Keywords

1 Introduction

In the Digital Era, digital interface has played an important role in many important fields of civilian applications and national defense. For example, in nuclear power plant control room, digital interface is the main way for operators to obtain information and make response. The main task of the operator is to make decisions based on information. Some scholars have studied the impact of information on decision making. Kevin found that information can affect customers decision making through two components, quality and quantity. High quality improved decision effectiveness while high quantity decreased decision effectiveness [1]. Nicholas found that a structural approach such as information theory can better predict information overload [2]. However, few studies focus on digital interface, the relationship between information and decision making in digital interface still remains unknown.

Among the many factors affecting information, quantity is the most basic and most important one. There are many ways to measure information quantity. One of the most efficient measurement methods is information entropy [3,4,5]. Image entropy, which is the amount of average information in the image. Pixels located at different positions in the image have different gradations, and image entropy is expressed as the average number of bits of the set of gray levels of an image, the same applies to static interfaces. The image entropy can be expressed by the formula as shown in Eq. (1):

$$ H = - \sum\nolimits_{i = 1}^{L} {p_{i} \log_{2} p_{i} } $$
(1)

Decision making is the main task for complex information system operators. For example, nuclear power plant operators need to switch the button accordingly when indicators reach a critical threshold [6]. Fighter pilot need to decide when to engage or escape and which tactic to use [7]. Making right decisions is the fundamental guarantee for the safety of operators and facilities.

A large amount of data and high complexity in the digital interface of complex information systems poses great challenges to users’ decision making. Providing appropriate information to help user make decision is an important and promising research topic. We aim to find the image entropy limit value suitable for user’s decision-making and provide effective reference and new evidence for the design and evaluation of complex information system interface.

2 Methodology

2.1 Materials

The experimental materials were a series of interfaces with linearly increasing information quantity which were designed according to the real display interface of nuclear power plant, as shown in Fig. 1. There were ten information quantity levels in the experimental materials, the image entropy of each level was proportionally increased, from 0.2 to 2.0 with an interval of 0.2. In order to facilitate the experimental record, the interface images were named as information quantity 1 to 10, the larger the number, the higher the entropy, as shown in Table 1.

Fig. 1.
figure 1

Experiment interfaces with ten information quantity levels

Full size image
Table 1. Image entropy of experiment interfaces
Full size table

2.2 Subjects

20 subjects were enrolled from school of mechanical engineering in Southeast University. Subjects included 10 males and 10 females between the ages of 20 and 30 years (M = 24), all of whom were right handed and had normal vision or corrected vision. Subjects all had basic knowledge of computer operation and were trained to make rational decisions in the experiment.

2.3 Experimental Equipment and Experimental Procedures

The experiment was carried out in a soundproof lab under the normal lighting condition. Experimental equipment was a desktop computer and a Tobii Eye Tracker. The desktop computer had a 23-in. LCD monitor which can provide required resolution (1920 pixels * 1080 pixels). Tobii Eye Tracker was mounted at the bottom edge of computer monitor to obtain the correct recording position and angle.

The data needed to be collected included behavioral data and eye movement data. Both types of data were collected by Tobii Studio which is a supporting software of Tobii Eye Tracker. Tobii Studio needed to be installed on the computer in advance. Behavioral data included the correct rate and decision-making time. Eye movement data included pupil diameter, gaze plots and heatmaps.

Before the experiment, subject was asked to keep their eyes 620–660 mm away from the screen and both horizontal and vertical viewing angles were controlled in 2.3°. The subject was then asked to move his eyes to perform the software calibration. After the software calibration was successful, the experiment could be carried out.

The experimental tasks were divided into two parts: learning part and decision-making part. The learning part allowed subjects to achieve the level of operation required for the experiment, primarily through explanations, exercises, and feedback. The learning part began with the general introduction. After the Space key was pressed, the interface would display the normal state of the 10 interfaces of different information quantity in turn. Subjects needed to learn and remember the normal state of each interface. Then press the Space key to continue. The explanation interface would display typical abnormal values and show corresponding decision-making methods. After that came the practice, the practice would provide feedback to help subjects familiar with the operation, the learning part had no time limit. The flow chart is shown in Fig. 2.

