
ABA non-members can visit www.heinonline.org for more information.
inluencing and transforming our legal system. he expertise of media scholars and
Requests to reprint articles should be sent to ABA Copyrights & Contracts,
www.americanbar.org/reprint; please send all other correspondence to
professionals and their systemic exploration of the impact of video on society ofer
The SciTech Lawyer Managing Editor Lori Lyons, lori.lyons@americanbar.
org. For more information, visit ambar.org/SciTechMagazine. The materials
valuable perspectives to solve issues arising from video evidence in the courts. I
contained herein represent the opinions of the authors and editors and hope this issue piques your curiosity as much as it did mine about the role of video
should not be construed to be those of either the American Bar Association
or The Science & Technology Law Section unless adopted pursuant to evidence in court proceedings, and I look forward to hearing your reactions and
the bylaws of the Association. The materials contained herein are not
intended as and cannot serve as a substitute for legal advice. Readers are relections on the articles.
encouraged to obtain advice from their own legal counsel. These materials
and any forms and agreements herein are intended for educational and
informational purposes only. Copyright © 2024 American Bar Association.
All rights reserved.
Published in The SciTech Lawyer, Volume 20, Number 2, Winter 2024. © 2024 American Bar Association. Reproduced with permission. All rights reserved. This information or any portion thereof
CONTENTS
2 MESSAGE FROM THE CHAIR
BY LAURA POSSESSKY
4 JUSTICE BY VIDEO: DO COURTS NEED LEGAL GUIDELINES FOR

VIDEO EVIDENCE?
BY SANDRA RISTOVSKA AND YAEL GRANOT
8 CONSTRUCTING MEANING IN VIDEO EVIDENCE

BY CHRISTINA SPIESEL AND NEAL FEIGENSON
13 HOW COGNITIVE BIAS CAN IMPAIR FORENSIC FACIAL

IDENTIFICATION
BY DEVON E. LABAT AND JEFF KUKUCKA
17 POLICE BODY CAMERAS: ANALOG THINKING ABOUT A DIGITAL

QUESTION
BY THE HON. RODERICK KENNEDY (RET.)
22 INTERNATIONAL VERIFICATION STANDARDS FOR OPEN SOURCE

VIDEOS: STRENGTHENING FAITH IN FACTS
BY ALEXA KOENIG
28 DEEPFAKES IN THE DOCK: PREPARING INTERNATIONAL JUSTICE

FOR GENERATIVE AI
BY RAQUEL VAZQUEZ LLORENTE
34 GET THE SCITECH EDGE: MEMBERSHIP AND DIVERSITY

COMMITTEE NEWS
BY JOANNE CHARLES
35 FUTURE SCITECH LEADERS: LAW STUDENT ENGAGEMENT

COMMITTEE NEWS
BY DAVID HUSBAND AND CAYLAN FAZIO
WINTER 2024 TheSciTechLawyer 3

F
rom cell phones to police body
cameras, today’s courts increas-
ingly use video as evidence. A
common assumption is that video can
help people bear witness to an event as
if they themselves were transported to
the complicated scene of its unfolding.
Rather than the secondhand testimony
provided by eyewitnesses, video is
assumed to provide an unmediated and
irsthand account directly to the deci-
sion-maker. Video has thus been praised
as a seemingly objective tool for justice
and accountability. At the same time,
video, just like eyewitness testimony, can
be subject to a host of biases of which
people are largely unaware.
Despite the potential for bias, U.S.
courts, at all levels, lack clear guide-
Justice by lines on how video can be used and

presented as evidence. As a result,
courts can interpret video evidence dif-
Video: Do Courts ferently, even within the lifespan of the
same case. Scott v. Harris is a renowned
U.S. Supreme Court case from 2007 that
Need Legal provided an early cautionary tale about
the challenges of video evidence.1 In this
Guidelines for ot-cited case, the court had to decide

whether a police car chase, which let a
driver paralyzed, violated the constitu-
Video Evidence? tional protection against unreasonable

seizure. he car chase was recorded on
two dashboard cameras. Lower courts
By Sandra Ristovska ruled that reasonable jurors could dif-
and Yael Granot fer as to whether the police had used
unreasonable force to end the chase,
requiring a jury trial. The Supreme
Court, though, ruled 8–1 in favor of
the oicer, explaining that the case was
“clear from the videotape” and that no
reasonable juror could agree with the
plaintif ’s account that the police had
used excessive force to stop the vehicle.
In an unprecedented move, the Court
uploaded the video to its website, invit-
ing the public to conirm that the video
“speak[s] for itself.”2 Social scientists
accepted the challenge and conducted
an experimental study showing that
people indeed had diferent interpre-
tations about the reasonableness of
the oicer’s behavior depicted in the
video. In fact, people’s perceptions dif-
fered depending on their cultural and
ideological backgrounds, such as racial
identity, education, income level, and
4 TheSciTechLawyer WINTER 2024

political ailiation. he scientists who researchers categorized “estimator Researchers found that this wider angle
conducted the study thus stated that variables” like lighting conditions and led to reduced perceptions of guilt and
what a video says “depends on to whom distance, whose impact could only be confession sincerity. he angle of the
it is speaking.”3 In other words, seeing estimated and weighted, versus “system camera directs viewers’ attention, oten
is not only about what the eyes physi- variables,” such as lineup construction, leading them to ascribe more agency,
cally see but the experiences and ideas which were more malleable to proce- and sometimes blame, to the target they
a viewer brings to an image. dural intervention.7 Distinguishing can see.
Photographic and videographic where biased interpretations of video his process of overweighting can
images work through two forces, which evidence come from can provide insight be so powerful that people may ascribe
visual communication researchers have into when and how instructions, reg- veracity to what they see, even when
termed denotation and connotation. ulations, or other solutions could be viewing something they objectively
Denotation refers to the literal mean- necessary and efective. know did not occur. In one set of stud-
ing, while connotation describes how In the examples we present below, ies, participants sat in a room with a
an image draws from the wider cultural we highlight three main issues identi- video camera and completed a comput-
context in lending meaning to what is ied in the literature8 that we think help erized task. When they got a question
depicted. When Justice Scalia famously capture the ways that interpretations correct, denoted by a green check on
compared the car chase in Scott v. Har- of video evidence may be biased. First, their computers, they were instructed to
ris to a scene in the Hollywood movie people struggle with how to distinguish take money from a bank. When they got
he French Connection, he drew on the between relevant and irrelevant infor- questions wrong, denoted by a red X on
connotative level of meaning, couch- mation, thus overweighting what they the screen, they could not take money.
ing the dashcam footage in a symbolic see. Second, people disregard the infor- Researchers then had participants come
frame consonant with broader under- mation they miss when viewing a video, in for a second session and accused
standings of the world as facilitated by thus underweighting what they do not them of taking money when they should
popular culture. he interplay between see. hird, people are overconident not have. Half of the participants were
denotation and connotation, between in their interpretations, thus lacking simply told there was video evidence of
the particular and the symbolic meaning awareness of their own potential for them committing this misdeed, whereas
of images, complicates the evidentiary bias. he seemingly direct, irsthand the other half actually saw doctored
status of video in court. experience they feel they get from video footage of themselves committing the
These two levels of visual mean- evidence makes it particularly diicult infraction.10 While most participants
ing help us distinguish between what to question their interpretations and ended up signing a confession for the
an image shows versus what a viewer subsequent judgments. experimenter, only those who watched
infers from it. And what viewers infer the doctored footage confabulated
from images is subject to bias from both OVERWEIGHTING: WHAT IS details to explain what happened, sug-
endogenous factors (e.g., prior attitudes SEEN IS SALIENT gesting they genuinely came to believe
and group identities) and exogenous Video evidence makes certain informa- they had committed the act. he weight
factors (e.g., features of the medium or tion particularly salient, and viewers of what they saw eclipsed their own
the viewing context). Lab experiments, ascribe greater agency or causal force knowledge of the facts.
for example, have demonstrated the to the targets or elements that they per- In these ways, people may overem-
endogenous inluence of identiication ceive as most salient in a video. his phasize the relevance of what they see,
with police on interpretation of videos phenomenon is potentially best exem- leading video’s salient features to have an
of police–civilian altercations.4 Similarly, pliied by work on the camera perspective outsized inluence over legal judgments.
cultural worldviews predicted interpre- bias.9 Across multiple studies involving Further, people may do this overweight-
tations and legal judgments regarding videotaped police interrogations, social ing even when they have clear reason to
videos of protest.5 On the other hand, psychologists found that the angle of the question their own eyes.
exogenous inluences account for why camera greatly inluenced how people
videos played in slow motion, compared evaluated the confession they saw. If, UNDERWEIGHTING: WHAT IS
to normal speed, resulted in judgments for example, they saw only the suspect, MISSED IS TRIVIALIZED
of greater intentionality in the depicted who might have been sweaty or nervous, Just as people may overweight the
action.6 they were likely to think he appeared information they see, they also con-
The binary distinction between guilty, thus believing in the sincerity of versely underweight the information
endogenous and exogenous influ- his confession. If, however, people were they miss. Information may be missed
ences is useful because it can facilitate given more visual information—such as because viewers direct their attention
the development of a taxonomy of the when the angle of the camera included away from it. However, sometimes view-
sources of systematic bias in apprais- both the police oicer and the suspect— ers even miss things directly in their
als of video evidence. his process can people suddenly had another actor to line of sight, a phenomenon known
mimic work on eyewitness testimony; whom they could ascribe causal agency. as inattentional blindness. In the most

well-known demonstrations of this his perceptual “illing in” extends to toward the incorrect information their
phenomenon, commonly called the real-world scenes. Researchers have eyes supplied.15
“invisible gorilla” studies, over half of identiied a phenomenon called bound- People consider their visual experi-
people missed a man in a gorilla suit ary extension, where people, informed ences to be accurate and reliable even
walking through the middle of a ball- by past experiences, misremembered as they are willing to acknowledge
passing game between players in white images as having been more expansive others’ susceptibility to biases when
and black shirts.11 his study was in part than they were, potentially making up viewing video. In considering video
inspired by the real legal case of Boston details that were never shown.14 evidence of a police–civilian interac-
police oicer Kenneth Conley, who was In other words, people might ascribe tion, participants said they were less
involved in a foot pursuit. In the course less weight to potentially relevant infor- susceptible to bias than the average
of chasing a suspect, Conley ran by fel- mation if it is not presented or they do American, indicating that “If I’m pay-
low oicers beating a Black man who not attend to it. Further, missed infor- ing very close attention to the event, I
was later established to be an under- mation may carry such little weight can prevent my worldview from afect-
cover oicer. he court found Conley because people fail to question its ing my understanding of [the video]”
guilty of perjury when he claimed he absence. was more true for themselves as com-
had not witnessed the beating despite pared to other viewers.16
passing directly by it. Yet, research on Courts are not yet well equipped to
inattentional blindness might explain address the problems that overconi-
how it is possible to miss dynamic dence creates. Some of the elements
actions happening within one’s ield of built into the legal process that might
vision. mitigate these biases do not seem to
Information may also be missed be efective. For example, one might
because the camera angle fails to capture imagine that the opportunity to watch
it. Police body-worn cameras (BWCs) video evidence multiple times through-
usually show very limited visual cues out legal proceedings would undo any
of the oicer, as compared to dashboard weighting errors that might occur
cameras mounted on police cars, which upon irst appraisal. In one study, how-
tend to have a broader visual perspec- ever, researchers showed participants
tive. In one study, scientists showed the same videotaped police–civilian
participants the same police–civilian altercation twice, using eye-tracking
altercation captured by either BWC technology to measure their patterns
or dashboard camera. hey found that of visual attention. Results revealed
body camera footage systematically a visual conirmation bias, such that
decreased perceptions of police inten- people seemed to systematically follow
tionality and culpability for the same the same visual path they had already
actions relative to dashboard camera tread.17 Unless viewers are encouraged
footage.12 he lack of visual salience to gather new information, they might
of the oicer prevented people from OVERCONFIDENCE: WHAT IS become even more certain about their
ascribing agency to him, potentially PERCEIVED IS CERTAIN initial interpretation.
underweighting his role in the depicted The tendencies to overweight and
event. underweight may be particularly per- REGULATING VIDEO EVIDENCE
People may not realize they are nicious in that people do not realize The challenges that overweighting,
engaging in underweighting because the their own vulnerability to these biases. underweighting, and overconidence
brain oten ills in the blanks, perceptu- Viewers are incredibly conident in their create speak directly to the discrep-
ally and cognitively, in such a way that visual experiences. his conidence is ancies between what a video shows
people do not feel like they may be miss- so strong that they may trust errone- and how a viewer interprets it. Judges,
ing information when viewing images. ous visual information even when it is lawyers, and jurors operate within a
In one set of studies, for example, par- directly contradicted by relevant infor- legal framework that has not yet given
ticipants saw incomplete images like a mation from the other senses. In a images the same procedural scru-
half-completed picture of a butterly. On classic psychology study, participants tiny as it has given to words. When
a subsequent memory task, participants were asked to estimate the size of an video is presented as an opportunity
were asked to recall what they saw and, object held in their hand, which they for those in the courtroom to witness
when given a chance to choose between looked at through a size-distorting lens. with their own eyes the events at the
a full butterly or the partial image they hough their hands were giving them core of a litigation, the consideration of
had in fact seen, they were more likely to an accurate tactile perception, peo- the various inluences on how people
misreport seeing the complete image.13 ple’s estimates were more likely to err interpret images may get sidelined. Yet

courts tend to see video as an objective existing challenges with video evi- 9. G. Daniel Lassiter et al., Accountability
record, partly due to the uncertainty dence, we argue that safeguards for and the Camera Perspective Bias in Videotaped
regarding the classification of video as visual interpretation in court are Confessions, 1 Analyses of Soc. Issues &
evidence and the insufficient attention urgently necessary. Short of clear Pub. Pol’y 53 (2001).
to visual legal literacy. Together, these guidelines for the use of video as evi- 10. Robert A. Nash & Kimberley A. Wade,
factors contribute to an inconsistent dence, judges, lawyers, and jurors are Innocent but Proven Guilty: Eliciting Internal-
legal approach to video evidence.18 left treating video in highly varied ized False Confessions Using Doctored-Video
Video’s classification as evidence ways that can lead to uneven render- Evidence, 23 Applied Cognitive Psych. 624
is significant because it can influence ings of justice. Research on guidelines (2008).
how judges decide on video’s admissi- that can optimize the overall consider- 11. D. J. Simons & C. F. Chabris, Gorillas in
bility or how attorneys engage with the ation of video as evidence is evermore Our Midst: Sustained Inattentional Blindness
footage in court.19 The introduction of important for promoting consistency for Dynamic Events, 28 Perception 1059
photography in the nineteenth century and fairness for those who use the (1999).
gave rise to demonstrative evidence as courts. 12. Broderick L. Turner et al, Body Camera
a category governing the use of images Footage Leads to Lower Judgments of Intent
as evidence.20 Under this view, photo- Sandra Ristovska, PhD, is an assistant han Dash Camera Footage, 116 Proc. Nat’l
realistic images cannot prove facts on professor of Media Studies at the Acad. Sci. 1201 (2019).
their own. They can merely illustrate University of Colorado Boulder. As 13. Mary Ann Foley et al., Remembering
what witnesses say or what other evi- part of a 2021 Mellon/ACLS Scholars More han Meets the Eye: A Study of Memory
dence shows. Demonstrative evidence, & Society Fellowship, she was a Confusions About Incomplete Visual Informa-
however, has always been an uncertain resident researcher with the Scientiic tion, 15 Memory 616 (2007).
category ranging from illustration to Evidence Committee of ABA Science & 14. H. Intraub & M. Richardson, Wide-
proof.21 It remains an ambiguous term Technology Law Section. Angle Memories of Close-up Scenes, 15. J.
because both courts and legal schol- Experimental Psych.: Learning, Memory
ars disagree about its meaning.22 Video Yael Granot, PhD, is an assistant & Cognition 179 (1989).
can be used as an illustrative aid, but professor of Psychology at Smith 15. Irvin Rock & Jack Victor, Vision and
its role as direct evidence, capable of College and an ailiated scholar with Touch: An Experimentally Created Conlict
independently proving facts, is also Yale Law School’s Justice Collaboratory. Between the Two Senses, 143 Science 594
common. Video has also always been (1964).
subject to the “silent witness” excep- ENDNOTES 16. Kristyn A. Jones, William E. Crozier &
tion when recording events no one 1. 550 U.S. 372 (2007). Deryn Strange, Objectivity Is a Myth for You
else was around to see.23 At the heart 2. Id. at 378 n.5. But Not for Me or Police: A Bias Blind Spot for
of these shifting and ambivalent cat- 3. Dan M. Kahan, David A. Hofman & Viewing and Remembering Criminal Events,
egories under which video can be Donald Braman, Whose Eyes Are You Going 24 Psych., Pub. Pol’y & L. 259, 261 (2018).
admitted as evidence is an uncer- to Believe? Scott v. Harris and the Perils of 17. Jennie Qu-Lee et al., he Relationship
tainty about how to evaluate video’s Cognitive Illiberalism, 122 Harv. L. Rev. 837, Between Visual Confirmation Bias, Belief
probative value, or the degree to which 903 (2009). Consistency, and Belief Polarization, 6 Com-
it can prove the facts it is offered to 4. Roseanna Sommers, Note, Will Putting prehensive Results in Soc. Psych. 1 (2022).
prove. Cameras on Police Reduce Polarization?, 125 18. Sandra Ristovska, When Believing
The appraisal of video evidence is Yale L.J. 1304 (2016). Can Be Seeing: he Unregulated Approach to
further complicated by cognitive and 5. Dan M. Kahan et al, “hey Saw a Protest:” Video Evidence in the U.S. and the Need for
social factors that influence how peo- Cognitive Illiberalism and the Speech-Conduct Archival Legal Standards, 28 First Monday,
ple see and interpret images. Visual Distinction, 64 Stan. L. Rev. 851 (2012). no. 7, 2023.
literacy training that addresses these 6. Eugene M. Caruso, Zachary C. Burns & 19. Yael Granot et al., In the Eyes of the
factors is still not broadly represented Benjamin A. Converse, Slow Motion Increases Law: Perception versus Reality in the Appraisals
in law school curricula and profes- Perceived Intent, 113 Proc. Nat’l Acad. Sci. of Video Evidence, 24 Psych., Pub. Pol’y &
sional programs, though there are 9250 (2016). L. 93 (2018).
more trial advocacy programs today 7. Gary L. Wells et al., Policy and Procedure 20. Jennifer L. Mnookin, he Image of
that teach how to interrogate video. Recommendations for the Collection and Pres- Truth: Photographic Evidence and the Power
Visual literacy can help judges and ervation of Eyewitness Identiication Evidence, of Analogy, 10 Yale Journal of Law & the
attorneys learn strategies for how to 44 Law & Hum. Behav. 3 (2020). Humanities 1 (1998).
analyze evidentiary video—how to 8. Yael Granot & David Igliozzi, Psycho- 21. Id.
probe and ask relevant questions of logical Perspectives on the Presentation of 22. Neal Feigenson, Experiencing
the underlying content. Video Evidence: How Perceivers Weight What Other Minds in the Courtroom (2016).
At times when generative AI and Is Seen and Unseen, 28 First Monday, no. 7, 23. Robert P. Mosteller et al.,
deepfakes threaten to exacerbate July 2023. McCormick on Evidence (8th ed. 2020).

