ll

NeuroView
Can Neuroscience Change the Way We View Morality?
Clare Kelly1,2,* and Redmond O’Connell1,*
1TrinityCollege Institute of Neuroscience & School of Psychology, Trinity College Dublin, Dublin, Ireland
2Department of Psychiatry at the School of Medicine, Trinity College Dublin, Dublin, Ireland
*Correspondence: clare.kelly@tcd.ie (C.K.), reoconne@tcd.ie (R.O.)
https://doi.org/10.1016/j.neuron.2020.10.024

Neuroscience has cast new light on the nature of human morality by exploiting simplified paradigms. To
enhance our understanding of everyday moral decisions, the field should complement computational ap-
proaches with naturalistic paradigms and a focus on narratives and stories.

Morality is as firmly grounded in chain a shift in perspective on morality— harmful actions that use someone as a
neurobiology as anything else we from a rulebook etched in our minds means to an end, even if that end is saving
do or are. —Frans de Waal (Good through socialization to a diverse set of five lives) were associated with relatively
Natured: The Origins of Right and evolutionarily deep-rooted psychological weaker activation of brain areas associ-
Wrong in Humans and Ani- capacities which represent nature’s solu- ated with emotional appraisal (such as
mals, 1996) tion to the need to put Us ahead of Me ventromedial prefrontal cortex [vmPFC])
(Greene, 2013). These insights created and stronger activation of areas associ-
As humanity faces unprecedented health an imperative to understand the moral de- ated with deliberative reasoning (such as
and environmental crises, routine daily cisions that humans make in practice; dorsolateral prefrontal cortex [dlPFC])
decisions have gained a new moral neuroscience has claimed a position at (described in Greene, 2013). Further evi-
weight. Should you forgo a social gath- the vanguard of this field of inquiry. Over dence from neuropsychological and neu-
ering to reduce your chances of transmit- the last three decades, great strides roimaging studies showed that a person
ting COVID-19? Should you opt to have been achieved in elucidating the can be shifted more toward utilitarian or
commute to work by bicycle, rather than neural bases of moral choice, but deontological judgments by disruption or
by car, to reduce your carbon footprint? what are the implications of this knowl- damage to vmPFC or dlPFC functioning,
Will your actual behavior match what edge? In this piece we adopt a neurohu- respectively.
you believe you should do? At a time manities perspective to examine whether These initial studies suggested that
when individual choices can have such neuroscience’s purview is limited to moral decision making could be captured
serious and far-reaching consequences describing the physiological processes by a ‘‘dual process’’ model in which fast,
for our species, it has never seemed that underlie moral decisions, or if it can automatic emotional responses and
more important to understand how hu- change the way we view ourselves as slow, deliberative reasoning processes
mans make moral decisions. moral agents. independently influence choice behavior
Questions about the nature of human and battle for supremacy when conflicts
morality have long been the domain of Shared Neural Machinery for Moral arise (Greene, 2013). Subsequent work
the humanities. For centuries, philoso- and Non-moral Decisions has elaborated on this model, piecing
phers have sought to identify a set of One of neuroscience’s most striking con- together a rich patchwork of cognitive
universal moral principles that separate tributions has been to cast new light on abilities that play pivotal roles in morality,
right from wrong and thereby specify what it is that enables us to make sophis- including visualization (hippocampus),
how we ought to act. When psychologists ticated moral decisions. This work has es- empathy (anterior cingulate, anterior in-
first began to apply scientific methods to tablished that, rather than reflecting the sula), perspective-taking and mentalizing
study the development of human morality, output of a single evolutionarily special- (dorsomedial PFC, temporoparietal junc-
their theories were strongly rooted in ized neural module, our moral decisions tion [TPJ]), emotional responses (amyg-
these normative philosophical frame- are the product of many of the same dala and anterior insula), value and reward
works. In time, however, anthropological perceptual, cognitive, and socioemo- representations (vmPFC, ventral stria-
and psychological evidence demon- tional processes that we invoke for all tum), and the biasing of behavior in accor-
strated that, in reality, our moral decisions kinds of decisions. These processes dance with long-term goals (dlPFC and
are strongly influenced by a wealth of make independent, sometimes conflict- dorsal caudate) (e.g., Buckholtz and Mar-
social, emotional, cultural, and contextual ing contributions to the final choice. For ois, 2012).
factors. Together with ethological re- example, a famous series of studies of The observation of a shared neural cir-
search that identified rudimentary forms moral dilemmas showed that decisions cuitry for moral and non-moral decisions
of empathy and compassion in non-hu- favoring utilitarian outcomes (e.g., killing implies that many phenomena pertaining
man animals and advances in social evo- one person to save the lives of five others) to the latter will also manifest in the
lution theory, these observations set in over deontological ones (i.e., rejecting former. Indeed, cognitive science has