Fig. 2.
figure 2

Experimental flow chart: Learning part

Full size image

The decision-making part was the main part of the experiment, and this part began with the experimental instruction. After the Space key was pressed, the visual guidance center appeared in the center of the screen for 1000 ms. Then subject needed to observe whether the data such as flow rate and temperature in the presented interface was normal, if abnormal data was found, press the corresponding number in the number pad. The number keys 0–9 represented 10 areas in the interface, and the subject decided which area to select by pressing the corresponding number key. Interfaces with different information quantity appeared randomly in the experiment to eliminate the influence of image changes on the subject. The flow chart is shown in Fig. 3.

Fig. 3.
figure 3

Experimental flow chart: Decision-making part

Full size image

The main purpose of the experiment is to study the impact of interface information quantity on user’s decision-making. By analyzing user’s eye movement data such as pupil diameter, gaze plots and heatmaps, the efficiency and load difference of decision making under different image entropy interfaces are explored, and the influence of interface information quantity on decision making is obtained.

3 Results

3.1 Behavioral Data

Table 2 shows the correct rate of decision-making by subjects under different information quantity levels. The correct rate is calculated by dividing the number of interfaces that are correctly responded by the total number of interfaces at the same information quantity level. The “Information quantity” column represents ten kinds of information quantity levels as listed in Table 1. “All” column represents the average correct rate of all subjects. “Male” column represents average correct rate of males (10 male subjects) and “Female” column represents average correct rate of females (10 female subjects).

Table 2. Correct rate
Full size table

Table 3 shows the decision-making time of subjects under different information quantity levels. Time is presented in seconds. The decision-making time is calculated from the time when gaze point enters AOI (Areas of Interest) which is an area containing abnormal value on the interface until the keyboard button is pressed. “All” column represents the average time of all subjects. “Male” column represents average time of males (10 male subjects) and “Female” column represents average time of females (10 female subjects).

Table 3. Decision-making time (s)
Full size table

Decision efficiency is defined as the ratio of correct rate to decision-making time. As a comprehensive indicator of decision making, the higher the decision efficiency, the better the quality of decision-making. The decision efficiency is calculated by dividing the value in Table 2 by the corresponding value in Table 3, the data is represented by a line graph as shown in Fig. 4. As can be seen from the figure, the decision-making efficiency starts to rise, reaches the maximum at the information quantity level 3, and then gradually decreases. Male and female subjects show differences.

Fig. 4.
figure 4

Relationship between information quantity change and decision efficiency

Full size image

3.2 Eye Movement Data

The change of pupil diameter is calculated from the maximum value of pupil diameter minus the minimum value. The change of pupil diameter reflects subject’s decision load. The greater the change, the higher the decision load. The pupil diameter changes at each information quantity level were represented by a line graph as shown in Fig. 5.

Fig. 5.
figure 5

Relationship between information quantity change and pupil diameter change

Full size image

The decision-making load increases with the increase of information quantity generally. At the information quantity level 7, male subjects have a big increase in the value while female subjects have a drop in the value.

Gaze plots analysis is performed on the interfaces with information quantity levels of 1, 3 and 10 respectively as shown in Fig. 6. Gaze plots can present the order of the fixation point and location, the size of the dot say fixation time, the number in the dot means the order of fixation points. As can be seen from Fig. 4 that the number of gaze points on information quantity 1 interface is 20, the number of gaze points on information quantity 3 interface is 19. Although the quantity of information varies greatly, the number of gaze points is almost the same. This is mainly because the information quantity 3 interface has short paths in the region and no backtracking paths between the regions. In contrast, the information quantity 1 interface has more backtracking paths. As information quantity increases further, the number of gaze points on the interface will increase and the number of gaze points on the information quantity 10 interface reaches 47.

Fig. 6.
figure 6

Typical gaze plots

Full size image

Heatmaps analysis is then performed on the same interfaces respectively as shown in Fig. 7. Heatmaps use cloud marks to show whether an area is concerned and how long it is concerned. The duration is reflected by color, with red being the longest, followed by yellow, and green the shortest. As can be seen from Fig. 7 that red cloud marks of different information quantity interfaces are all concentrated in abnormal data areas. The difference is that the number of green and yellow cloud marks increases with the increase of information quantity.