Constructing
Meaning in
Video Evidence
By Christina Spiesel and Neal Feigenson

P
eople intuitively believe that video together. In Kennedy v. Bremerton School or know little to nothing about it. he
puts them directly in touch with District,2 a high school football coach point is that what the photo means to
what really happened. hey tend made it a practice to pray ater games viewers can be as varied as their diver-
to think that the camera objectively at the 50-yard line with members of gent views about whether this sort of
records what is unfolding in front of the both teams. He claimed that the school prayer by a public employee on school
lens and that the resulting video looks district had ired him for violating its grounds ought to be allowed. Advo-
like what they would have seen had policy against public religious practice cates face the challenge of deciding
they been there. Of course, video never by public employees while on the job. which cultural community they most
just shows us what happened. And it When the case came before the Supreme need to address. If the case teaches us
never just “speaks for itself,” as Supreme Court, Justice Gorsuch, writing for the anything, it is that how to decide that is
Court Justice Antonin Scalia famously majority, described the coach’s actions as nonobvious.
incanted in Scott v. Harris, where the ofering “a quiet prayer of thanks.”3 Jus- One last observation. he original
Court accorded dashcam video the tice Sotomayor, dissenting, wrote: “his photo that was part of the trial court
status of conclusive proof.1 Video is case is about whether a school district is record is in color. Does the black and
often incomplete, often ambiguous required to allow one of its employees white photo in Justice Sotomayor’s opin-
with respect to the issues in dispute, to incorporate a public, communicative ion reduce its emotional intensity and
sometimes misleading, sometimes in display of the employee’s personal reli- bring the image closer to writing, mak-
conlict with other video, and almost gious beliefs into a school event.”4 She ing the markings on the ield and the
always open to reasonable diferences included in her opinion three photo- numbers on the uniforms of the play-
in interpretation. Given the prolifera- graphs from the record, visual evidence ers more salient—and if so, might that
tion of surveillance cameras and now that Justice Gorsuch did not address. distract our attention from the issues
with cameras on every cell phone, there Justice Sotomayor sets the visual text in before the Court? And how might the
may also be multiple videos of litigated dialogue with the written text, almost black and white image afect the public
events, many of them posted to the mano a mano, inviting us to see for audience for the opinion?
internet, where even more people may ourselves.
try to decode what has happened. he irst photo from her opinion VIEWING VIDEO
Understanding what video means, shows the coach surrounded by his FRAME-BY-FRAME
what it is and is not capable of providing team (and possibly players from the Sometimes information in a video can
evidence of, always demands look- opposing team) while raising two hel- be discovered only if we watch it frame
ing inward toward the medium itself. mets up high.5 his certainly seems to by frame. The current standard for
Sometimes critical information can be refute the majority’s description of the full-resolution video is 24 frames per
elicited only by scrutinizing the individ- prayer as “private” (it is hard to see the second, but that is just what the monitor
ual frames; sometimes the crucial fact people in the stands who are watching) sees and the timecodes indicate; it’s not
occurs between the frames, so it’s not and perhaps also “quiet” (although the what our eyes see. People can respond
actually in the video at all but implied photo doesn’t have sound). What else to stimuli in fractions of a second, but
by what comes before and ater. Under- might it signify? Are the raised helmets not that fast, and they respond more
standing video also requires looking trophies? Objects of veneration? Does slowly to visual stimuli (about a quarter
outward toward the contexts in which raising them elevate the boys who pre- of a second) than to auditory or tactile
every video is embedded. Advocates and sumably wore them as having engaged ones.7 he important message here is
witnesses provide verbal context; other in a kind of sacred activity? that human eyes can miss the details in
images, including other video, can pro- Many viewers will be aware, if only video whizzing by at 1/24th of a second
vide visual context. Videos also carry vaguely, of the deep-seated ties between per frame—not to mention that impor-
cultural associations of all sorts, which religion, speciically Christianity, and tant things that happen between frames
inlect the meanings that legal decision- American football, as evidenced by are not in the video at all.
makers take away from their viewing phrases like the “Hail Mary” pass and Lawyers who need to explore a
(and listening) and can lead lawyers the “Touchdown Jesus” mural on the piece of video for its full evidentiary
to make assumptions, right or wrong, Notre Dame University campus.6 Many value have to upload it into video edit-
about how others will interpret what of them would also bring to the picture ing sotware, which may enable them to
they are seeing. We illustrate these gen- some awareness that the 50-yard line is get more information out of it than they
eral points with six observations based a special place on a football ield—not can by watching the more lifelike, fast-
on speciic cases. just because it’s center stage, but also moving stream. One of us (Spiesel) was
because it oten features a logo associ- asked to explore a piece of body-worn
PRESENTING STILL ated with the home team and hence the camera video in a case involving police
PHOTOGRAPHIC EVIDENCE home institution. For other viewers, the oicers’ use of excessive force.8 he one
We begin with still photographs because photo would prompt none of these asso- video submitted for analysis was 6½
video is made up of still frames strung ciations; they might just enjoy football minutes long—over 9,000 frames. he

oicer whose video was submitted was close quarters and close to the recording this case, read opening statements from
the second to arrive at the scene. Did the device, coordinating with each other, the prosecution and defense attorneys,
other oicer not have a bodycam? If he cannot be heard even with the ruckus in which the lawyers tell them (or not)
did and he turned it on, what happened in the background. Examining the what to look for in the video.11 he study
to his data? It is reasonable to assume soundtrack in a video editing program has found that verbally framing mock
that the plaintif ’s lawyer was looking reveals the dense siren, which masked jurors’ viewing by instructing them
for evidence of excessive force, but the any sound waves from conversation. on what they will see afects what they
review of the video was also looking for his prompts one last observation use- report having seen. For instance, when
evidence that the bodycam data may ful to attorneys (and others): In a video the prosecutor said, “Oicer Tensing
have been altered. editing program that shows the frames, was already pointing his gun directly
he video begins with the sound audio is a picture—a waveform—and at the driver’s head before the driver
of everyday communications between not only a sound. hose waveforms can reached his right hand to put the key
the second oicer and the dispatcher be very revealing (as well as a tool for in the ignition,” mock jurors were sig-
while the oicer is driving to the scene. precisely editing sound, should that be niicantly more likely to say that that’s
Upon arrival, the oicer moves to the desired). It takes patience to view all what they saw than when the prosecu-
suspect’s parked vehicle, removes him those still frames, but alert observation tor did not tell them to look for this.
from the car, and puts him on the can discover things otherwise unseen. What is particularly noteworthy about
ground. he direction of the camera this example is that, based on the video
constantly and rapidly changes—the VERBALLY FRAMING VIDEO itself, the prosecutor’s statement is
bodycam was attached to the oicer’s VIEWING FOR JURORS unambiguously false: he driver’s hand
torso and moved as he did—confus- When video material like this does go is clearly holding the key in the igni-
ing what can be seen. he oicer who to court, it is seen in the verbal con- tion well before the oicer pulls out his
was already present joins in. here is a texts the lawyers give it. We turn now gun. hus, verbally framing the video
lot of tussling as the suspect ostensibly to experimental evidence indicating that can lead viewers to think they saw what
tries to get up. Some frames show one the words advocates use to frame jurors’ didn’t happen, as well as incline them to
and then both oicers putting pressure viewing of a video can afect what jurors resolve in the advocate’s favor what the
on the suspect’s neck with their hands. think they see. video leaves more ambiguous. his is
At least part of that time, we can infer In State of Ohio v. Tensing,9 a police one very important way in which advo-
from the camera angle that the oicer oicer pulled a driver over for not dis- cates can try to make video evidence
wearing the camera is straddling the playing his front license plate in the speak as they want it to.
suspect’s hips. Exactly one frame in the proper place. Ater the driver did not
whole vast stream of pictures shows the produce his driver’s license, the oicer USING MULTIPLE VIDEOS TO
right hand of an oicer making a ist, asked him to step out of the car. When TELL A STORY
which seems to be hanging from above the driver refused and instead began to On October 1, 2017, Stephen Paddock
and aimed directly at the camera. But start the engine as if to leave the scene, used a hammer to break windows fac-
there is no frame showing the ist strik- a brief struggle ensued, and the oicer ing the street in his two suites on the
ing the suspect. At the end, the suspect fatally shot the driver point-blank in the thirty-second loor of the Mandalay
is shown walking, with assistance, away head.10 he crucial issue in the oicer’s Bay Hotel, opposite the Harvest Coun-
from the scene. murder trial was whether he was justi- try Music Festival, then in full swing. At
he video leaves us with many ques- ied in doing so because, as the driver 10:05 p.m. he began shooting through
tions. Why do we not see the suspect started up the car, the oicer reasonably those broken windows, resulting in the
getting up and starting to walk? How believed his life was in danger. And that largest massacre carried out by a sin-
can we interpret very fast changes in the largely depended on a factual question: gle gunman, leaving sixty people dead
positions of the oicers, both of whom Was the oicer in harm’s way, about to and over 800 injured. Paddock died of
appear to be well-muscled guys? Just be run over, when he ired his weapon, a self-inlicted gunshot wound before
how much time does it take for them or did he shoot the driver irst? It’s hard authorities could reach him. Had he
to change positions? Are any frames to determine this from the officer’s lived, there might well have been a trial
missing? body-worn camera video. Everything and a public accounting, perhaps even
And why do we not hear the oicers’ happens fast: he oicer’s own move- a satisfactory explanation for why he
voices when they seem to be commu- ments create motion blur in the images, did it.
nicating with one another, although the isheye lens distorts the view, and Victims and survivors sued the hotel,
we cannot see their mouths moving in it is diicult to sync the sounds on the claiming that it bore some responsibil-
speech? Police cruiser sirens are scream- video to what can be seen. ity for allowing Paddock to accumulate,
ing in the background throughout. Were One of us (Feigenson) has been con- over the course of a week, the arsenal
they really on the whole time? It seems ducting a study in which mock jurors, he used in the massacre. Video of Pad-
improbable that two men working in before watching the actual video from dock’s movements in and out of the

hotel’s public spaces was captured on
several surveillance cameras. he case
thus ofers an opportunity to explore
how multiple video streams may be
deployed to tell a compelling story. Our
discussion is based on the New York
Times video reporting, which includes
security camera footage of Paddock
from his checking in on September 25
to the terrible night of October 1, as
well as diagrams and other documen-
tary images.12 he video segments show
Paddock, an unostentatiously dressed,
60-something white man, calmly
entering the hotel over and over again,
carrying various suitcases and satch-
els, assisted by porters and others, with
whom he chatted and whom he tipped.
He repeatedly enters elevators with a lot
of luggage. No one seems to be taking be sure you have it all? And when you value of giving jurors a single emblem-
any special notice. do, how are you going to make sense of atic picture to hold in their minds, and
he lawsuit settled, but we can ask: all of it to tell the story of your case? Be they showed it repeatedly throughout
If something like the Times video com- prepared to be a bit of a ilm director. the trial.
pilation had become part of a legal he prosecution also incorporated
argument, which side would it serve? TURNING A VIDEO INTO A the same still frame from the video, as
For instance, does the video help the MULTIMEDIA PRESENTATION well as other images and text, into a
plaintiffs make the case that hotel Lawyers will oten want to go beyond timeline of the crucial events that pro-
employees should have foreseen that integrating multiple videos and using batively summarized their narrative.16
this well-behaved guest, who (based on their own words to frame the audience’s Among other things, by placing the
past visits) seemed his usual self, was viewing. Everyone saw Darnella Fra- representations of Chauvin’s conduct
planning mass murder? Or would it help zier’s video of the killing of George across the top of the slide, above but
the defendant more? Paddock certainly Floyd.13 One might have been tempted separated by empty background from
did not it the now-common stereotype to think that just seeing the video its efects on Floyd below, prosecutors
of the out-of-control angry adolescent would have been enough for jurors to visually emphasized the psychological
or early adulthood shooter. He looked convict Oicer Derek Chauvin of mur- distance between the two, and hence
like a middle-aged man on a gambling der. But prosecuting a police oicer, Chauvin’s disregard for everything
trip, the sort of person the hotel was especially for murder, is never a slam that was happening that should have
used to accommodating. Could the dunk. How can one prove beyond a made him realize, as the minutes went
hotel have put the developing story reasonable doubt that Chauvin caused on, that he was risking serious harm
together at the time from reviewing Floyd’s death “by perpetrating an act . . . to Floyd.
its own security footage? It is doubtful evincing a depraved mind, without In closing argument, prosecutors
that Paddock would have stood out in regard for human life” (part of the emphasized the moral dimensions of
the many lower-quality video feeds that third-degree murder charge)?14 the video evidence. hey valorized the
guards would have been monitoring. If he prosecution team did not let bystanders’ actions of videorecording
anything, the video provides a kind of the Frazier video speak for itself. hey the crime in the religious language of
character evidence: Paddock appeared wove it into their theory of the case, “bearing witness.” And they described
to be meticulous and aware of his envi- both visually and verbally, each per- the bystanders’ videos as “precious”
ronment and related equally to men and suasive tactic reinforcing the others. gits, which the bystanders “gave to
women as he encountered them. All of Here are just a few examples. In their you,” the jurors, which—according to
this might be useful to the defendant, opening statement, before playing the anthropologists who have studied cul-
but it emerges only from seeing the video, prosecutors displayed a single tures of git exchange—imposed on the
thirty or so video bits knitted into this frame taken from it, showing Chauvin jurors the obligation to reciprocate by
documentary. kneeling on Floyd’s neck and glaring rendering the desired guilty verdict.17
Increasingly, advocates have access to at the camera.15 Of course, the Frazier hus, with video as the focal point,
much more video than was imaginable video (and others) would be central prosecutors constructed a multime-
even ten years ago. How are you going to to the case, but prosecutors knew the dia presentation combining diferent