604 Neuron 108, November 25, 2020 ª 2020 Elsevier Inc.

 ll
NeuroView

already established that our moral deci- application of computational models, subjective value of the different choice
sions are subject to many of the same which originated in research examining options presented to participants and
biases, heuristics, and capacity limita- basic perceptual and economic deci- was then regressed against fMRI data to
tions that we exhibit in our economic sions, to parse and estimate the latent isolate choice-relevant brain activations.
choices. In a similar vein, Kameda and appraisal dimensions that inform our Within the identified set of regions, activa-
colleagues (2016) provided evidence that moral choices. A powerful aspect of this tion of the dorsal striatum in response to
interindividual differences in distributive approach is that these model-derived es- money gained from shocking others was
justice judgments share a common origin timates can be used as parametric re- relatively reduced in more altruistic partic-
with choice policies for risky personal de- gressors for neurophysiological data, ipants whereas no behaviorally predictive
cisions. In their study, individuals who thus reducing the need for informal neural correlates of anticipated pain were
preferred high-risk, high-gain personal reverse inference in the interpretation of observed, suggesting that only the former
gambles also tended to opt for more utili- observed neural correlates. Computa- was being factored into the decision-
tarian distributive choices that maximized tional approaches thus offer a principled making process. The lateral PFC was
the overall financial gain among three framework within which to develop also found to track judgments of the
unknown others, whereas individuals models of morality that provide parsimo- blameworthiness of harmful actions and
who were riskaverse tended to make nious yet comprehensive accounts of exhibited functional connections with the
distributive choices that maximized the key behavioral and neural phenomena. same region of dorsal striatum. This latter
amount that the least fortunate benefi- For example, numerous studies have observation suggests that moral actions
ciary would receive. The latter strategy observed that choice response times are not achieved by inhibiting the tempta-
was interpreted as being consistent with and vmPFC activation are reduced for tion to further our own selfish interest but
the ‘‘maximin’’ principle outlined by the selfish relative to generous choices, and by devaluing choices that would cause
moral philosopher John Rawls, according this difference tends to be smaller in harm to others.
to which, benefit to the neediest (the more generous participants. This pattern The adoption of models that have
worst possible outcome) should be maxi- has previously been accounted for by been extensively road-tested in investi-
mized. Individual differences in the appli- invoking a dual-process competition be- gations of perceptual and value-based
cation of the maximin strategy were tween fast, automatic systems and slow, decision making has been a central
correlated with differences in the strength deliberative ones. However, a recent driver of recent progress in research on
of functional connectivity between the study (Hutcherson et al., 2015) showed moral decision making. The neurosci-
right TPJ (a region associated with mental that these behavioral and neural data pat- ence of morality will further benefit from
simulation of another perspective) and the terns can be more parsimoniously ex- ongoing developments in these closely
caudate (which is thought to compute plained by a multi-attribute version of the related fields, which currently strive to
decision value). These intriguing findings Drift Diffusion Model (DDM)—a dominant pinpoint how our decision processes
underline the centrality of perspective tak- model in research on perceptual deci- are shaped by key contextual factors
ing and mental simulation in decision sions—in which altruistic choices are like time pressure, prior knowledge, un-
making and suggest that even purport- based on a single deliberative process certainty, delay discounting, and relative
edly non-moral decisions can have a that integrates two independent attributes outcome value. Electrophysiological re-
moral flavor—when such decisions have representing the subjective value of gains cordings in non-human animals have
consequences for future selves. More- for oneself and for others. played a pivotal role in this work,
over, in demonstrating that individual A significant body of recent work providing an invaluable means of
differences in solutions to distributive has also adapted economic utility and measuring the neural decision process
justice problems partly reflect neurocog- Bayesian reinforcement learning models in action and, hence, of testing and
nitive diversity in the weighting of other to examine alternative ways that the moral refining mathematical model predictions.
perspectives in elementary value-based decision process could be formulated. Such data can also contribute more
decisions, this study uncovers a potential For example, Crockett et al. (2017) asked directly to our understanding of moral
explanation for why our positions on participants to trade personal financial decisions. To date, comparative studies
questions of morality are often deeply en- gain for electric shocks that would have focused predominantly on empathy
trenched (Kappes et al., 2016). Studying be administered either to themselves or and related processes such as social
these differences and their malleability to an anonymous other. Participants communication and emotion recognition,
may simultaneously shed light on how behaved very altruistically on this task— since evidence of evolutionary conserva-
people can be shifted toward one or the they were willing to pay more to avoid tion of these fundamental building blocks
other moral stance and toward less risky shocks for others than for themselves. of morality is strongest. Such studies
behavior. Here, the combination of neural data and have revealed, for example, that the
Recognition of the considerable com- computational modeling allowed the au- same (‘‘mirror’’) neurons respond to
monalities between moral and non-moral thors to test whether this tendency re- both the direct, first-hand experience of
decisions has also created an impetus to flected a greater aversion to others’ pain positive (e.g., reward) and negative
embed research on moral decision mak- compared to one’s own pain or, alterna- (e.g., pain) outcomes and the observa-
ing within the broader field of decision tively, a devaluation of profit gained from tion of the same outcomes in conspe-
neuroscience. This has led to the fruitful harming others. A model estimated the cifics. Remarkably, the intensity of