Fig. 7.
figure 7

Typical heatmaps (Color figure online)

Full size image

4 Discussion

4.1 Decision Efficiency

Results of decision-making efficiency indicate that the efficiency increases first and then decreases with the increase of information quantity. Information quantity has an optimal range, within this range, subjects have high decision-making efficiency. ANOVA is carried out to further analyze decision-making efficiency data, it is found that different quantity of information has no significant impact on decision efficiency (F = 1.414, p = 0.194, α = 0.05, p > α). Alvarez’s research shows that visual short-term memory has a fixed capacity [8]. Therefore, visual short-term memory may have an influence on the reception of visual information, so that there is an optimal range of information quantity. The findings from Jie Gao seem consistent with this opinion [9]. The optimal range of information quantity may be related to the upper and lower limits at which people receive information.

4.2 Decision Load

Results of decision load show that the load increases with the increase of information quantity. ANOVA is carried out to further analyze decision load data, it is found that different quantity of information has significant impact on decision load (F = 21.277, p = 0.000, α = 0.05, p < α). The rise in decision load may be due to the brain calling more resources to process more information. Therefore, information quantity should not exceed a certain upper limit.

4.3 Optimal Range

According to the experimental data of decision efficiency and the questionnaire survey conducted on the subjects after the experiment, the difference between the maximum and minimum values of decision efficiency is taken as the interval, and the 80–100% of the maximum value within this interval is defined as high decision-making efficiency, the 60–80% is defined as medium decision-making efficiency, and the 60% or less is defined as low decision-making efficiency. Similarly, the difference between the maximum value and the minimum value of the decision load is taken as the interval, the 80–100% of the maximum value within this interval is defined as high decision load, the 60–80% is defined as medium decision load, and the 60% or less is defined as low decision load. Taking the intersection of the information range of high decision efficiency and that of low decision-making load, the optimal range of information quantity suitable for decision-making is obtained as image entropy 0.4 to 0.6.

4.4 Gender Differences

Subjects of different genders showed differences in the decision-making process. The maximum value of decision efficiency curve of male is more inclined to high information quantity than that of female. The difference between male and female may be because male have better ability to process high information quantity. The decision load of male is higher than that of female in almost all information quantity, which may be due to the difference in thinking mode between male and female. Therefore, for male, the optimal range of information quantity can be adjusted to 0.4 to 1.2, while for female, it is 0.4 to 0.6.

The above research provides the optimal range of information quantity for decision-making in the interface, which makes it possible to quickly judge the quality of the interface through image entropy in the early stage of interface design.

Finally, it should be pointed out that decision-making is a complex process and information quantity is only one of the influencing factors. Therefore, it is necessary to comprehensively consider various factors to determine whether an interface is suitable for decision-making. This paper provides a quick evaluation method, but it is not perfect enough, more factors such as quality and prominence need to be studied in the future.

5 Conclusions

The aim of this study is to understand the influence of information quantity on user’s decision-making. Stimuli with image entropy ranging from 0.2 to 2.0 are designed by varying the number of interface information elements. Stimuli are designed to have the same size and contain the same basic elements so that they only differ in the quantity of information.

The experimental data was collected by Tobii eye tracker. By analyzing the subjects’ behavioral data and eye movement data, we found that stimuli with different image entropy showed different influence on user’s decision-making. As expected, decision load increased as the quantity of information increased. However, decision efficiency did not decrease as the quantity of information increased but rose first and then fell. This showed that people have lower and upper limits on the quantity of information when making decisions. Measured by image entropy, lower limit was 0.4 and upper limit was 0.6. In other words, the quantity of information suitable for user decision-making was between 0.4 and 0.6. We also found that males and females show differences in decision-making. The decision load of female was lower than that of male in almost all levels of information quantity. And female had higher decision efficiency in the range of image entropy from 0.4 to 0.6, while male is 0.4 to 1.2.

Complex information system digital interface that has high requirements for user decision-making can use image entropy method to quickly evaluate and design the interface information quantity to a reasonable range. Our research provided effective reference and new evidence for the design and evaluation of digital interface.