videos, still images, and diagrams, as decision-makers. And for that reason, 11. Jaihyun Park, Neal Feigenson & Ngayin
well as words, to hammer home their it remains imperative for advocates to Cheng, Efects of Verbal Framing and Attitudes
theory of the case. appreciate what video evidence can Toward Police on Mock Jurors’ Judgments of
mean and can be made to mean. Being Body-Worn Camera Video (unpublished
CITING IMAGES IN ADS, VIDEO visually literate depends on closely manuscript on ile with authors).
GAMES, AND SOCIAL MEDIA observing what is actually there before 12. Malachy Brown et al., 10 Minutes.
Vast as the world of video evidence is, the eyes, and then thinking carefully 12 Gunire Bursts. 30 Videos. Mapping the
lawyers need to be aware of an even about how what is there, in all of its Las Vegas Massacre, N.Y. Times (Oct. 21,
wider visual environment, including many dimensions, can help (or hin- 2017), https://www.nytimes.com/video/
video games and social media, both der) the advocate’s persuasive aims. us/100000005473328/las-vegas-shooting-
as potential sources of compelling timeline-12-bursts.html.
visual evidence and also for famil- Christina Spiesel is senior research 13. Frazier video: Shaun King, The
iarizing themselves with the cultural scholar in law and ailiated Minneapolis Police Choke an Unarmed
surround of their cases. After the 2012 faculty fellow at Yale Law School’s Handcufed Black Man to Death, YouTube
massacre at the Sandy Hook Elemen- Information Society Project, co-author (May 2020), https://www.youtube.com/
tary School in Newtown, Connecticut, of Law on Display: The Digital watch?v=g8hGKB5QDhw.
the parents of several of the murdered Transformation of Legal Persuasion 14. Criminal Complaint, State v.
children sued the manufacturer of the and Judgment, author of numerous Chauvin, No. 27-CR-20-12646 (Hen-
assault rifle the killer used. The plain- chapters and articles, professor of nepin Cnty. Dist. Ct. 2020), https://web.
tiffs’ lawyer would need to prove that visual persuasion, and a legal visual archive.org/web/20200604085340/http://
the defendant’s wrongful marketing of consultant. mncour ts.gov/mncour tsgov/me dia/
the weapon causally contributed to the High-Proile-Cases/27-CR- 20-12646/Amend-
massacre. He was, of course, prepared Neal Feigenson is a professor of edComplaint06032020.pdf.
to show the defendant’s ads.18 But he law at Quinnipiac University Law 15. Derek Chauvin Trial Opening State-
also used imagery from video games School. He writes about the uses of ments (State v. Chauvin), YouTube (Mar. 29,
like the one the killer spent countless visual media in legal practice and has 2021) (from Wash. Post), https://www.youtube.
hours playing to show that the defen- authored or co-authored three books, com/watch?v=IWbA2brcEbg (this was Exhibit
dant used product placement to put including Law on Display: The Digital 17 at the trial).
realistic renditions of its assault rifles Transformation of Legal Persuasion 16. Trial of Derek Chauvin Day 1, YouTube
in those boys’ virtual hands—and, in and Judgment and Experiencing (Mar. 29, 2021) (from Fox News Now), https://
juxtaposition with crime scene photos, Other Minds in the Courtroom. www.youtube.com/watch?v=HehKMmW0pjc
to connect that advertising with the (beginning about 36:20).
crime.19 This computer- and culture- ENDNOTES 17. Prosecution Closing Argument Full
savvy plaintiffs’ lawyer knew where 1. 550 U.S. 372, 378 n.5 (2007). Transcript: Derek Chauvin Trial for Murder of
to look for images that would help his 2. 142 S. Ct. 2407 (2022). George Floyd, Rev.com (Apr. 19, 2021), https://
case and knew what to do with those 3. Id. at 2415. www.rev.com/blog/transcripts/prosecution-
images when he found them. As advo- 4. Id. at 2441 (Sotomayor, J., dissenting). closing-argument-full-transcript-derek-
cates try to persuade decision-makers 5. Id. at 2436. chauvin-trial-for-murder-of-george-floyd
and the public at large, being aware of 6. For both terms, see Hail Mary (day 15).
the sorts of images that are out there Pass, Wikipedia (Nov. 23, 2023), https:// 18. Gawker.com, Bushmaster Firearms:
in mainstream culture is more impor- en.wikipedia.org/wiki/Hail_Mary_pass. “Your Man Card Is Revoked.”
tant than ever before. 7. For example, see Robert Kosinski, A 19. Elizabeth Williamson, How hey Did
The presumptive credibility of Literature Review on Reaction Time (Sept. It: Sandy Hook Families Gain Long-Awaited
video is now under attack. Computers 2013), https://www.sciencebuddies.org/ Legal Wins, N.Y. Times (Feb. 21, 2022), https://
can generate photorealistic images of Files/16478/4/clemson.rt.pdf. www.nytimes.com/2022/02/20/us/politics/
things that never appeared in front of 8. Because the case settled and is not a sandy-hook-legal-victories.html (images from
any lens.20 In the words of W. J. Mitch- matter of public record, we are not able to Koskof, Koskof, & Bieder PC). he crime
ell, “the referent has become unstuck,” provide more details or the visual evidence scene photo close-up is on the let; the screen
and with it, the axiomatic reliability of itself. capture from Call of Duty is on the right.
video that derives from its indexical 9. No. B1503961 (Ct. Common Pleas, 20. “Deep fake” videos abound in other
relationship to the reality it depicts.21 Hamilton Cnty., 2016). domains. E.g., Orange—la Compil des
While the availability of deep- 10. Samuel DuBose: Full Police Dash- Bleues (English version), YouTube (June
fakes may make the authentication of cam Video of Shooting, The Guardian 29, 2023), https://www.youtube.com/
video evidence more laborious and (Sept. 1, 2015), https://www.theguardian. watch?v=QVNZRHIZVL8.
sometimes more contested, that evi- com/us-news/video/2015/sep/01/ 21. William J. Mitchell, The Reconfig-
dence will continue to preoccupy legal samuel-dubose-killing-full-dashcam-video. ured Eye 31 (1994).

H
ave you ever been told that you
resemble a relative who actually
looks quite diferent than you?
You’re not alone. In a 2002 study, partici-
pants were asked to judge the similarity
of numerous pairs of faces, and they per-
ceived the faces as more similar if told
that the two people were genetically
related—even if they weren’t.1 In fact,
participants’ beliefs about whether the
two people were related had a stronger
efect on their perceptions than whether
the people were actually related.
Psychologists have long understood
that our beliefs, desires, and expecta-
tions inluence how we see the world
around us, such that people with dif-
ferent mindsets oten interpret the very
same information in markedly diferent
ways. Imagine, for example, watching
a sporting event with a friend and dis-
agreeing over a referee’s decision, even
though you both saw the same play. Or
remember when society lost its collec-
tive mind over whether “the dress” was
black and blue or white and gold?2 hese
disagreements relect a phenomenon
known as cognitive bias.
In a nutshell, cognitive bias is our
brain’s way of simplifying our environ-
ment by using our personal beliefs and
experiences to select, organize, and inter-
pret new information more eiciently.
Cognitive bias is oten beneicial because
it subconsciously distorts our perception
How Cognitive in ways that encourage adaptive behav-

ior. In one study, for example, people
who were wearing a heavy backpack
Bias Can Impair estimated the same hill to be steeper

than other people who weren’t wearing
a backpack, which is our brain’s subtle
Forensic Facial way of discouraging us from climbing

that hill.3
However, because cognitive bias pro-
duces diferences in perception, it can
Identification also impede eforts to determine the

truth—such as in forensic science anal-
yses. hat is to say, if cognitive bias leads
By Devon E. LaBat and Jef Kukucka two forensic examiners to form diferent
opinions of the same evidence, then at
least one of them must be wrong, which
can trigger a costly miscarriage of justice
if let unchecked.
Importantly, cognitive bias has a
stronger efect on judgments that are
challenging rather than clear-cut. In

forensic facial comparisons, an exam- practitioners to “rely solely on task-rel- similarities between the two prints and
iner compares an image of an unknown evant information when performing thus overlook minutiae that they would
person (oten the perpetrator of a crime) forensic analyses,”9 but, unfortunately, have otherwise noticed.
against one or more known individu- it appears that many forensic examiners Informed by these findings, psy-
als and decides which, if any, of those still receive potentially biasing informa- chologists now urge forensic examiners
individuals’ faces matches the unknown tion—such as information about the to analyze the unknown sample on its
person. his is a diicult task; even pro- suspect’s criminal history—as a matter own before making a side-by-side com-
fessionals who have specialized training of standard procedure.10 parison against a known sample—and,
and ample experience in face matching Relatedly, many forensic labs require more broadly, to make (and document)
tend to exhibit high error rates (7–31%) examiners’ opinions to be veriied by a thoughtful advance decisions about what
on tasks designed to mimic real-world colleague to ensure that they agree. In information they will review and in what
facial comparisons.4 theory, veriication is a useful quality order.14 In forensic facial comparison,
control measure—but, in practice, the where focusing on diagnostic facial fea-
CONTEXT DISTORTS FORENSIC verifying examiner is oten aware of tures (similar to ingerprint minutiae)
EXAMINERS’ JUDGMENTS the original examiner’s opinion, which has been shown to enhance identi-
Cognitive bias can further impede facial creates a prejudice to agree with it. For fication accuracy, examiners should
identiication accuracy, especially when example, one irearms laboratory found therefore analyze the unknown image
examiners receive information that is that veriiers were much less likely to for such features before comparing it
irrelevant to the facial comparison. It is disagree with the original examiner’s against any known images.15
well known that the same image can be opinion if they were aware (12.5%)
interpreted diferently depending on the rather than unaware (42.3%) of it.11 hus, THE COGNITIVE BIAS BLINDSPOT
context in which it appears; for exam- the standard method of veriication is FOR FORENSIC SCIENTISTS
ple, when researchers showed people treacherous in that it provides only the Can we prevent bias? And, if so, how?
an ambiguous drawing that could be illusion of independent corroboration. One influential theory explains that
perceived as either a squirrel or a swan, in order to overcome bias, we must
they tended to see it as a squirrel when it SEQUENCE OF IMAGES be (a) aware of the bias, (b) motivated
appeared on a tree branch but as a swan INFLUENCES FORENSIC to correct it, (c) aware of its direction
when it appeared in a pond.5 As for “the EXAMINERS and magnitude, and (d) capable of cor-
dress,” one study found that people who Even absent extraneous information, the recting it.16 As for awareness, a global
self-identiied as “night owls” more oten process by which examiners compare survey found that 71% of forensic exam-
saw it as blue and black, whereas “morn- unknown and known faces may prompt iners recognized cognitive bias as a
ing people” tended to see it as white and bias. In psychology, an order efect occurs problem for the forensic sciences as a
gold.6 when the same information presented in whole, but only 52% saw it as a prob-
Similarly, a 2009 study showed how a diferent order elicits a diferent judg- lem for their own discipline, and only
extraneous case information can bias ment. In one classic study, participants 26% believed that it could afect their
judgments of facial similarity in foren- evaluated a person who was described as own judgments.17 his pattern relects a
sic settings.7 Participants compared a “intelligent, impulsive, critical, stubborn, phenomenon called the bias blind spot,
computer-generated facial composite and envious” more favorably than other wherein people recognize others’ biases
of a perpetrator against a lineup of four participants who read the same traits in but not their own, and thus see no need
suspects, and they were told that one of reverse order.12 hey were also more to change their behavior.18
the suspects (chosen at random) had likely to interpret “critical” as a positive In that same survey, 71% of forensic
previously been identiied by two eyewit- (i.e., inquisitive) rather than negative examiners believed that bias can simply
nesses. When asked to rate the similarity (i.e., disparaging) trait. be willed away and 32% felt that experi-
of these faces, participants rated which- he sequencing of information can ence protects against bias19—but research
ever suspect was ostensibly identiied by likewise afect the rigor of forensic analy- suggests otherwise. With respect to the
the eyewitnesses as looking most simi- ses. In a 2011 study, ingerprint experts former, examiners must understand that
lar to the composite, and, accordingly, were asked to count the number of cognitive bias is a natural byproduct of
they believed most strongly in that sus- minutiae (i.e., distinctive features) in how the brain works; it is not a sign of
pect’s guilt. several latent prints, each of which was incompetence, carelessness, or dishon-
hese indings are not a luke; there shown either alone or with a poten- esty. However, one new study suggests
is now a wealth of research across many tial match.13 he results showed that that examiners may be able to self-reg-
disciplines showing that knowledge experts actually counted more minu- ulate bias to some degree by explicitly
of extraneous information can dis- tiae in the same latent prints when they considering alternative hypotheses.20 In
tort forensic examiners’ judgments.8 were shown alone. According to the that study, doctors who were given two
As a result, the National Commis- researchers, the presence of a potential possible explanations for the same injury
sion on Forensic Science has urged match led experts to selectively look for (e.g., assault with a weapon versus dog

bite) were less certain in their interpreta- prints. his revealed a bias; experts spent thus provides a means of proiciency
tions and less inluenced by extraneous more time analyzing whichever print testing as well.25
contextual information than doctors happened to appear at the top of the list In cases where no suspect has been
who considered only one possibility. and were more likely to select that print identiied, forensic facial examiners
With respect to experience, some as a match to the latent print, regardless should be very cautious about execut-
have argued that experts are actually of whether it actually matched. ing unconstrained database searches. In
more susceptible to bias insofar as they Forensic facial examiners also use such situations, every person who the
rely more on “cognitive shortcuts.”21 database searches to identify poten- algorithm identiies as a potential match
These shortcuts develop over time tial matches to an unknown face, but to the unknown face efectively becomes
and are generally helpful because they the inal judgment is made by a human a suspect. his practice in essence pro-
allow experts to process information who may likewise favor faces that the duces an “all-suspect lineup,” which is
more eiciently, but they can also limit sotware ranks more highly. Moreover, known to dramatically increase mis-
experts’ cognitive lexibility and cause the limitations of the algorithms that taken eyewitness identiications and is
them to miss crucial information. For underlie these searches are well-known, strongly discouraged by researchers in
example, chess experts are much better especially their relative inaccuracy for that area.26
than novices at remembering realistic women and people of color, which may
arrangements of pieces on a chessboard, afect who appears as a potential match PREVENTING COGNITIVE BIAS
but they perform no better—or even or is included in the database to begin IN FORENSIC DECISIONS WITH
worse—than novices if the pieces are with.23 Finally, these databases may also PROCEDURAL INTERVENTION
arranged at random. disclose biasing information (e.g., crim- To sum up, cognitive bias is an inher-
inal history) that taints the examiner’s ent feature of human cognition that can
INHERENT BIAS IN FORENSIC judgment. impede the search for truth by leading
SCIENCE TECHNOLOGY people to diferent interpretations of
Forensic science’s growing reliance on LESSONS FOR FORENSIC FACIAL the same information. Because it can-
technology is surely a positive trend as EXAMINERS FROM EYEWITNESS not be eliminated by willpower alone,
it encourages standardization and objec- IDENTIFICATION preventing cognitive bias in forensic
tivity. But does it eliminate bias? Take There are several parallels between decisions requires procedural interven-
ingerprint identiication, for example. forensic facial identiication and eyewit- tion. Research suggests that examiners,
Many ingerprint examiners can now ness identiication that suggest avenues including those who search large digi-
enter an unknown latent print into a for reform. For instance, eyewitness tal databases for potential matches to
digital database (e.g., AFIS) and quickly identification procedures customar- unknown samples (e.g., forensic facial
receive a rank-ordered list of potential ily embed the suspect’s photo in an examiners), should:
matches, which is extremely useful. array of known-innocent “iller” pho- • Have a case manager or super-
However, the inal decision is still made tos because showing only the suspect’s visor identify and redact any
by a human examiner with an imperfect photo is known to increase the risk of extraneous and potentially bias-
and bias-capable brain. misidentiication.24 Similarly, if a suspect ing information (e.g., criminal
hese technologies may even create has been identiied for comparison to an history) from the case materials
new biases. In a 2012 study, research- unknown face, the suspect’s face could prior to analysis.
ers entered latent prints into AFIS to be embedded among several similar but • Analyze and document the dis-
generate rank-ordered lists of potential known-innocent faces, which a database tinctive features of the unknown
matches.22 hen, they asked ingerprint search could generate. his procedure— sample (e.g., face) on its own
experts to decide which (if any) of those called an “evidence lineup”—also creates before making any side-by-side
prints matched the latent print—but the opportunity for the examiner to comparison(s) to known samples.
without the experts’ knowledge, they commit a known error (i.e., a misiden- • If no suspect has been identiied,
had randomized the rank order of those tiication of a known-innocent face) and have a colleague randomize the