Neuron 108, November 25, 2020 605

 ll
NeuroView

neuronal spiking in the anterior cingulate age, gender, ethnicity, culture, religion, digm approach to the examination of
cortex of one rat witnessing a cagemate class, and politics (Hester and Gray, moral narratives offers an opportunity to
receiving an electric shock could be 2020), as well as concerns about the role examine not only how stories are morally
used to decode the intensity of shocks of social desirability biases. Coupled effective, but also to help address dis-
that were witnessed as well as shocks with discipline-wide evidence of poor putes in the literature about how and
that the rat experienced first-hand. correspondence between the predictions when certain moral capacities emerge
These data provide a strong demonstra- of behavioral models developed under developmentally. Although moral devel-
tion of a shared neural code for the experimental conditions and real-world opment is a burgeoning area in terms of
distress of self and others (Carrillo behavior, these considerations motivate behavioral research, with highly innova-
et al., 2019), which is thought to provide a pivot toward ecologically valid and tive paradigms and robust debates, there
the basis for empathetic behavior. Estab- naturalistic stimuli (Nastase et al., 2020). is surprisingly little cognitive neurosci-
lishing the extent to which the building Naturalistic stimuli and paradigms chal- ence research in this area. This is under-
blocks of other facets of morality (e.g., lenge the experimentalist’s need for tight standable, due to the challenge of col-
inequity aversion, norm-following) are control over variables, but embracing lecting high-quality functional MRI or
shared with non-human animals awaits ‘‘the complexity of real life’’ (Nastase EEG data from young children—a chal-
further evidence. Regardless, sufficient et al., 2020) can open up new research lenge that the use of naturalistic stimuli
cross-species commonalities have been questions and avenues for investigation, such as movies can help overcome. By
demonstrated to suggest that compara- particularly in the context of complex collaborating with humanities scholars
tive neuroscience can make an important cognitive domains, such as morality. to examine age-related changes in pat-
contribution to advancing our under- Such approaches are being very success- terns of neural synchrony and asyn-
standing of moral decisions. fully applied in the study of other complex chrony during moral stories, we may
cognitive functions like perception of identify points of developmental change
Addressing the Complexity of speech and complex visual scenes, that in moral capacities and their relationship
Morality in Real Life are best understood using methods with brain maturation.
A key message of the work we have out- that closely approximate the real-world Much more remains to be gained from
lined is that moral decisions should be ecological contexts in which those func- the intersection of neuroscience and the
viewed as complex cognitive tasks whose tions evolved. humanities. For example, historical and
outcome is not determined solely by the Increasing adoption of naturalistic par- anthropological perspectives on societal
values we hold but also by the host adigms would also forge a new point of and cultural trends and differences,
of domain-general neurocognitive pro- connection between moral neuroscience changing moral frames, and generational
cesses that allow us to access and eval- and the humanities: stories. Throughout differences provide invaluable informa-
uate choice-relevant information. With human history, stories and narratives tion about the often rapidly changing
the increasing adoption of formal compu- have served as a primary vehicle for the context in which human brains are
tational models, the field is moving communication and transmission of immersed, yet neuroscience rarely ac-
beyond cartography of the moral brain to morals. Narratives are a primary mode knowledges these. A final consideration,
exploit the potential for neurophysiolog- through which children, and indeed hu- however, is what the neuroscientific
ical data to contribute directly to the mans of all ages, are taught about right study of morality can give back to the hu-
validation and refinement of cognitive and wrong. Stories activate a broad array manities. While we should always follow
models. However, current neuroscience of brain areas that support language, Hume’s advice, and avoid deriving
approaches to moral decision making emotion, memory, and imagination (e.g., ‘‘ought’’ from ‘‘is,’’ uncovering the neural
are not without shortcomings, such as hippocampus, TPJ, medial PFC)—many mechanics of morality may reveal that the
the use of highly stylized tasks to assess of which are also activated by moral de- principles guiding what we ‘‘ought’’ to do
a limited range of moral judgments (often cisions. A series of studies by Uri Hasson emerge naturally from brains built by
involving genderless and raceless anony- and colleagues (e.g., Nguyen et al., 2019) evolution to connect and cooperate. Not
mous others), and a focus on white, have used innovative approaches such only can this understanding serve to
educated, industrialized, rich, democratic as intersubject correlation of neural time strengthen the obligation to act morally,
(WEIRD) concerns and samples (Henrich series to show that stories synchronize but it can help us find effective ways of
and Muthukrishna, 2020). Learning about brain activity across listeners and be- nurturing our prosocial inclinations so
responses to abstract moral scenarios is tween listeners and the story-teller, and that the cooperation required to over-
undoubtedly important; both the COVID- that neural synchrony is greatest among come the current crises faced by our
19 pandemic and the climate crisis high- those who share the same interpretation species can be realized.
light the importance of grappling with of the story. In the humanities and social
questions such as how scarce resources sciences, stories are widely recognized
should be distributed among anonymous as an engine for social and moral REFERENCES
beneficiaries. Nonetheless, for the neuro- change. Yet the neural mechanisms
science of morality to gain greater rele- through which narratives exert their Buckholtz, J.W., and Marois, R. (2012). The roots
of modern justice: cognitive and neural founda-
vance it will have to address its current moral power have been little investi- tions of social norms and their enforcement. Nat.
neglect of identity-linked factors such as gated. Expanding the naturalistic para- Neurosci. 15, 655–661.