order of database search results 3. Mukul Bhalla & Dennis R. Proitt, A General Approach for Improving Decision
prior to analysis and retain the full Visual-Motor Recalibration in Geographical Making as Well as Minimizing Bias, 3 Foren-
list of search results for purposes of Slant Perception. 25 J. Experimental Psych.: sic Sci. Int’l: Synergy 100161 (2021).
discovery. Hum. Perception & Performance 1076 15. A. Towler et al., Diagnostic Feature
• If a suspect has been identiied, (1999). Training Improves Face Matching Accuracy,
embed their sample in a lineup of 4. Alice Towler et al., Diverse Types of 47 J. Experimental Psych.: Learning,
similar but known-innocent iller Expertise in Facial Recognition, 13 Sci. Rep. Memory & Cognition 1288 (2021).
samples from the database. 11396 (2023); P. Jonathon Phillips et al., Face 16. Timothy D. Wilson & Nancy Brekke,
• Document which materials they Recognition Accuracy of Forensic Examin- Mental Contamination and Mental Correc-
reviewed and in what order, when ers, Superrecognizers, and Face Recognition tion: Unwanted Inluences on Judgments and
and how (if at all) their opinion Algorithms, 115 Proc. Nat’l Acad. Sci. 6171 Evaluations, 116 Psych. Bull. 117 (1994).
changed over time, and any mea- (2018). 17. Jef Kukucka et al., Cognitive Bias and
sures they took to protect against 5. T. C. Kietzmann & P. König, Efects of Blindness: A Global Survey of Forensic Science
bias.27 Contextual Information and Stimulus Ambi- Examiners, 6 J. Applied Rsch. in Memory
• Have their opinion independently guity on Overt Visual Sampling Behavior, 110 & Cognition 452 (2017).
corroborated by a qualiied col- Vision Rsch. (Pt. A) 76 (2015). 18. Emily Pronin, Daniel Y. Lin & Lee
league who examines the same 6. Pascal Wallisch, Illumination Assump- Ross, he Bias Blind Spot: Perceptions of Bias
case materials but is kept unaware tions Account for Individual Diferences in in Self versus Others, 28 Personality & Soc.
of their opinion (i.e., blind the Perceptual Interpretation of a Profoundly Psych. Bull. 369 (2002).
veriication). Ambiguous Stimulus in the Color Domain: 19. Kukucka et al., supra note 17.
Implementing such protections “he Dress,” 17 J. Vision, no. 4, June 2017, 20. Moa Lidén, et al., he Role of Alterna-
against cognitive bias in the forensic sci- at 1. tive Hypotheses in Reducing Bias in Forensic
ences will reap extensive beneits. For 7. Steve D. Charman, Amy Hyman Medical Experts’ Decision Making, 63 Sci. &
defendants, it will reduce the risk of Gregory & Marianna Carlucci, Exploring the Just. 581 (2023).
wrongful conviction—as well as the con- Diagnostic Utility of Facial Composites: Beliefs 21. Itiel E. Dror, he Paradox of Human
comitant risk to public safety when the of Guilt Can Bias Perceived Similarity Between Expertise: Why Experts Get It Wrong, in The
actual perpetrators are let free to reof- Composite and Suspect, 15 J. Experimental Paradoxical Brain 177 (N. Kapur ed.,
fend. For practitioners, it will increase Psych.: Applied 76 (2009). 2011).
judicial trust in their decisions by inspir- 8. Jef Kukucka & Itiel E. Dror, Human 22. Itiel E. Dror et al., The Impact of
ing conidence that those decisions stem Factors in Forensic Science: Psychological Human-Technology Cooperation and Dis-
from their unique skill and not from out- Causes of Bias and Error, in The Oxford tributed Cognition in Forensic Science: Biasing
side inluences. And for fact-inders who Handbook of Psychology and Law 621 Efects of AFIS Contextual Information on
tend to trust those decisions indiscrimi- (D. DeMatteo & K. C. Scherr eds., 2023). Human Experts, 57 J. Forensic Sci. 343
nately, these simple and research-based 9. Nat’l Comm’n on Forensic Sci., (2012).
measures will elevate the “science” in Views of the Commission: Ensuring 23. Joy Buolamwini & Timnit Gebru,
forensic science. That Forensic Analysis Is Based upon Gender Shades: Intersectional Accuracy Dis-
Task-Relevant Information (Dec. 8, parities in Commercial Gender Classiication,
Devon E. LaBat is a doctoral candidate 2015), https://www.justice.gov/archives/ncfs/ 81 Proc. Mach. Learning Rsch. 77 (2018);
in legal psychology at Florida ile/818196/download. Sidney Perkowitz, he Bias in the Machine:
International University in Miami, Florida. 10. Brett O. Gardner et al., Do Evidence Facial Recognition Technology and Racial Dis-
Submission Forms Expose Latent Print parities, MIT Case Stud. in Soc. & Ethical
Jef Kukucka, PhD, is an associate Examiners to Task-Irrelevant Information?, Resp. of Computing, Winter 2021, https://
professor of Psychology at Towson 297 Forensic Sci. Int’l 236 (2019). doi.org10.21428/2c646de5.62272586.
University in Towson, Maryland, and 11. Erwin J.A.T. Mattijssen et al., Cogni- 24. Gary L. Wells et al., Policy and Pro-
co-vice chair of the OSAC Human tive Biases in the Peer Review of Bullet and cedure Recommendations for the Collection
Factors Task Group. Cartridge Case Comparison Casework: A Field and Preservation of Eyewitness Identiication
Study, 60 Sci. & Just. 337 (2020). Evidence, 44 Law & Hum. Behav. 3 (2020).
ENDNOTES 12. S. E. Asch, Forming Impressions of 25. Jef Kukucka et al., he Impact of Evi-
1. Paola Bressan & Maria F. Dal Martello, Personality, 41 J. Abnormal & Soc. Psych. dence Lineups on Fingerprint Expert Decisions,
Talis Pater, Talis Filius: Perceived Resem- 258 (1946). 34 Applied Cognitive Psych. 1143 (2020).
blance and the Belief in Genetic Relatedness, 13. Itiel E. Dror et al., Cognitive Issues in 26. Wells et al., supra note 24.
13 Psych. Sci. 213 (2002). Fingerprint Analysis: Inter- and Intra-expert 27. Adele Quigley-McBride et al., A
2. See The Dress, Wikipedia (Nov. 4, Consistency and the Efect of a “Target” Com- Practical Tool for Information Management
2023), https://en.wikipedia.org/wiki/ parison, 208 Forensic Sci. Int’l 10 (2011). in Forensic Decisions: Using Linear Sequential
the_dress. 14. Itiel E. Dror & Jef Kukucka, Linear Unmasking-Expanded (LSU-E) in Casework, 4
Sequential Unmasking–Expanded (LSU-E): Forensic Sci. Int’l: Synergy 100216 (2022).

P
olice body-worn cameras (BWCs)
embody hope and promise for the
objective preservation of criminal
procedure evidence and an enhanced
means for police to provide transparency
and accountability regarding their activ-
ities and practices. Yet, BWCs can only
operate to capture phenomena within
optical and audio recording limits of
the camera itself. his is a developing
capability, for which standards for use
are lexible. In the “real world,” BWC
value can be limited by operator fac-
tors, departmental regulation (or lack
thereof), and the same human observer
efects and cognitive biases of any other
visual evidence. With the advent of new
technologies that can be incorporated
into BWC hardware and systems—
facial recognition and AI scrubbing, to
name two—a critical analysis of the use
of BWC evidence, its limitations, and its
foibles is necessary.
here are enough aspects concern-
ing the use of BWCs to make this article
a virtual drop in a bucket. I hope to
introduce some of the numerous issues
Police Body
that are involved in police departments’
use of BWCs and suggest a diversity of
resources on the subject as well.
POLICE BODY-WORN CAMERA
Cameras: BASICS
BWCs are a component of the chang-
ing face of “evidence verité.”1 First used
in Rialto, California, in 2012 and soon
Analog ater in New York City,2 BWCs are now

ubiquitous in the United States. BWCs
are small enough to go where oicers
go, and, if activated, should record what
Thinking about
transpires in oicers’ presence. Rialto’s
small pilot program correlated with an
88% decline in complaints against oi-
cers and a 60% reduction in use-of-force
incidents.3 BWCs have been occasion-
a Digital ally resisted by police oicers and their

unions for the potential to record non-
incident-related behaviors.4
Early on, BWCs were touted to be
Question part of “best practices” in policing.5 he

Department of Justice (DOJ) stated,
“Body-worn cameras hold tremendous
promise for enhancing transparency, pro-
By the Hon. Roderick Kennedy (Ret.) moting accountability, and advancing
public safety for law enforcement oi-
cers and the communities they serve.”6

It suggested that BWCs are “thought to on Evidence. Introducing photogra- evidence beyond the “normal view-
improve police conduct and transpar- phy, ilm, and video in court as evidence point” of a human observer, involving
ency, especially regarding police use of required recalibration by courts of the similar cognitive foibles. Courts and
force.”7 DOJ undertook a grant program impact of how to present potentially lawyers must guard against the poten-
to assist local/tribal law enforcement to graphic and disturbing evidence ver- tial for cognitive mischief in the minds
purchase and deploy BWCs.8 By 2018, ité to lay fact inders. With video from of fact inders at trial.
about 80% of large departments had citizen cell phones and BWCs, digital
BWCs.9 More than 80% had policies for evidence has caused great acceleration EVALUATING FACT FINDERS’
their use, but the policies are not stan- in both the capabilities and dangers of ACCURACY AMID COGNITIVE
dardized, nor is their enforcement. contemporarily recorded images com- BIASES REGARDING VIDEO
A BJA 2022 practice evaluation ing to court in the twenty-irst century. Intuitively, a viewer can believe what
mentions BWCs’ ability to capture an In 2016, Mary Fan pointed to courts’ they see on video is an objective record
“objective record” of events that does not relying on testimony to ind the facts of of events. he viewer has no more acuity
force “agencies to rely solely on written a case and being a step away from the than a dirty lens and can only observe
reports from oicers or accounts from circumstances as to what happened for for the length of the recording of a
citizens . . . potentially encouraging events they adjudicate, to a time where particular incident, all with no more
mutual accountability during encoun- “the staple events of criminal procedure perspective than the camera afords.
ters.”10 hese are valid considerations, law will be recorded.”13 he 2014 death Since events exist in a 360-degree
although skepticism of both the objec- of Eric Garner in New York at the hands world, lawyers must take care to recog-
tivity of the visual record, its value to of police oicers and the 2020 murder nize, identify, and explain to fact inders
objectivity and accountability, and how of George Floyd in Minneapolis were the different meanings that can be
jurors use the evidence persist. he qual- both vividly recorded by bystanders, found in this visual evidence. his can
ity of policing, especially as it afects uploaded on social media, broadcast be as much circumstantial as visual. For
racial groups, revealed by the volume of by news media, and used in court. he instance, what is portrayed in a video
videos of police interactions entering the impact on both judicial proceedings might be an interaction with a person
public forum, has rightly been called into and public opinion is substantial as with a mental disorder that is not rec-
serious question. he evidence was oten video of police/public interactions are ognized by the oicer, who is facing a
clear; its release to the public by police both aired to the public and brought to verbal onslaught and passive resistance
agencies varies from a trickle to a cat- court as evidence. that could be explained by the behav-
aract, and, thus, its efect to encourage Mary Fan advocated for a reexam- ioral problem. Without understanding
police accountability and thus provide ination of court rules to protect the the circumstantial environment, the
a moderating efect on police/citizen integrity of the audiovisual record when fact inder may be no better at assess-
encounters is uncertain. Studies have ofered as evidence, both to present a ing the evidence than the oicer.
indicated a decrease in use of force and complete and contextualized picture Fact inders may place overreliance
quality of police/citizen interaction, with and to actively provide a legal stan- on video evidence owing to its imme-
citizens reciprocating; in another, BWC dard to protect against unquestioned diacy of presentation, its irst-person
use correlated with an increase in police assumptions on the part of fact inders appearance and image salience com-
use of force and arrests.11 that video presents unmediated truth. pared to written or verbal testimony,
For police, use of BWCs carries he basis: here is a diference between their preexisting biases toward video as
increased costs to departments of acquir- legally relevant truth and the depiction “reality,” or its emotional content, com-
ing sotware, developing new procedures of it on video. Attention to “perceptual pounding the “eyewitness” efect with
and policies for oicers, increasing data yardsticks” and providing interpretive credulity as to the nature of video evi-
storage, and hiring personnel to handle tools for fact inders is a critical neces- dence itself.
redaction, editing, retention, and review sity—so is a detailed understanding of Neal Feigenson and Christina Spie-
of footage. Prosecutors and defense the limitations of the BWC medium. sel’s work15 on the psychology of video
attorneys must have access to review, cat- he ability to replay in court videos of evidence highlights common practices
egorize, and notate footage, adding cost things that witnesses previously only and problems with presenting unstaged,
to the justice system.12 BWC programs, described brought a revolution in the unedited footage of legally relevant
while promising, are still a police instru- capabilities of presenting evidence, but events captured by a range of recording
ment that is in lux. also the need for regulation and skepti- devices from BWCs and surveillance
cism concerning its power.14 Why is this cameras to citizens’ cell phones.
VIDEO EVIDENCE COMES TO supposedly “objective” source of evi- Combined with assumptions from day-
COURT dence drawing caveats and skepticism? to-day experience from the news, social
he impact of witnesses on trials has been BWCs are increasingly employed media, and other sources, jurors may
studied since the nineteenth century, as in the ield, while their capabilities are come to think and expect that video
evidenced in the venerable Wigmore expanding to capture visual and other is likely to be reliable, probative, and

persuasive, even while their perceptions Oicer’s activation of cameras prior to their on-camera actions and words,
and interpretations of this evidence are engaging in interactions is of paramount for example, by yelling for a subject to
beholden to cognitive biases of which importance. Failure in this regard is crit- quit resisting in the absence of resis-
they are not aware. hese can afect how ical. BWCs have a bufer that records a tance. Oral statements made by police
jurors see, hear, and construe the evi- period of time (about 30 seconds) prior that are recorded on the video create
dence. But their views also can change to actual activation, without audio, but new problems regarding hearsay and
in the course of trial.16 he ubiquity of oicers oten fail to capture a complete confrontation rights in evidence law.25
raw BWC footage released in public by incident, losing important evidence. Behavioral changes by oicers owing to
police agencies may exacerbate a ten- Similarly, departments have problems being on camera are noted. Oicers can
dency to desensitize the public to the with oicers placing their cameras on be prone to play to the camera or narrate
nuance of legal and constitutional limi- their equipment belts suiciently low the video as it unfolds to supplement
tations on police behavior, and perhaps as to obscure the ield of view. And, the images of the unfolding action. his
operate to dehumanize the subjects of inevitably, despite the high hopes and attempt to inluence a viewer’s percep-
police interactions owing to a process goals for cameras attached to oicers tion of the action is worthy of skeptical
of seeing a never-ending panoply of producing objective and trustworthy judicial scrutiny, particularly given the
worst behaviors without critical con- primary evidence, police oicers have credulity with which the video evidence
text by which to evaluate them. been called to task for fabricating video can be received. How video inluences
Viewers of BWC video thus share evidence of drug discoveries.23 However, oicer narratives arises in the context of
many characteristics of eyewitnesses, more prosaic manipulation of police report writing as well.26
and the same problems regarding ability video is not uncommon.
to perceive and the existence of factors Here is a strong example of the limi- OFFICER ACCESS TO BODY-
afecting what is “remembered” assume tation of police video: In 2014, deputies WORN CAMERA VIDEO AND
similar weight when applied to video were pursuing Derrick Price from the STANDARDS FOR REPORTING
from BWCs.17 Observers of evidentiary scene of a SWAT activation. he close- USE OF FORCE
video are prone to “naïve realism,” a in police body camera footage depicted Police officers write reports, which
belief that the video shows them real- deputies attempting to restrain Price, are oten the irst thing the prosecu-
ity as it was.18 Given that many police who was face down on a sidewalk. tor and defense attorney see. In most
videos depict emotionally charged sit- Oicers were yelling for him to “stop police departments, oicers are gen-
uations on both sides, the emotional resisting” while attempting to hand- erally allowed, if not encouraged, to
impact can increase observers’ vulner- cuf him. Much of the video faced away review their video in the course of writ-
ability to many biases that foreclose from the action. However, a surveil- ing their reports. he ability to conform
recognizing their subjective involve- lance video from the building in front of recollections to a virtual playback of the
ment or fragmentation of perspective which the sidewalk was located showed incident requires evaluation. For use-
or enhance a tendency toward identi- the entire incident: Price running from of-force (UOF) purposes, the extent
ication bias.19 Jurors are as susceptible deputies who were chasing him and, to which access to data that they did
to conirmation bias—using evidence to prior to their catching up, putting up not possess during the stress and dis-
reinforce prior belief—as any human. In his hands and passively laying on the tractions of an incident is later made
this reality, a viewer may become over- sidewalk before the deputies proceeded available to restructure their narrative
conident in their beliefs and judgments to kick and punch him.24 Was the police is an issue.
they derive from the video—prompt- video manipulated to face away from In the context of BWCs as being vital
ing them to be less inclined to look for much of the beating? Where were videos evaluative tools in UOF investigations,
alternative explanations or be aware of from the other deputies? N.B.: he video the existence of departmental and legal
how their conclusions are shaped by also was not produced by the Sherif ’s standards regarding what is reasonable
preconceptions.20 Department for sixteen months ater or excessive force under the Consti-
As video conveys both less and more the incident, a phenomenon regarded tution governs evaluations of UOF.
than appearances, lawyers need to take as inimical to promoting departmental Evaluating excessive force is judged
control of multiple meanings and edu- transparency and public trust. by legal standards like Graham v. Con-
cate judges and juries in ways that best As vividly seen in the Price video, the nor,27 requiring assessment of oicer
serve their theories of the case.21 video sufered from late activation and perceptions of the need to use neces-
a limited view, showing only the violent sary, minimal, and proportional force
BODY-WORN CAMERA VIDEOS: motion by body parts of oicers who to counteract an immediate threat at
EFFECTS OF POLICE POLICIES were yelling at Price to “quit resisting!” the time of the incident. he objective
AND PROCEDURES he same happens in audio content of reasonableness of an oicer’s subjec-
Using BWCs is an aspect of police prac- BWC recordings in the author’s experi- tive perceptions of an immediate deadly
tice that must include comprehensive ence. Police oicers commonly attempt threat at the moment force is used is criti-
regulation by policies and procedures.22 to structure the narrative of a video in cal, not the arc of circumstance recorded

in real time by a camera. Under Gra- searches or seizures under the First and as images of persons not relevant
ham, “hindsight” is not relevant to Fourth Amendments abound. Tradi- to the officer’s immediate activi-
judging the reasonableness of the force tionally, the use of cameras that mimic ties are flagged for other reasons by
used. This requires assessing unfil- the extent to which human vision can facial recognition technology—ran-
tered recollections to evaluate oicers’ operate has been free of strong restric- dom persons recorded anywhere may
actions and conduct at the time they tions on their use and, to a fair degree, have their data stored and recalled
used force, under the totality of those treated with some deference by the for other purposes later. The abil-
circumstances. courts as presenting an image worthy of ity to identify persons involved with
he availability of BWC video to oi- prima facie accuracy. Scott v. Harris rep- First Amendment–protected activi-
cers creating their written reports and resents a signiicant example.30 However, ties for later investigation can chill
the timing of oicer access to them are intrusion into areas subject to “rea- the right to associate and expose pro-
important concerns. he oicer pre- sonable expectations of privacy” may testers to persistent surveillance and
pares an incident report at the end of expand by BWCs’ ability to collect and identification in police databases. AI
shit and frequently prepares a sepa- visualize data outside the visible spec- video analytics is an algorithm-driven
rate report dealing speciically with trum, as in the Kyllo case,31 or collateral process that can isolate persons and
the underlying basis for the UOF—a data from the whole BWC ield of view. objects appearing on the video, also
legal purpose relecting the severity of here is nothing to prevent BWCs from identifying clothing, tattoos, and other
the crime involved, the level of threat acquiring and transmitting incidental objects, along with hair color, facial
from the subject, and the oicer’s sub- data to which artiicial intelligence can hair, and skin color, further impact-
jective perception as to whether it was be applied to discern possibly relevant ing the right to associate or be free of
an imminent or immediate threat to information. suspicionless searches.
the oicer personally. Oicers writ- BWCs are not limited to recording The continuing and advancing use
ing reports unaided by access to BWC only images. A BWC also records audio; of BWCs by police can affect a pano-
video are removed from corroborative the microphone is nondirectional in its ply of legal issues and areas of concern
or contradictory “factual” information ability to gather audio or visual evidence from civil rights through criminal pro-
that may be present, which could allow that might be outside the constitutional cedure. There does not seem to be an
for “hindsight”-based justiications and requirement of particularity if what is to aspect of it without some cause for
explanations that are impermissible be searched were to become an issue. concern, from the gathering of the
under some department UOF policies Conversations that are not relevant to data to its storage, retention, and use
and the Graham line of cases.28 Using the investigation but may contain con- by the police agencies that use the
BWCs for report writing should be tent that on later listening shows itself cameras. The disparate landscape of
explained and examined. to be otherwise germane to police pur- regulation of body cameras—despite
Last, it is not uncommon for BWC poses can be recorded. Juvenile victims serious effort from organizations such
video or audio to be redacted by law or suspects and their statements may as Police Executive Research Forum
enforcement agencies prior to release be outside the scope of proper record- and the New York University Policing
outside the organization. Redactions ing and need special treatment. GPS Project—deserves continued atten-
can include conidential information, capability allows police departments to tion. This is particularly so as the
graphic content, and information pro- collect information on locations where capabilities for enhanced data acqui-
tected by privacy provisions.29 there are multiple incidents, or places sition and real-time analysis increase.
particular persons who are not at the
CONSTITUTIONAL time of recording legitimate subjects for Judge Roderick Kennedy is a retired
CONSIDERATIONS INCREASE investigation may frequent. chief judge of the New Mexico Court
AS BODY-WORN CAMERA he ability to interface BWC video of Appeals. He is a retired fellow in
CAPABILITIES INCREASE with facial recognition and AI is also the American Academy of Forensic
Fourth Amendment and privacy con- concerning. First, facial recognition Sciences and a fellow in the Chartered
siderations become important as both systems have a recognized tendency Society of Forensic Sciences in the
the scope and particularity of infor- to perform poorly with the faces of UK. He teaches scientiic evidence
mation gathered by BWCs may exceed women, children, and persons with law at the University of New Mexico
constitutional and other legal standards darker skin tones, leading to over-polic- School of Law.
for evidence gathering. First Amend- ing among vulnerable populations and
ment rights may be chilled by incidental increased danger of misidentiications. ENDNOTES
surveillance that algorithmic searches With the advent of real-time video from 1. Neil R. Feigenson & Christina O.
can pinpoint individuals for proper, BWCs, the possibility of video dragnets Spiesel, The Psychology of Surveillance and
improper, or mistaken purposes. or suspicionless investigations being Sousveillance Video Evidence, in Criminal
Existing models of thinking about undertaken in real time increases as Juries in the 21st Century: Contem-
police body cam video as impacting another Fourth Amendment concern porary Issues, Psychological Science,

and the Law 173 (Cynthia Najdowski & Cameras: Understanding the Factors hat Body Camera Evidence, 87 Fordham L.
Margaret C. Stevenson eds., 2019). Influence Variation in Body-Worn Cam- Rev. 1425, 1427 (2019), https://scholarship.
2. Dina Demetrius & Michael Okwu, era Activation, 40 Just. Q. 315 (2023), law.wm.edu/facpubs/1904.
Meet the First U.S. Police Department to https://www.tandfonline.com/doi/ful 24. See Ethan Magoc, Judge Calls Marion
Deploy Body Cameras, Aljazeera Am. l/10.1080/07418825.2022.2071325; O. County Case a “True Tragedy” as He Sen-
(Dec. 17, 2014), http://america.aljazeera. Nicholas Robertson et al., Body Cameras tences Former Deputies to Prison, WUFT
com/watch/shows/america-tonight/ and Adjudication: Views of Prosecutors (Apr. 20, 2016), https://www.wuft.org/
articles/2014/12/17/body-cams-california. and Public Defenders, Crim. J. Review news/2016/04/20/former-marion-county-
html; Lindsay Whitehurst, Police Body (Oct. 2, 2022), https://journals.sagepub. deputies-sentenced-in-derrick-price-beating.
Camera Ramp up Started a Decade Ago. How com/doi/10.1177/07340168221124458. See 25. Id. passim.
Well Have hey Worked?, Associated Press also Richard E. Myers II, Police-Generated 26. E.g., Craig D. Uchida et al.,
(June 29, 2023), https://apnews.com/article/ Digital Video: Five Key Questions, Multiple Managing Digital Evidence from Body-
police-body-cameras-shooting-914729c87b Audiences, and a Range of Answers, 96 N.C. Worn Cameras: Case Studies in Seven
65c102acc766e713031f63#. L. Rev. 1237 (2018). Sites (Just. & Sec. Strategies, Inc. 2022),
3. Demetrius & Okwu, supra note 3 13. Mary D. Fan, Justice Visualized: https://bwctta.com/sites/default/iles/2022-
(note: results may vary by department). Courts and the Body Camera Revolution. 03/1%20DEM%20Final%20Report%2003%20
4. See, e.g., Sean Smoot, Body-Worn 50 U.C. Davis L. Rev. 897 (2017). 01%2022_CONTENT_0.pdf.
Cameras—Understanding the Union 14. Craig Uchida et al., Police 27. 490 U.S. 386 (1989) (establishing an
Perspective, Body-Worn Camera Train- Accounts of Critical Incidents: A Descrip- objective reasonableness standard by which to
ing & Tech Assistance, http://www. tive and Empirical Assessment, J. Crime judge claims of excessive force used by police
bwctta.com/resources/commentar y/ & Just. (May 30, 2022), https://www. based on the perceptions of the oicer at the
view-commentary-body-worn-cameras- tandfonline.com/doi/abs/10.1080/07356 time force was used, under the totality of
understanding-union-perspective. 48X.2022.2080100. the circumstances existing at that moment;
5. See, e.g., Am. Bar Ass’n, Resolution 15. E.g., Neal Fiegenson & Christina seizures of persons accomplished by exces-
604, ABA Principles on Law Enforcement Spiesel, Law on Display: The Digital sive force violate the Fourth Amendment
Body-Worn Camera Polices (adopted Transformation of Legal Persuasion of the U.S. Constitution’s requirement of
Aug. 2021), https://www.americanbar.org/ and Judgment (2009). reasonableness).
content/dam/aba/directories/policy/annual- 16. See id.; Feigenson & Spiesel, supra 28. Rémi Boivn & Annie Gendron, An
2021/604-annual-2021.pdf; Wash. Legis. note 1. Experimental Study of the Impact of Body-
Off., ACLU, Strengthening CBP with 17. See Siegfried L. Sporer, Editorial, Worn Cameras on Police Report Writing, 18
the Use of Body-Worn Cameras (June The Science of Eyewitness Testimony Has J. Experimental Criminology 747 (2022),
2014), https://www.aclu.org/wp-content/ Come of Age, 7 Psych. Sci. in the Pub. https://link.springer.com/article/10.1007/
uploads/document/14_6_27_aclu_handout_ Int. i (2006), for a two-page history of psy- s11292-021-09469-8 (discussing tension
re_body-worn_cameras_for_cbp_inal.pdf. chology and witness testimony, especially between viewing video and Graham’s eschew-
6. Press Release, Of. of Pub. Afairs, eyewitness testimony. ing “20/20 hindsight”).
Dep’t of Just., Justice Department 18. Ronald J. Coleman, Measuring 29. See Sci. Working Grp. on Digit.
Announces $20 Million in Funding to Police Body Camera Infrastructure, 62 Evidence, SWGDE Video and Audio
Support Body-Worn Camera Pilot Program Santa Clara L. Rev. 273, 289 n.77 (2022). Redaction Guidelines, ver. 2.0 (Sept.
(May 1, 2015), https://www.justice.gov/ 19. Id. at 289. N.B. Coleman’s 291-word 17, 2020), https://drive.google.com/ile/d/1-
opa/pr/justice-department-announces- footnote 77 is a significantly comprehen- yAVF8u-YAUn4YOIDWSaMbH23QzLISF0/
20-million-funding-support-body-worn- sive compendium of research on why there view.
camera-pilot-program. is significant cause for worry over why 30. 550 U.S. 372 (2007).
7. Id. “juries [may] be overly reliant on BWC 31. 533 U.S. 27 (2001). See, generally, Sci.
8. Id. his was part of a total package video or . . . reach improper conclusions Working Grp. on Digit. Evidence, SWGDE
worth more than $200 million. based on it.” Id. at 288. Guidelines for the Use of Infrared
9. Research on Body-Worn Cameras and 20. See Fan, supra note 13 (also dis- Radiation (IR) in Forensic Photog-
Law Enforcement, Nat’l Inst. of Just. (Jan. cussing hindsight and outcome biases and raphy, ver. 1.0 (Sept. 17, 2020), https://
7, 2022), http://tinyurl.com/3837kvxb. observer effects). drive.google.com/ile/d/1loQzzeckoFyX6z
10. Id. 21. Feigenson & Speisel, What Do You BiD2o7B1sVwdtSDK2P/view?usp=sharing;
11. See, e.g., Program Profile: Police See in Video? The Hidden Dimensions of Sci. Working Grp. on Digit. Evidence,
Body-Worn Cameras (Phoenix, Ari- Visual Evidence, ABA Sci. Tech. Sec. SWGDE Guideline for the Use of
zona), Nat’l Inst. of Just. (Dec. 6, Webinar: Video in Legal Decision- Reflected Ultraviolet Radiation (UV)
2021), https://crimesolutions.ojp.gov/ Making (Oct. 14, 2022). in Forensic Photography, ver. 1.0 (Sept.
ratedprograms/730#eo. 22. Katz & Huff, supra note 12. 17, 2020) https://drive.google.com/ile/d/1_
12. See Charles M. Katz & Jessica Huf, 23. See Jeffrey Bellin & Shevarma rfIC3VhsJeRJE09MzoUE-cjmBawrosB/
The Achilles Heel of Police Body-Worn Pemberton, Policing the Admissibility of view?usp=sharing.

International
Verification
Standards for
Open Source
Videos:
Strengthening
Faith in Facts
By Alexa Koenig

S
unlight and a faint breeze loated developing creative methods for ind-
through the windows overlook- ing and verifying the accuracy of online
ing Lake Como as eighteen people videos for close to a decade, the vid-
positioned themselves around a rectan- eos oten were not being collected in
gular set of tables in October 2017. he ways that documented their chain of
lightness of the morning contrasted with custody, or with notes that explained
the weight of the subject matter they how an item might have been modi-
were there to discuss: how to bridge their ied since collection and why.3 Much of
various areas of practice to strengthen what was being shared with courts was
the future of digital investigations and relatively useless for international tri-
improve the ability of international law- als, documenting, say, that a school had
yers to bring the world’s highest-level been destroyed or a civilian killed, but
perpetrators to account. Even more not the contextual cues, such as a street
speciically, they were tasked with deter- sign or surrounding architecture, that
mining whether international guidelines would be critical to locate the incident
were needed to help standardize the rap- in geographic space. hese contextual
idly emerging ield of digital open source data could also be helpful for linking
investigations.1 the incident to an alleged perpetrator,
Each person in the room had been providing lead information by depict-
invited because of their particular per- ing people who could potentially serve
spective on digital evidence, especially as witnesses, or suggesting physical evi-
video evidence, or their expertise in dence that might still be available.4 Few
international criminal legal processes. had yet articulated the legal issues that
One was a journalist who, years earlier, could arise from the mass collection of
had found creative ways to identify evi- digital data, ranging from the diiculty
dence of atrocities on Twitter. Another of meeting disclosure obligations given
had established an archive to preserve how hard it would be to know what was
YouTube footage of possible inter- in a dataset when entire social media
national crimes in Syria. Still others channels were scraped, to identifying
had abundant experience investigat- who might qualify as a witness for pur-
ing atrocities related to international poses of explaining how an open source
crimes committed in the former Yugo- investigation had been conducted.
slavia and Rwanda. NGO representatives One of the irst tasks thrown to those
included experts in training activists present was to clarify a handful of rel-
how to use smartphones to capture foot- evant terms. his seemed important
age of atrocities. Two of those present to ensure everyone was talking about
had previously helped drat international the same thing when they used a par-
protocols to aid the collection of other ticular word or phrase. For example,
types of evidence. One lawyer who had OSINT—an abbreviation for Open
worked with the International Criminal Source Intelligence—had become a com-
Court (ICC) was there because she had mon reference for digital information
painstakingly analyzed how ICC judges pulled from public spaces on the inter-
were approaching digital data, while net. But as several pointed out, much
another—a trial attorney—was there of the information they were interested
to keep the discussion grounded in the in would not be used for intelligence
practical realities of how he would need purposes and thus for decision-mak-
to eventually present any digital evidence ing, but as evidence, or perhaps as lead
in court.2 information. While the organizers had
While many of the participants had anticipated that it would take an hour or
been working with video and other two to clarify deinitions, it took nearly
open source evidence for years, almost eight hours to reach a consensus for just
all agreed that online information was ive terms—proof in itself that standards
being created or collected to varying might be needed to bridge the disparate
standards and thus varying degrees of communities beginning to do this work.
usefulness. Although journalists and By the end of the three-day workshop,
human rights researchers had been the conclusion was clear: A protocol was

needed to (1) help the international jus- Pompeii,7 to the printing press in the nominal fee to access an article online).
tice community communicate more iteenth century, photographs and ilms he deinition excludes information
efectively within and across areas of during the nineteenth century, and that you need special status to access,
practice and (2) strengthen the qual- radio and television in the twentieth such as law enforcement status or some
ity and coordination of information century.8 he digital revolution made kind of subpoena, or that you acquire
increasingly lowing from social media possible by the invention of transistors by illegal means, such as unauthorized
sites towards courts.5 in the 1940s ushered in a new era: By hacking.10 Information that qualiies can
hree years later, in 2020, the United the end of the twentieth century, inter- take a wide array of forms, from vid-
Nations Oice of the High Commis- net use was ubiquitous and phones with eos uploaded by members of the public,
sioner for Human Rights and the cameras (irst released in Japan in 1999 to the audio of police scanners, photo-
Human Rights Center at the Univer- and internationally popularized with graphs shared by major media, PDFs of
sity of California, Berkeley, launched the launch of Apple’s iPhone in 2007) oicial reports issued by human rights
the resulting Berkeley Protocol on Digital enabled the capture and transmission organizations, and government data-
Open Source Investigations.6 Originally, of visual information by anyone with bases and spreadsheets.
the launch was to have taken place in access to that technology. During the In addition to providing key deini-
Nuremberg’s historic courtroom 600, irst two decades of the twenty-irst tions, the Berkeley Protocol establishes
but those plans had been derailed by century, the scale of digital data online minimum standards for ethically and
the onset of a global pandemic. he exploded. As just two examples, by 2022 efectively inding and evaluating open
shit to Zoom, however, proved fortu- more than 500 hours of video were source data. It outlines professional
itous, allowing hundreds of additional being uploaded to YouTube every min- principles (the competencies investi-
people to be present as then-High Com- ute, while 34 million videos were being gators should have to conduct online
missioner for Human Rights Michele uploaded to TikTok each day. research, at a minimum), methodolog-
Bachalet announced the protocol’s In the early 2010s, journalists, gam- ical principles (which focus on how the
launch in its advance English version. ers, activists, and others all began work should be done; for example, by
In 2024, the Berkeley Protocol becomes jumping on the justice bandwagon.9 always engaging in a three-step veri-
oicial, when inally released in all of Alerted to videos of grave crimes and ication process as described below),
the languages of the UN. abuses posted to social media in rela- and ethical principles that outline how
In this article, I briefly describe tion to the Arab Spring, they invented to responsibly do the work.11
those guidelines and others that have new methods for mining the internet he protocol was drated to be tool
been developed to shape the handling for facts about those events and sharing agnostic, meaning that no speciic plat-
of digital open source evidence in an that information with legal investigators forms or tools are referenced. his was
international context. I also summarize or the public. hese collective eforts done to “future proof ” the protocol,
new international veriication standards would launch an increased awareness given the relative speed with which
for such information, with an emphasis of the potential of digital open source digital tools rise and fall out of favor, or
on videos. information to strengthen the eviden- become otherwise obsolete.
tiary foundations of cases. Methodologically, the protocol also
INTERNATIONAL GUIDELINES According to the Berkeley Proto- explains how digital investigations can
FOR OPEN SOURCE DATA col, digital open source information is be incorporated into traditional docu-
Each new communication technology information that is publicly accessible mentation and investigation life cycles.
has allowed for information to low on the internet, by either observation Perhaps most important are the pre-
farther and faster from its source. Such (e.g., you conduct a search and see what investigative steps, first of which is
technologies have ranged from wax results), request (e.g., you submit your conducting a risk assessment—ideally
tablets used to transmit news from email address to get access to a website), one that is holistic in that it considers
household to household in ancient or purchase (e.g., you pay a relatively the digital, physical, and psychosocial