606 Neuron 108, November 25, 2020

 ll
NeuroView
Carrillo, M., Han, Y., Migliorati, F., Liu, M., Gazzola, Annu. Rev. Psychol. Published online October 2, chor in distributive justice and risky decisions.
V., and Keysers, C. (2019). Emotional mirror neu- 2020. https://doi.org/10.1146/annurev-psych- Proc. Natl. Acad. Sci. USA 113, 11817–11822.
rons in the rat’s anterior cingulate cortex. Curr. 081920-042106.
Biol. 29, 1301–1312.e6. Kappes, A., Kahane, G., and Crockett, M.J. (2016).
Hester, N., and Gray, K. (2020). The moral psychol- From risk to fairness. Proc. Natl. Acad. Sci. USA
Crockett, M.J., Siegel, J.Z., Kurth-Nelson, Z., ogy of raceless, genderless strangers. Perspect. 113, 11651–11653.
Dayan, P., and Dolan, R.J. (2017). Moral transgres- Psychol. Sci. 15, 216–230.
sions corrupt neural representations of value. Nat. Nastase, S.A., Goldstein, A., and Hasson, U.
Neurosci. 20, 879–885. (2020). Keep it real: rethinking the primacy of
Hutcherson, C.A., Bushong, B., and Rangel, A. experimental control in cognitive neuroscience.
Greene, J.D. (2013). Moral Tribes: Emotion, (2015). A neurocomputational model of altruistic Neuroimage 222, 117254.
Reason, and the Gap Between Us and Them choice and its implications. Neuron 87, 451–462.
(Penguin). Nguyen, M., Vanderwal, T., and Hasson, U. (2019).
Kameda, T., Inukai, K., Higuchi, S., Ogawa, A., Kim, Shared understanding of narratives is correlated
Henrich, J., and Muthukrishna, M. (2020). The H., Matsuda, T., and Sakagami, M. (2016). Rawlsian with shared neural responses. Neuroimage 184,
Origins and Psychology of Human Cooperation. maximin rule operates as a common cognitive an- 161–170.

Neuron 108, November 25, 2020 607