risks that come with online research— INTERNATIONAL VERIFICATION accents or patterns of speech consistent
and how to put a plan in place to STANDARDS FOR OPEN SOURCE with that region and the alleged identi-
mitigate those risks. Second is a digital DATA ties of the people depicted? What was
landscape assessment, which includes he Berkeley Protocol recommends tak- the weather that day in that part of the
identifying who has access to the inter- ing a three-step approach to verifying world? Are people dressed consistent
net and how they’re communicating all digital open source data, including with that weather?
online, as well as whose perspectives video.19 he irst objective is to assess Content analysis is where the major-
may be missing. Finally, online investi- any technical information affiliated ity of open source veriication methods
gation planning includes what protocols with the item. A technical analysis takes are aimed but also can be the most
will be deployed, what tools will be used, advantage of the fact that metadata is uncertain and/or dangerous step in the
and who will do what and when.12 he embedded in every digital item at the veriication process, given its depen-
protocol provides a series of templates time of its creation—whether that item dence on visual comparisons.22 Social
to help investigators map all of this, such is a photo, video, Word document, PDF, science research has repeatedly shown
as templates for assessing the risks and spreadsheet, or otherwise—and oten that people’s ability to visually ana-
opportunities of using digital tools, doc- contains critical information about lyze information and their conidence
umenting key information identiied where, when, and via which device that that they have accurately assessed that
during the digital landscape analysis, item was created. Several tools are avail- information do not perfectly correlate:
and more.13 able to easily and quickly assess whether People are oten less accurate than they
Finally, the protocol touches on pub- metadata is still attached.20 think.23 A few diferent tactics can help
lication possibilities. he open source If the metadata has been stripped minimize risk. For example, people can
investigative era has been marked by a (for example, if the video was posted develop multiple working hypotheses
wave of creativity in how to communi- to a social media site), the investigator about the item they’re investigating and
cate digital data to various audiences. can search for metadata in the contex- then test those hypotheses against the
he largest potential audience is the tual information around the item (for facts they know and the other evidence
general public: Visual explainers, tweet example, a date or timestamp from the they’ve collected. hey can also use peer
threads, and story maps have all become social media platform to which it has review to try to ofset any biases or mis-
popular outputs to amplify investigative been posted, or details provided in the takes. Finally, it is important to have
indings. A more targeted audience is comments, tags, or text accompanying some sort of protocol for avoiding com-
composed of judges. For example, SITU the item). Of course, these technical mon visual traps, such as overanalyzing
Research, “an unconventional architec- data are not irrefutable; metadata can lighter parts of an image, information in
ture practice” based in New York City,14 be spoofed and contextual information the foreground, or faces.24
created a digital platform to help judges can be deliberately or unintentionally he third step is source analysis:
navigate the visual evidence (both misleading. But these technical data can establishing who irst posted the video
closed and open source) for a case at provide a possible hypothesis for where and whether they are reliable for the
the ICC.15 Charges in that case, Prosecu- and when a video was created or can be information that has been shared. A
tor v. Al Mahdi,16 centered on the alleged triangulated with other information as related issue is identifying whether the
destruction of cultural heritage prop- a check on its veracity. source is human or machine generated;
erty in Timbuktu, Mali. By clicking on he second step is content analysis: today, a signiicant portion of online
the name of one of the nine sites where looking within the item’s “four cor- activity is perpetrated by automated
buildings had been destroyed, judges ners.”21 Is what you’ve been told about a accounts known as “bots.”25
could pull up all of the satellite imag- video consistent with what you can see? All three strategies—technical, con-
ery, photographs, and videos relevant Methods for testing the content include tent, and source analysis—are critical to
to that location to better understand processes like reverse image searching ofset human and machine biases,26 as
what had happened.17 hey could also stills pulled from a video to see if an discussed in this issue by Devon LaBat
pull up satellite imagery that revealed image has appeared on the internet pre- and Jef Kukucka,27 and to check for
how the nine locations were positioned viously. Another would be geolocation. misinformation and disinformation,
in relation to each other. For a later case For example, if the video is allegedly including inauthentic digital data, as
involving international crimes in Tim- of events that took place in San Fran- discussed by Raquel Vazquez Llorente.28
buktu, Prosecutor v. Al Hassan,18 SITU cisco, California, does satellite imagery Once an item has been thoroughly
Research took the visual aids even fur- or drone footage or other videos or pho- analyzed—and ideally either veriied
ther, digitally reconstructing the city tos of the same alleged location show or debunked—questions may arise as
and empowering digital “visitors” to similar built or natural markers, like the to admissibility. At the ICC, the admis-
explore the various sites remotely, turn- same pattern of buildings, mountains, sibility standard is relatively loose.29
ing corners to ind (and play) the videos or trees? Is the lora or fauna what you’d Anything that is relevant will likely
of speciic incidents at the sites where expect to see? How about the clothes be admitted. he primary barrier is
they had been captured. people are wearing? If there is audio, are whether the method of obtaining the

item violates a provision of the Rome reliability of evidence acquired through beyond, those in the legal profession
Statute or an internationally recognized open source investigations. Second are in order to strengthen the foundations
human right, that is, whether the viola- guidelines that build of the Murad of justice.
tion throws the reliability of the evidence Code to help open source investigators
into question or admitting it “would be adopt a victim-centered approach to
antithetical to and . . . seriously damage documenting systematic and conlict- Alexa Koenig, JD, PhD, is an
the integrity of the proceedings.”30 related sexual violence that are being adjunct professor at the University
developed by UC Berkeley’s Human of California, Berkeley School of
LEGAL TOOLS FOR OPEN SOURCE Rights Center and the Institute for Law, where she is also co-faculty
INVESTIGATORS International Criminal Investigations, director of the Human Rights Center
Digital technologies have radically with guidance from an expert working and co-founder of the center’s
afected how international investiga- group. he third is a template for cre- Investigations Lab (Berkeley). She is
tions are conducted. his evolution is ating a psychosocial security plan for co-editor of Digital Witness: Using
particularly notable for encouraging online investigators. Open Source Information for Human
cross-institutional and cross-disci- On the horizon are new frontiers for Rights Investigation, Documentation
plinary collaborations.31 Tools like the legal practice that rely heavily on visual and Accountability and co-author of
Berkeley Protocol and the guidelines materials. hese include virtual court- Graphic: Trauma and Meaning in our
following in its wake are designed to rooms, such as one recently held in the Online Lives.
help legal actors become better able to metaverse;32 virtual visits to alleged
communicate with others, but also to crime scenes;33 and the potential to ENDNOTES
help nonlegal actors understand how use augmented reality when physically 1. See Human Rights Ctr., The New
to make sometimes subtle shits in the visiting the site of an atrocity.34 All of Forensics: Using Open Source Informa-
ways they work that can increase the these possibilities raise new opportu- tion to Investigate Grave Crimes (2018),
value of what they produce, collect, or nities and new risks for justice35 and https://humanrights.berkeley.edu/sites/
analyze for court purposes. will require an ever-widening and more default/iles/publications/bellagio_report_
Ultimately, such guidance is criti- diverse community of practice. july2018_inal.pdf.
cal to ensuring that new individuals 2. Id.; see also Alexa Koenig & Lindsay
can be introduced into the fact-ind- CONSISTENCY AND TRAINING IN Freeman, Advancing Human Rights Investiga-
ing ecosystem who have the skills OPEN SOURCE VIDEO EVIDENCE tions Now and for the Future, Medium (Dec. 1,
needed to make this work as efective, Numerous scholars have identified 2020), https://unhumanrights.medium.com/
eicient, and ethical as possible. hese the strengths and vulnerabilities of advancing-human-rights-investigations-now-
may include computer scientists, who using digital open source informa- and-for-the-future-67ae9be1129a.
can help investigators bring the deluge tion—especially video content—as 3. For a summary of international attempts
of digital data down to human scale evidence. This includes a relative to strengthen the forensic value of user videos,
by automating various aspects of the lack of consistency in how courts see Hum. Rts. Ctr., Digital Fingerprints:
review process; specialists in artiicial treat such evidence. 36 Hopefully, Using Electronic Evidence to Advance
intelligence, who can deploy their tal- guidelines, standards, and training Prosecutions at the International
ents to do everything from helping to will strengthen consistency. As law Criminal Court 9 (2014); Hum. Rts. Ctr.,
detect synthetic videos (“deepfakes”) to professor Jennifer Mnookin pointed Beyond Reasonable Doubt: Using Scien-
automatically identifying and blurring out in an interview with media stud- tific Evidence to Advance Prosecutions
graphic material to minimize investi- ies professor Dr. Sandra Ristovska, at the International Criminal Court
gators’ exposure to potentially harmful “law school, legal education, and the (2012).
content; to architects like those at SITU legal profession can be very focused 4. See Beyond Reasonable Doubt, supra
research who are pioneering new ways on words. We should work harder to note 3.
to introduce visual materials into court. ensure that our students gain mean- 5. Andrea Lampros, Bellagio Work-
Now that the Berkeley Protocol is ingful exposure to working with and shop Examines Open Source Information
available to facilitate collaboration understanding both numbers and as Evidence, Medium (Oct. 8, 2017),
and communication across organiza- images, both the interpretive chal- https://medium.com/humanrightscenter/
tions and sectors, several other tools lenges they raise and how to use bellagio-workshop-examines-open-source-
are being developed to further support them as effective tools for advo- information-as-evidence-dad2475fac7d.
such investigations. hree are partic- cacy.” 37 Given the growing diversity 6. Berkeley Protocol on Digital Open
ularly noteworthy. he irst consists of individuals and organizations now Source Investigations: A Practical Guide
of materials for judges that are being involved in the documentation, pres- on the Effective Use of Digital Open
pulled together by the University of ervation, analysis, and presentation Source Information in Investigating
Essex in London with a consortium of open source videos, that train- Violations of International Criminal,
of civil society actors that discuss the ing must include, but also extend Human Rights and Humanitarian Law

(2022) [hereinater Berkeley Protocol], see Sandra Ristovska, When Believing Can Open: Fair Trial Rights and Open Source
https://www.ohchr.org/en/publications/ Be Seeing: The Unregulated Approach to Evidence at the ICC, OpinioJuris (Feb. 10,
policy-and-methodological-publications/ Video Evidence in U.S. Courts and the Need 2023), http://opiniojuris.org/2023/02/10/
berkeley-protocol-digital-open-source. for Archival Legal Standards, FirstMonday symposium-on-fairness-equality-and-
7. See, e.g., Tom Standage, Writing on (June 12, 2023), https://irstmonday.org/ojs/ diversity-in-open-source-investigations-
the Wall: Social Media, the First 2,000 index.php/fm/article/view/13231/11050. out-in-the-open-fair-trial-rights-and-
Years (2013). 23. See, e.g., id.; Image Comparison open-source-evidence-at-the-icc/; Judy
8. See, e.g., Lindsay Freeman, Digital Evidence: The United States Experience, Mionki, Periodismo y OSINT para los
Evidence and War Crimes Prosecutions: he Jonathan Hak (Mar. 17, 2018), https:// DDHH y law persecución de crímines de
Impact of Digital Technologies on International w w w.jonathanhak.com/2018/03/17/ atrocidad: Digital Evidence, Open Source
Criminal Investigations and Trials, 41 Ford- image-comparison-evidence-the-united- Evidence, and International Criminal Law
ham Int’l L.J. 283 (2018). states-experience/; Opposing Experts (conference presentation), YouTube (Sept.
9. See, for example, the pioneering work Diverge on Video Authentication: Case 27, 2023), https://www.youtube.com/
of Bellingcat into the downing of light MH17 Study, Jonathan Hak (Sept. 20, 2023), watch?v=aUOv9URZBOM; see also Free-
over Ukraine—MH17, bellingcat (2023), https://www.jonathanhak.com/2023/09/20/ man, supra note 8.
https://www.bellingcat.com/tag/mh17/—and opposing-experts-diverge-on-video- 31. See, e.g., Joint Investigation Team
on the Al Werfalli case at the International authentication-case-study. into Alleged Core International Crimes in
Criminal Court—Bellingcat Investigation 24. See, e.g., Amy Herman, Visual Ukraine: One Year of International Col-
Team, How a Werfalli Execution Site Was Geo- Intelligence: Sharpen Your Perception, laboration, EuroJust (Mar. 24, 2023),
located, bellingcat (Oct. 3, 2017), https:// Change Your Life (2017). https://www.eurojust.europa.eu/news/
www.bellingcat.com/news/mena/2017/10/03/ 25. For an overview of bot types joint-investigation-team-alleged-core-
how-an-execution-site-was-geolocated. Also and related statistics, see Homeland international-crimes-ukraine-one-year-
see the groundbreaking work of the New York Security, Social Media Bots Over- international; Alexa Koenig, From “Capture
Times Visual Investigations Team, which view (May 2018), https://niccs.cisa. to Courtroom”: Collaboration and the Digital
began releasing its video-based stories in gov/sites/default/files/documents/pdf/ Documentation of International Crimes in
2017 (https://www.nytimes.com/spotlight/ ncsam_socialmediabotsoverview_508. Ukraine, 20 J. Int’l Crim. Just. 829 (2022).
visual-investigations). p d f ? t r a ck D o c s = n c s am _ 32. See Isabel Woodford, Colombia
10. Berkeley Protocol, supra note 6, socialmediabotsoverview_508.pdf. Court Moves to Metaverse to Hold Hear-
at 6. 26. For an overview of biases that com- ing, Reuters (Feb. 24, 2023), https://
11. Id. ch. II. monly affect the use of visual evidence w w w. re ut e r s . c om / w or l d / a m e r i c a s /
12. Id., ch. V; see also id., annex I, at 83. in digital open source investigations, see colombia-court-moves-metaverse-host-
13. Id., annexes I–V, at 83–87. Yvonne McDermott, Alexa Koenig & hearing-2023-02-24.
14. SITU, https://situ.nyc. Daragh Murray, Open Source Information’s 33. See, e.g., Catrin Dath, Crime Scenes
15. For a demonstration of the platform, Blind Spot: Human and Machine Bias in in Virtual Reality: A User Centered Study
see ICC Digital Platform: Timbuktu, Mali, International Criminal Investigations, 19 J. (June 26, 2017) (M.S. thesis, KTH Royal
SITU/Research, https://situ.nyc/research/ Int’l Crim. Just. 85 (2021). Inst. of Tech.), https://kth.diva-portal.org/
projects/icc-digital-platform-timbuktu-mali). 27. Devon E. LaBat & Jef Kukucka, How smash/get/diva2:1115566/FULLTEXT01.
16, Prosecutor v. Ahmad Al Faqi Al Mahdi, Cognitive Bias Can Impair Forensic Facial pdf.
ICC-01/12-01/15. Identiication, 20 SciTech Law., no. 2, Win- 34. See Rafael Golomingi et al., Aug-
17. Id. ter 2024, at 13. mented Reality in Forensics and Forensic
18. Prosecutor v. Al Hassan Ag 28. Raquel Vazquez Llorente, Deepfakes Medicine—Current Status and Future Pros-
Abdoul Aziz Ag Mohamed Ag Mahmoud, in the Dock: Preparing International Justice pects, 63 Sci. Just. 451 (2023).
ICC-01/12-01/18. for Generative AI, 20 SciTech Law., no. 2, 35. See, e.g., Brittan Heller, Watching
19. Berkeley Protocol, supra note 6, Winter 2024, at 28. Androids Dream of Electric Sheep: Immersive
at 62–65. 29. Rome Statute art. 69(3); Lindsay Free- Technology, Biometric Psychography, and the
20. One example is InVid, available at man & Raquel Vazquez Llorente, Finding the Law, 23 Vanderbilt J. Ent. & Tech. L. 1
https://www.invid-project.eu/invid-mobile- Signal in the Noise: International Criminal (2020).
application. Search engines can be used to Evidence and Procedure in the Digital Age,” 36. See, e.g., Alexa Koenig & Lindsay
identify simple instructions for checking 19 J. Int’l Crim. Just. 163 (2021). Freeman, Cutting-Edge Evidence: Strengths
metadata in any digital document, including 30. Rome Statute art. 69(7); see also and Weaknesses of New Digital Investigation
a Word ile, PDF, or spreadsheet. Freeman & Llorente, supra note 29. For Methods in Litigation, 73 Hastings L.J. 1233
21. Berkeley Protocol, supra note 6, additional information on the jurispru- (2022).
at 64–65. dence of the International Criminal Court, 37. Sandra Ristovska, On the Ontology
22. For an excellent summary of the see Sarah Zarmsky & Judy Mionki, Sym- and Epistemology of Visual Legal Evidence:
various cognitive and social factors that may posium on Fairness, Equality, and Diversity Interview with Jennifer L. Mnookin,” 28
afect how people interpret video evidence, in Open Source Investigations: Out in the FirstMonday, no. 7, June 14, 2023.

DEEPFAKES
IN THE DOCK:
Preparing
International
Justice for
Generative AI
By Raquel Vazquez Llorente

T
he history of audiovisual media (AI) tools that allow anyone with an
is the history of audiovisual internet connection to create realis-
manipulation. he irst camera tic synthetic images or audio, are now
came into the world in 1816, although bringing novel challenges to the court-
it was not until 1888 when American room. While the impact of generative
businessman George Eastman started AI and synthetic media in the informa-
marketing a device under the name of tion landscape has been the subject of
“Kodak.” Cameras were commercial- heightened scrutiny since shortly ater
ized more widely in the early twentieth ChatGPT was made publicly avail-
century, but even before the average able in November 2022, the effects
consumer was able to take their own that deepfake technology can have on
photographs, the world had already international criminal justice are yet
seen its irst trial for audiovisual manip- to be discussed with similar intensity.
ulation back in 1869. William Mumler, his article aims to start a deeper, more
a jewelry engraver from Boston, took a informed conversation about how to
selie (or, back then, a “self-portrait”) prepare international courts and tri-
that revealed the shape of his deceased bunals, especially the International
cousin on the image. As the story goes, Criminal Court (ICC or the Court),
initially Mumler shared the ghostly for developments in the ield of digital
photograph with a friend as a joke, but evidence and media synthesis.
seeing the amazement of his colleague,
he thought he could make a lucrative DEMYSTIFYING THE TECH: GANS
business out of “spirit photography” by VS. DIFFUSION MODELS
taking images of people and conjuring Deepfakes, a portmanteau of “deep
their loved ones to appear on cam- learning” and “fake,” have evolved rap-
era. hese manipulations seem to have idly since they irst appeared in 2017 in
been the result of nondivine interven- a Reddit group that was trading con-
tion, showing us an early example of tent of celebrities whose faces had been
double exposure, by which a previous swapped into videos (including porno-
image made its way into another pho- graphic movies). Deepfake technology
tograph that used the same glass plate harnesses AI to create realistic images,
for producing the negative. Ater a few videos, and audio recordings that can
years of cashing in on the grief that otentimes be indistinguishable from
the American Civil War had brought, reality. Some of the most common tech-
he was accused of fraud. During the niques leverage Generative Adversarial
trial, another photographer testiied Networks (GANs) and difusion mod-
as a witness, having produced himself els. GANs involve two neural networks
a fabricated image of a client with the that are pitted against each other. One
“ghost” of Abraham Lincoln to dem- is the generator that creates the exam-
onstrate the manipulation technique. ples; the other one is the discriminator
Mumler was eventually acquitted.1 that evaluates whether they are real or
he Mumler trial not only repre- fake. With each iteration, the generator
sents one of the earliest recorded cases reines the output.2 A GAN trained on
of malicious audiovisual manipula- images of faces can produce new syn-
tion, but it also highlights two powerful thetic faces that look realistic.
dynamics that carry throughout his- On the other hand, difusion mod-
tory. First, many of us are willing to els are a type of deep generative model
place trust in what we see as long as it that add noise to the training data and
aligns with our worldview. Second, even then reconstruct the data by revers-
when demonstrating the ease of forg- ing this process. While GANs are like
ery, it may still be diicult to prove in a legal debate where each side sharp-
court that any editing took place on a ens the other’s skills, difusion models
given image. Advancements in audiovi- are more about taking a broad idea and
sual manipulation, speciically deepfake reining it into something clear and
technology, combined with the com- detailed. hey are remarkably adept
mercialization of artiicial intelligence at generating high-quality images that

can sometimes beat the capabilities of on oicial social media channels of the alleged crimes by the RSF.19 According
GANs. Whereas GANs excel in creating Rapid Support Forces (RSF) or of their to a pre-trial Chamber decision from
faces and expressions, difusion models leader, Mohamed Hamdan Dagalo, who 2021, the ICC has jurisdiction over
can crat detailed and realistic textures is believed to be dead. We consulted Rome Statute crimes in Gaza and the
and patterns, making them a powerful AI media forensic experts, and they all West Bank, including East Jerusalem,
technique to fabricate convincing envi- concluded that there did not seem to be and over crimes committed by Pales-
ronments or contexts. Difusion models indications of AI manipulation—with tinian nationals or the nationals of any
are responsible for much of the prog- the caveat that models do not perform state parties on Israeli territory.20 For
ress over the past two years in the image to the same level of accuracy for non- the past two years, the Oice of the
domain, and they are behind tools now English languages. Neither does this Prosecutor has been “steadily increas-
known to the public, such as Dall-E or mean the absence of other forms of ing the resources and personnel for the
Stable Difusion. manipulation, such as more traditional Palestine investigation.”21
editing. In the recent conlict between Looking at the countries listed
THE IMPACT OF GENERATIVE AI Palestine and Israel, synthetic media above, it is not a question of whether
ON INTERNATIONAL JUSTICE portrays Israeli refugee camps,9 crowds an institution like the ICC will face a
Deepfakes, or claims of AI manipula- marching in support of Israel,10 military deepfake problem, but how badly it
tion, have begun to surface in countries attacks taken from video games,11 and will impact justice. U.S. Ambassador-
embroiled in armed conlicts, experi- children in need.12 at-Large for Global Criminal Justice
encing mass violence, or under the hese are just a few examples of how Beth Van Schaak has highlighted the
thumb of authoritarian regimes. In late AI is creating confusion and mistrust urgency of the situation: “Now we have
2018, a video of President Ali Bongo in what we see or hear. As deepfake the ability to use synthetic media, gen-
of Gabon, released by his oice partly technology grows more accessible, the erative AI, [. . .] and so you can imagine
to counter claims of his ill health, was volume of synthetic media is increasing, creating pieces of evidence that could
dismissed as a deepfake by the oppo- swamping the information ecosystem actually infect a legal process.”22 While
sition, precipitating an attempted and impacting most countries, includ- the proliferation of deepfakes risks false
military coup.3 Experts have not been ing those under the jurisdiction of the information being accepted as true, the
able to conclude unanimously whether ICC. Gabonese oicials and opposition larger threat is that the possibility of AI
the video is the result of AI manipu- members were once under preliminary manipulation is making it easier for the
lation (although it was probably not a examination for crimes against human- public or those in power to dismiss real
deepfake). In Myanmar in 2021, Phyo ity and incitement to genocide.13 he content as fake. If this general mistrust
Min hein, former Chief Minister of situation in Myanmar/Bangladesh is creeps into the courtroom, it may poi-
Yangon Region, confessed on camera to under investigation.14 he prosecu- son the well of all audiovisual digital
having paid gold and cash to Aung San tor of the ICC resumed in 2023 their evidence.
Suu Kyi as part of a corruption scheme. investigation into the situation in Ven- he ascent of deepfakes raises the
he video was likely to have been pro- ezuela (for transparency, I was one of specter of doubt, turning audiovisual
duced under duress rather than by AI the lawyers involved in the communica- content from important probative mate-
intervention.4 tion submitted in 2022 by the Clooney rial into potential red herrings. While
In Venezuela and Burkina Faso, gov- Foundation for Justice and Foro Penal deepfakes could be used to frame inno-
ernments or their sympathizers have to contribute to the the ICC’s investi- cent parties or for creating “evidence” of
misused commercial sotware intended gation).15 While the Court has yet to atrocities that never occurred, looking
to be for company training and product act on the allegations of crimes against back in history and at the jurisprudence
demos, to produce favorable propa- humanity in Mexico, victims’ groups at the ICC,23 synthetic media will most
ganda.5 In Mexico, candidates in the and civil society have been calling for likely have two efects on judicial pro-
2024 electoral cycle have seen them- an investigation into mass disappear- cesses for core international crimes if
selves or their voices deepfaked into ances for nearly a decade.16 he ICC has let unaddressed. First, the burgeoning
videos or audios.6 More notoriously, in issued arrest warrants for war crimes in volume of deepfakes will overwhelm
Ukraine, deepfaked President Zelen- Ukraine against Vladimir Putin, pres- the capacity for analysis of institutions
skyy called the troops to surrender in a ident of the Russian Federation, and in charge of justice and account-
video rapidly debunked by the govern- Maria Lvova-Belova, commissioner for ability. he conlict in Syria already
ment.7 In Sudan, “leaked recordings” Children’s Rights in the Oice of the sounded alarm bells and catalyzed the
of Omar Al Bashir, the former leader President of the Russian Federation.17 application of object recognition and
who has not been seen in public for a Al Bashir was the irst sitting head of frame analysis technology to images
year, were suspected of being manip- state with a warrant of arrest by the ICC circulated mostly online; however, tra-
ulated.8 he organization I work for, and the irst person to be charged by ditional veriication methods are no
WITNESS, has also received tips of the Court for the crime of genocide.18 longer suicient. Second, the leap for-
two audio iles and a video circulated he prosecution is also investigating ward in deepfake technology will lead

to acquittals of those guilty of crimes by “content credentials” that can help with trying to meet, there is a blatant barrier
inserting doubt on authentic evidence. the veriication. he human rights sec- to their efectiveness. hese technology
tor has been a pioneer in piloting apps solutions are ultimately only as strong
FUTURE-PROOFING TRIAGE for human rights defenders document- as their adoption, so they require
AND ANALYSIS AT THE ing abuses that automatically collect intense outreach to put the tools in the
INTERNATIONAL CRIMINAL metadata, for instance, CameraV and hands of those documenting interna-
COURT Proofmode. In the last few years, we tional crimes. Furthermore, without
he ICC prosecutor has the obligation have seen provenance technology gain- robust capacity-building programs
to investigate both incriminating and ing traction in the form of standards that give those behind the camera the
exonerating circumstances equally.24 spearheaded mainly by the Coalition knowledge to capture what is relevant to
Hence, siting through the avalanche of for Content Provenance and Authentic- a case, and those receiving the material
content to identify authentic evidence is ity (C2PA), of which WITNESS is part. the expertise to interpret the metadata,
critical. In an academic article I wrote Most of the solutions available can the potential of provenance technology
with Lindsay Freeman, we examined provide helpful markers to verify con- will be crippled.28
how the volume, velocity, and volatility tent but have crucial limitations from
of digital evidence would increasingly Detection Technology
afect trial proceedings at the ICC, and Just as deepfakes grow more sophisti-
we detailed speciic recommendations cated, so does the need for tools that
to help the Court address some of their can detect whether a piece of audio-
most immediate needs.25 With genera- visual content has been AI generated
tive AI rapidly reshaping the landscape or manipulated. It is unlikely that the
of truth in armed conlicts and situa- ICC will be able to develop, and keep up
tions of mass violence, diferentiating the maintenance of, their own state-of-
genuine content from manipulated the-art detection tools. For this reason,
media is now assuming a new urgency. partnerships with leading companies
his is not just a technical challenge but can help ill this detection gap (the
a fundamental threat to the course of Oice of the Prosecutor already has an
justice. ongoing collaboration with Microsot).
Institutional responses must evolve However, it is important to note three
in lockstep with technological innova- shortcomings that may likely arise.
tions. As deepfake technology becomes Generally speaking, detection tools
increasingly advanced and accessible, are still in their infancy and have yet
the ICC can counteract the threats to be deployed at scale in a suiciently
this brings by investing in three areas: reliable fashion. While there has been
provenance infrastructure, detection dramatic progress in audio synthesis
technology, and analytical and foren- over the year 2023, audio biometrics,
sic expertise. Given the significant a litigation point of view. For inter- which could help identify whether
resources required to develop and national criminal justice, the most someone uttered certain words, still
deploy state-of-the-art technology useful approach to provenance is what lag behind. As Sam Gregory, executive
and the capacity gap on media foren- is known as “controlled capture,” spe- director at WITNESS, puts it, “[i]t’s no
sics (not unique to the ICC), these cialized sotware that incorporates data use having a tool to spot whether con-
approaches will require multistake- points at the moment of recording an tent is generated by one company when
holder collaboration, notably with the image or audio ile and maintains the the same tool would give a false neg-
private sector. None of these solutions chain of custody throughout the life of ative on fake audio created by one of
will be suicient on their own, but com- the footage.26 To date, only one solution the many other tools on the market.”29
bined they can be what tips the balance has been developed that meets these When AI detection tools are trained on
from impunity to justice. criteria of the shelf and has success- data not speciic to an individual, their
fully been tested in a court of law: the accuracy is similar to that of humans.30
Provenance Infrastructure eyeWitness to Atrocities app, incubated If trained on the biometrics of a per-
he irst line of defense against deep- at the International Bar Association son, such as President Zelenskyy,31 their
fakes is provenance infrastructure—a (in the spirit of full disclosure, I was accuracy in determining the authentic-
suite of technology solutions that can involved in setting up the organization ity of a given audio can jump to 99%.
help trace the origin of a piece of con- and was one of the lawyers working on his said, current deepfake detection
tent and any modiications since its the irst case that used content collected technology has been largely optimized
creation. Provenance tools embed with the eyeWitness app).27 Regardless for facial recognition, and it is less adept
metadata into a piece of media, creating of the criteria that provenance tools are at discerning tampering in nonhuman

elements of footage. his limitation is speciic piece of content is not a deep- to provenance, detection, and expertise
critical: the manipulation of structures fake; it just means that the system did that can keep pace with the evolv-
or the environment in a video—like a not detect what it was looking for.32 For ing technology. he sheer quantity of
doctored image of a bombed building or example, it could be a deepfake gener- potentially AI-generated or manipulated
the insertion of a falsiied object such as ated in a new way that the detection content threatens to overwhelm investi-
illegal ammunition in a civilian area— tool had not yet been trained to iden- gators and analysts tasked with triaging
can signiicantly alter the course of an tify. Legal practitioners will also have to audiovisual content, potentially leading
investigation. grapple with the multiple variations of a to miscarriages of justice. We owe it to
Second, algorithms trained to spot manipulation. A video could be entirely the victims and survivors of these crimes
inconsistencies in images, videos, or synthetic, or the deception could lurk to invest now in preparing our interna-
audios underperform in non-English only in a fraction of the frames or the tional justice system for a future that is
languages, noisy environments, and voice-over. Current models struggle already here.
complex scenarios like conlict settings. with pinpointing targeted manipula-
Grainy footage, low-resolution, or lack- tions, such as the addition of an object or
ing a representative biometric sample the alteration of a background. Further- Raquel Vazquez Llorente is the head
will challenge computational assess- more, the credibility of a witness could of Law and Policy, Technology Threats
ments. Additionally, automated analysis be called into question as a sophisticated and Opportunities at WITNESS.
is more efective when the ile is as close deepfake could shake the conidence in She is an international criminal
as possible to the original version. Social the authenticity of legitimate evidence. lawyer specializing in how emerging
media platforms, by compressing images If it is possible to falsify one event, what technologies impact our trust in
and modifying file metadata upon is stopping any evidence from being audiovisual content. She has over a
uploading, can obscure the digital foot- doubted? his broad basis for mistrust decade of experience with digital
print, thereby confounding automated threatens the core of international jus- evidence in conlict and crises.
systems. As long as audiovisual content tice—the establishment of an objective
continues to be posted online, the digital truth. ENDNOTES
footprint will get muddled, confounding 1. Megan O’Hearn, But It Looks So Real!
automated systems. his is why detection Analytical and Forensic Expertise he Parallel Rise of Photography and Spiritual-
algorithms must be paired with contex- Given these shortcomings, detec- ism, JSTOR (Oct. 20, 2016), https://about.jstor.
tual analysis and corroborating material tion tools will most likely be better org/blog/but-it-looks-so-real-the-parallel-rise-
to ascertain the authenticity of content. employed to conduct the initial triaging of-photography-and-spiritualism/.
To enhance the sensitivity of these mod- of audiovisual content, instead of aim- 2. Ian J. Goodfellow et al., Generative
els, there is a pressing need for datasets ing at producing conclusive veriication Adversarial Nets, arXiv 1406.2661v1 [stat
of manipulated videos and audio, spe- reports. Ascertaining authenticity will ML] (2014), https://arxiv.org/abs/1406.2661.
cifically derived from social media. require investing in human analysis in 3. Sarah Cahlan, How Misinformation
Collaboration with these platforms is a twofold manner, by strengthening the Helped Spark an Attempted Coup in Gabon,
then essential to help researchers reine knowledge and capacity of staf, particu- Wash. Post (Feb. 13, 2020), https://www.
detection algorithms. Similarly, the larly analysts and the judiciary, but also washingtonpost.com/politics/2020/02/13/
integration of biometric information through secondment of expertise from how-sick-president-suspect-video-helped-
is necessary. However, we should also governments, other accountability bod- sparked-an-attempted-coup-gabon/.
understand that acquiring potentially ies, specialized institutions like Justice 4. Sam Gregory, he World Needs Deepfake
sensitive data could put at risk those Rapid Response (JRR), and the private Experts to Stem his Chaos, WIRED (June
to whom we are trying to bring justice, sector. Interpreting the events under 24, 2021), https://www.wired.com/story/
should this information fall in the hands question will also necessitate updating opinion-the-world-needs-deepfake-experts-
of malicious actors or oppressive gov- the Court Registry’s roster to incorpo- to-stem-this-chaos/; he Irrawaddy, Myanmar
ernments. History teaches us that mass rate expert witnesses who are skilled in Junta Accused of Using Deepfake Technology to
atrocities are oten preceded by mass media synthesis and forensic analysis Prove Grat Case Against Daw Aung San Suu
surveillance, enabled by the collection and are able to testify in court to explain Kyi, Irrawaddy (Mar. 25, 2001), https://www.
of data and the monitoring of targeted provenance techniques such as digital irrawaddy.com/news/burma/myanmar-junta-
communities. signatures, provenance standards like accused-using-deepfake-technology-prove-
Last, from a legal standpoint, the C2PA, detection techniques, and other grat-case-daw-aung-san-suu-kyi.html.
deployment of automated detection technical aspects related to the authen- 5. Nathaniel Janowitz, Venezuela Is Using
technology raises complex questions. If ticity of evidence. Fake AI American Newscasters to Spread
such evidence is to be used in court, how As deepfake technology becomes Disinformation, Vice (Feb. 24, 2023), https://
will it be presented and understood by more accessible and convincing, the www.vice.com/en/article/z34jge/venezuela-ai-
the parties? he fact that a model did ICC and accountability institutions must newscaster-disinformation; Adam Satariano
not ind manipulation does not mean a pivot towards sophisticated approaches & Paul Mozur, he People Onscreen Are Fake.

he Disinformation Is Real, N.Y. Times (Feb. 7, 15. Venezuela: Investigating Crimes 23. Lindsay Freeman & Raquel
2023), https://www.nytimes.com/2023/02/07/ Against Humanity, Clooney Found. for Vazquez Llorente, How to Prepare the
technology/artiicial-intelligence-training- Just. (2023), https://cfj.org/the-docket/ International Criminal Court for Our
deepfake.html. venezuela/#:~:text=Our%20team%20at%20 Digital Future,” OpinioJuris (Oct. 12,
6. Rodrigo Soriano, Las Imágenes de he%20Docket,of%20security%20forces%20 2021), https://opiniojuris.org/2021/10/12/
Xóchitl Gálvez Editadas con Inteligencia in%20Venezuela. how-to-prepare-the-international-criminal-
Artiicial: Oportunidades y Peligros de Una 16. FIDH, Mexico Requires the Support court-for-our-digital-future/.
Nueva Arma para la Comunicación Política, of the ICC to Eradicate Structural Impu- 24. Rome Statute of the International
El País (July 10, 2023), https://elpais.com/ nity, reliefweb (May 26, 2020), https:// Criminal Court, art. 54.1.a, July 17, 1998,
mexico/2023-07-10/las-imagenes-de-xochitl- www.idh.org/en/region/americas/mexico/ 2187 UNTS.
galvez-editadas-con-inteligencia-artiicial- mexico-requires-the-support-of-the-icc-to- 25. Lindsay Freeman & Raquel Vazquez
oportunidades-y-peligros-de-una-nueva- eradicate-structural; FIDH, IDHEAS and Llorente, Finding the Signal in the Noise:
arma-para-comunicacion-politica.html; Jorge Colectivo Solecito, “Hasta encontrarlos”: International Criminal Evidence and Procedure
Ramis, Las Voces Generadas por IA se Vuelven enforced disappearances by security forces in in the Digital Age, 19 J. Int’l Crim. Just. 163
Indistinguibles de las Humanas (También en Veracruz constitute crimes against humanity (2021).
Español),” WIRED (Nov. 8, 2023), https:// (Feb. 2022), https://www.idh.org/IMG/pdf/ 26. Raquel Vazquez Llorente & Wendy
es.wired.com/articulos/voces-generadas-por- enforced_disappearances_in_veracruz.pdf. Betts, Coding Justice? he Tradeofs of Using
ia-indistinguibles-las-humanas-en-espanol. 17. Situation in Ukraine: ICC Judges Technology for Documenting Crimes in
7. Tom Simonite, A Zelensky Deep- Issue Arrest Warrants Against Vladi- Ukraine,” OpinioJuris (Oct. 22, 2023),
fake Was Quickly Defeated. The Next mir Vladimirovich Putin and Maria https://opiniojuris.org/2022/10/22/coding-
One Might Not Be, WIRED (Mar. 17, Alekseyevna Lvova-Belova, ICC (Mar. justice-the-tradeofs-of-using-technology-for-
2022), https://www.wired.com/story/ 17, 2023), https://www.icc-cpi.int/news/ documenting-crimes-in-ukraine/.
zelensky-deepfake-facebook-twitter-playbook. situation-ukraine-icc-judges-issue-arrest- 27. Chiara Gabriele, Kelly Matheson &
8. Jack Goodman & Mohanad Hashim, AI: warrants-against-vladimir-vladimirovich- Raquel Vazquez Llorente, he Role of Mobile
Voice Cloning Tech Emerges in Sudan Civil War, putin-and#:~:text=Today%2C%2017%20 Technology in Documenting International
BBC (Oct. 5, 2023), https://www.bbc.co.uk/ March%202023%2C%20Pre,Ms%20Maria%20 Crimes: he Afaire Castro et Kizito in the
news/world-africa-66987869. Alekseyevna%20Lvova%2DBelova. Democratic Republic of Congo, 19 J Int’l Crim.
9. Rania (@umyaznemo), Do you see what I 18. Darfur, Sudan: Situation in Darfur, Just. 107 (2021).
see? An Israeli refugee camp!, X (Oct. 22, 2023, Sudan, ICC-02//05, ICC, https://www.icc-cpi. 28. Chiara Gabriele, Kelly Matheson
8:42 PM), https://twitter.com/umyaznemo/ int/darfur. & Raquel Vazquez Llorente, Incorporating
status/1716173594818932808. 19. Karim A. A. Khan KC, Statement of ICC Digital Technology in the Investigation of Inter-
10. Aleksandra Wrona, Are hese Real Prosecutor, Karim A. A. Khan KC, to the United national Crimes: Lessons from the Democratic
Pics of a Massive Crowd of Israelis Demon- Nations Security Council on the Situation in Republic of Congo,” Just Sec. (Oct. 14,
strating Support for Israel?, Snopes (Oct. 23, Darfur, Pursuant to Resolution 1593 (2005), 2021), https://www.justsecurity.org/76780/
2023), https://www.snopes.com/fact-check/ ICC (July 13, 2023), https://www.icc-cpi.int/ incorporating-digital-technology-in-the-
fake-photo-israel-crowds. news/statement-icc-prosecutor-karim-khan- investigation-of-international-crimes-lessons-
11. Shayan Sardarizadeh (@Shayan86), his kc-united-nations-security-council-situation- from-the-democratic-republic-of-congo.
Video Certainly Doesn’t Show a New Air Assault darfur-0. 29. Morgan Meaker, Deepfake Audio
on Israel by Hamas Militants, Because It’s Actu- 20. Decision on the “Prosecution Request Is a Political Nightmare, WIRED (Oct. 9,
ally from the Video Game Arma 3, X (Oct. 9, Pursuant to Article 19(3) for a Ruling on the 2023), https://www.wired.co.uk/article/
2023, 1:44 PM), https://x.com/Shayan86/statu Court’s Territorial Jurisdiction in Palestine, keir-starmer-deepfake-audio.
s/1711180889453953238?s=20. ICC-01/18-143, ICC (Feb. 5, 2021), https:// 30. Kimberly T. Mai et al., Warning:
12. Shayan Sardarizadeh (@Shayan86), www.icc-cpi.int/court-record/icc-01/18-143. Humans Cannot Reliably Detect Speech Deep-
Some AI-Generated Images Are Being Shared 21. Karim A. A. Khan KC, Statement of fakes, 18 PLoS One e0285333 (2023), https://
in Relation to the Israel-Hamas Conlict, X ICC Prosecutor on the Situation in the State of journals.plos.org/plosone/article?id=10.1371/
(Oct. 24, 2023, 3:54 PM), https://twitter.com/ Palestine and Israel, United Nations (Oct. 30, journal.pone.0285333.
Shayan86/status/1716830625238544859/ 2023), https://www.un.org/unispal/document/ 31. Matyáš Boháček & Hany Farid, Protect-
photo/1; Shahin Hazamy (@shahin_ statement-of-icc-prosecutor-on-the-situation- ing President Zelenskyy Against Deep Fakes,
hazamy), It’s Genocide, Instagram (Oct. in-the-state-of-palestine-and-israel/. arXiv (June 24, 2022), https://arxiv.org/
17, 2023), https://www.instagram.com/p/ 22. Beth Van Schaak, Event: We Hold pdf/2206.12043.pdf.
Cygt-uWLKwS/?hl=en. hese Truths: How Veriied Content Defends 32. Mohamed Kambal (@Muhammed-
13. Preliminary Examination: Gabon, ICC- Democracies, CSIS (Mar. 7, 2023), https:// Kambal), Other Types of Forms of
01/16, ICC, https://www.icc-cpi.int/gabon. csis-website-prod.s3.amazonaws.com/ Manipulations Could Hatve Taken Place,
14. Information for Victims: Bangladesh/ s3fs-public/2023-03/230327_Verified_ Twitter (Aug. 1, 2023, 7:59 PM),
Myanmar, ICC, https://www.icc-cpi.int/ Content_Democracies.pdf?VersionId=2Gcq https://twitter.com/MuhammedKambal/
victims/bangladesh-myanmar. VaJZzK6XqUaRoQNbD9ZQT6SRtiI. status/1686451575978344470.

Get the SciTech Edge
MEMBERSHIP AND DIVERSITY COMMITTEE NEWS
BY JOANNE CHARLES
THE CARE AND FEEDING OF Some are useful. his year, make networking part of
YOUR LEGAL NETWORK your professional schedule. Digital tools can set meet-
A
s we embark upon the new ings and reminders for check-ins with those already in
year, we look to our calen- your network and note colleagues’ work anniversaries
dars for the events that will or birthdays. hese tools can streamline the networking
mark the milestones of 2024. If pro- process, so don’t skip them.
fessional development features high Communicate authentically. Ater you’ve met new
on your list of New Year’s resolutions, make networking colleagues, establish and maintain consistent commu-
the heart of your plans. Networking can feel overwhelm- nication. It’s not enough to send one “nice to meet you”
ing, but with thoughtful planning, your next networking email. hese communications shouldn’t be transactional,
opportunity may yield satisfying engagements. Below but instead should focus on relationship development.
are tips to help you build and improve your network- Consider using these communications to facilitate
ing skills. knowledge sharing or introductions to those in your
Utilize conferences. Establish a conference calen- community or your existing network.
dar and attend programs that support your educational he ABA Science & Technology Law Section will host
goals. Before any conference, research speakers and pan- its inaugural Spring Meeting and Privacy Institute in
elists by visiting their professional social media page or Washington D.C., April 17–19, 2024. his three-day
irm website. Your research goal is to enable thoughtful event ofers a place to network and learn about the lat-
engagements with the presenters you speak with and est legal developments in scitech law. We invite members
conversations that go beyond supericial comments or to join us and come ready to take advantage of the net-
questions. hese conversations have the potential to working opportunities the conference will ofer. For more
start a long-term relationship, so invest time in getting information, visit ambar.org/scitech. More ABA in-per-
prepared. son events are at https://www.americanbar.org/events-cle.
Get involved. Join a local chapter of your state bar
association and take advantage of opportunities to con-
tribute. Volunteer as a mentor, or host pro bono working
events for more impact. Working on pro bono events
builds practical experience and connects legal profes- Joanne Charles is chair of the MAD Committee and
sionals to like-minded colleagues. Developing these associate general counsel at Gilead Sciences. Her
relationships can be as fulilling as your community work focuses on AI, privacy, and data ethics. Prior to
service. joining Gilead, Joanne was senior corporate counsel
Leverage technology. We get bombarded with ads in Microsoft’s Corporate, External and Legal Afairs
for apps promoting the latest AI or eiciency product. Department.

Future SciTech Leaders
LAW STUDENT ENGAGEMENT COMMITTEE NEWS
BY DAVID HUSBAND AND CAYLAN FAZIO
FIRESIDE CHAT WITH THE into solutions that use contracts, policy, and other legal
SCITECH CLEANTECH structures to create change. he view Brammer and Solo-
& CLIMATE CHANGE way provided into clean energy is just that, a balance of
COMMITTEE the technical and legal work. But this ield is not exclusive
he Law Student Engagement to scientists. he overlap in environmental issues is boun-
Committee (LSEC) continues to tiful with areas for intersectional development of young
host events for law students that professionals of all backgrounds.
allow students to interact with Later in the month, Brammer and CTC3 held a webi-
experienced attorneys who vol- nar titled, Analyzing U.S. Billion-Dollar Weather & Climate
unteer their time for the ABA. Disasters: he Latest Data & Tools Used in Weather and Cli-
his December, LSEC hosted a mate Legal Proceedings, with Adam Smith of the National
discussion with the leadership of Oceanic and Atmospheric Administration (NOAA). his
the Cleantech & Climate Change talk highlighted the tools NOAA is using to monitor the
Committee. Law student and U.S. climate, particularly the costs and risks associated with
LSEC committee member Caylan Fazio writes about natural disasters and weather events. his discussion raised
her impressions of the chat and another climate webi- another crucial issue: accessing and leveraging high-qual-
nar below. ity data to make decisions. It is clear that data can improve
—David Husband, Co-Chair of LSEC environmental outcomes at every level, across government
and private industry, from planning to recovery eforts.
O
n December 5, LSEC hosted the leadership of the As law students and young professionals enter the
Cleantech & Climate Change Committee (CTC3) clean energy sector, they are aforded opportunities to
to explore the overlap of law and policy with solve some of the most pressing challenges the world faces
today’s most pressing environmental issues. his ireside through the legal profession. he CTC3 provides resources
chat was moderated by LSEC’s Co-Chair David Husband and support for student involvement. As students explore
and Vice Chair Julie Park. hey were joined by Dr. Robert their interests and career options, CTC3 and the ABA
Brammer and Joshua Soloway, co-chairs of CTC3. ofer free Section membership to law students to attend
he conversation explored Brammer and Soloway’s webinars, volunteer with committees, and network with
unique paths into clean energy, AI’s impact on climate professionals in areas like clean energy and many others
action, and the legal issues that arise in this ield. Clean that spark their interest. More ireside chats with SciTech’s
energy is growing, requiring more attention from lawyers substantive committees are planned in 2024. You can ind
with interdisciplinary skills across inance, data science, them at http://ambar.org/explorescitech.
and law. he clean energy sector is increasingly impacted
by regulations, tax incentives, and the growth of AI, all of David Husband is co-chair of the Law Student Engage-
which can afect the quality of climate data. ment Committee in ABA’s SciTech Section and works as
As someone with a background in mathematics and a senior counsel for the Board of Governors of the Fed-
analytics, I appreciate how data can both afect and be eral Reserve System. Caylan Fazio is a second-year law
afected by its governance. My hope in pursuing law school student at Cleveland State University College of Law. Her
Image: alexey_boldin / iStock / Getty Images Plus
was to expand my perspective beyond merely data and legal interests include space law, data privacy, and tax law.

SCIENCE & TECHNOLOGY LAW SECTION Nonproit
Organization
American Bar Association U.S. Postage
321 North Clark Street PAID
American Bar
Chicago, Illinois 60654-7598 Association
PC54580002002
ALSO FROM THE

SCIENCE &
TECHNOLOGY LAW
SECTION
» SciTech e-Merging News quarterly

electronic newsletter
» Jurimetrics Journal quarterly
electronic law review
SAVE THE DATE
» Free hot-topic committees ABA Science & Technology Law Section
and listservs
» Networking opportunities SPRING MEETING
April 17–19, 2024
» CLEs, webinars, and podcasts
» Books to enhance your practice
» Members-only discounts
» And much more WA S H I N G TO N , D C
Join us for:
ONE-DAY PRIVACY INSTITUTE - APRIL 17, 2024
CALENDAR • Section CLE programs

• Section business meetings
APRIL 17–19, 2024 • Networking events
ABA SciTech Section’s Inaugural • SciTech’s 50th Anniversary
Privacy Institute and Spring
Conference
Washington, D.C.
ambar.org/scitech
OCTOBER 14–15, 2024

2024 Artiicial intelligence & Robotics
National Institute
Santa Clara University School of Law
ambar.org/scitech
CONNECTWITH#SCITECH
ambar.org/SciTech www.facebook.com/ABASciTech twitter.com/ABASciTech www.linkedin.com/company/aba-scitech

20 Sci Tech Law 1

Uploaded by

Copyright:

Available Formats

20 Sci Tech Law 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

20 Sci Tech Law 1

Uploaded by

Copyright:

Available Formats

DATE DOWNLOADED: Tue Jun 25 21:34:45 2024

SOURCE: Content Downloaded from HeinOnline

Bluebook 21st ed.

ALWD 7th ed.

APA 7th ed.

Chicago 17th ed.

AGLC 4th ed.

OSCOLA 4th ed.

4 JUSTICE BY VIDEO: DO COURTS NEED LEGAL GUIDELINES FOR

8 CONSTRUCTING MEANING IN VIDEO EVIDENCE

13 HOW COGNITIVE BIAS CAN IMPAIR FORENSIC FACIAL

17 POLICE BODY CAMERAS: ANALOG THINKING ABOUT A DIGITAL

22 INTERNATIONAL VERIFICATION STANDARDS FOR OPEN SOURCE

28 DEEPFAKES IN THE DOCK: PREPARING INTERNATIONAL JUSTICE

34 GET THE SCITECH EDGE: MEMBERSHIP AND DIVERSITY

35 FUTURE SCITECH LEADERS: LAW STUDENT ENGAGEMENT

WINTER 2024 TheSciTechLawyer 3

Justice by lines on how video can be used and

Guidelines for ot-cited case, the court had to decide

Video Evidence? tional protection against unreasonable

4 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 5

6 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 7

8 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 9

10 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 11

12 TheSciTechLawyer WINTER 2024

How Cognitive in ways that encourage adaptive behav-

Bias Can Impair estimated the same hill to be steeper

Forensic Facial way of discouraging us from climbing

Identification also impede eforts to determine the

WINTER 2024 TheSciTechLawyer 13

14 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 15

16 TheSciTechLawyer WINTER 2024

POLICE BODY-WORN CAMERA

Analog ater in New York City,2 BWCs are now

a Digital ally resisted by police oicers and their

Question part of “best practices” in policing.5 he

WINTER 2024 TheSciTechLawyer 17

18 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 19

20 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 21

22 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 23

24 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 25

26 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 27

By Raquel Vazquez Llorente

28 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 29

30 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 31

32 TheSciTechLawyer WINTER 2024

WINTER 2024 TheSciTechLawyer 33