1. Home>
  2. Publicity>
  3. Press Release>
  4. 2025>
  5. Facial Expressions of Avatars Promote Risky Decision-Making
日本語
Print

Facial Expressions of Avatars Promote Risky Decision-Making

- The amygdala plays a key role in driving increased risk-taking -
April 23, 2025

National Institute of Information and Communications Technology

Highlights

  • Humans take more risks when interacting with facial expressions shown on avatars rather than real human faces.
  • This shift in risk-taking behavior is linked to activity in the amygdala.
  • The findings offer new insights into both the advantages and the cautionary aspects of communication via avatars.

Abstract

Figure 1 Example scenario
How expressing opinions in an online meeting may feel different depending on whether a supervisor’s face is shown directly or represented as an avatar.
 A research team led by Dr. TANAKA Toshiko and Dr. HARUNO Masahiko at the National Institute of Information and Communications Technology (NICT, President: TOKUDA Hideyuki Ph.D.), investigated how avatar-mediated communication affects human decision-making. They discovered that participants were more likely to take risks when facial expressions (such as admiration or contempt) were displayed by avatars than when the same expressions were shown on real human faces. This increase in risk-taking was found to result from a more favorable valuation of the "uncertainty" of facial feedback in the avatar condition. Furthermore, fMRI analysis revealed that this valuation of uncertainty depends on activity in the amygdala.
These results offer important insights into how avatar-based social communication, such as in virtual and augmented reality, can affect human decision-making, and highlight the amygdala's critical role in this process. The findings were published in the April 22, 2025 issue of the journal PLOS Biology, which features high-impact research in the life sciences.

Background

In recent years, avatar-mediated communication has rapidly spread in contexts such as online meetings and customer service. However, scientific studies on how avatars influence human cognition and decision-making are still in their early stages. In particular, little is known about how behavior changes when a communication partner's facial expressions are shown via avatars rather than real faces. For example, situations such as expressing opinions during online meetings or shopping in virtual stores increasingly involve avatars as communication partners. Understanding how avatars' facial expressions influence risk-related decisions in such settings is crucial.

Achievements

In this study, participants performed a risk decision-making task while undergoing fMRI scanning. In each trial, they were presented with two options and asked to choose their preferred one:
  • A "safe" option: a guaranteed small reward (e.g., 80 yen with 100% probability)
  • A "risky" option: a larger reward with lower probability (e.g., 300 yen with 33% probability)
Before the task, participants had a face-to-face meeting with a same-gender observer. During the task, when participants selected the risky option:
  • Success was followed by the observer showing an admiring facial expression
  • Failure was followed by a contemptuous expression
These expressions were shown either using the observer's real face (Human condition) or an avatar face (Avatar condition), switching every 6 to 13 trials (see Figure 2, left).
Results showed that participants made more risky choices in the Avatar condition than in the Human condition, particularly when the risky option had only a slightly higher expected value (see Figure 2, right).
Figure 2. (Left) Risk decision-making task. Participants chose between a "safe" and "risky" option. While making choices, they were observed via camera by an on-screen face—either a real human or an avatar. If they selected the risky option, the outcome (win or no win) was followed by the observer displaying a facial expression: admiration for wins, contempt for losses. The avatar and human face conditions alternated every 6 to 13 trials. When the safe option was selected, the participant received the reward and moved on to the next trial. Reward values and probabilities changed each trial, with one option always being 100%.(Right) Increase in risk-taking under the avatar condition. The red line shows the average risk-taking rate under the Avatar condition, and the blue line under the Human condition. The x-axis represents the expected value difference between the two options. When the risky option was only slightly better (moderate expected value differences), participants were more likely to choose it under the Avatar condition.
Figure 3 Amygdala activity reflecting valuation of feedback uncertainty. The amygdala encoded individual differences in how uncertain facial feedback was interpreted, explaining variations in risk-taking behavior across participants.
The increase in risky decision-making was explained by a more favorable valuation of the uncertainty of which facial expression would be shown. This valuation was more positive in the Avatar condition.
fMRI analysis further revealed that this valuation of uncertainty was associated with activity in the amygdala (see Figure 3).

Future Prospects

This study demonstrates that avatar-mediated communication influences human risk-taking behavior and that this effect is linked to the amygdala. The researchers plan to further investigate how factors such as the avatar's gender or age, individual personality traits, and other types of decision-making tasks are influenced by avatars.
They also aim to explore how avatars can be effectively used to support decision-making in real-world settings such as education and interpersonal support, while also identifying potential risks associated with avatar use.

Article information

Authors: Toshiko Tanaka, Masahiko Haruno
Title: Feedback from an avatar facilitates risk-taking by modulating the amygdala response to feedback uncertainty
Journal: PLOS Biology
DOI: 10.1371/journal.pbio.3003122

This research was supported in part by the following programs: JST Moonshot R&D Program “Realization of a society where people are free from limitations of body, brain, space, and time by 2050,” project: “Realizing a society where avatars coexist and empower all individuals”,  JST CREST Program “Exploration and application of multi-sensing systems in biological environments,” project: “Decoding multi-world predictive coding in cyber society” (Principal Investigator: HARUNO Masahiko) and Grant-in-Aid for Transformative Research Areas (A) “Decoding and manipulating brain dynamics that induce early behavioral changes, opening up multi-layered biology.”

All experiments in this study were approved by the NICT ethics committee. Informed consent was obtained from all participants after explaining the procedures of the study.
Illustrations in the background section were adapted from images generated using Microsoft Copilot.

Appendix

Details of the experimental findings

Risk decision-making Task

Participants (Behavioral experiment: 28; fMRI experiment: 51) performed a risk decision-making task while being observed via camera by a same-gender peer (referred to as the “observer”). In each trial, participants were presented with two options:
  • Risky Option: High reward, low probability of success
  • Safe Option: Lower reward, guaranteed outcome
The reward amounts and probabilities varied across trials. Participants were instructed to choose the option they preferred each time. 
Participants were also informed in advance that the observer would evaluate each of their choices from multiple perspectives and record these evaluations on a designated form.
We hypothesized that the uncertainty surrounding facial expression feedback about which expression might appear would differ depending on whether the face was the observer’s real face or an avatar. To test this, the observer responded with admiration when the participant's risk choice was successful, and contempt when it failed. These expressions were displayed via monitor and alternated between the observer’s real face and a pre-recorded avatar face every 6 to 13 trials. 
Four avatars (two males and two females, all of average attractiveness) were used; each participant was assigned one avatar throughout (see Figure 2). Although observers were physically located outside the MRI scanner room, the facial expressions were pre-recorded and edited to appear natural.
In analyzing behavior, the risk-taking rate was defined as the frequency with which the risky option was chosen. We examined how this rate varied as a function of the difference in expected value between the two options, comparing the real-face (human) and avatar-face (avatar) conditions (see Figure 2, right). Across both the behavioral and fMRI experiments, participants chose the risky option more often under the avatar condition—particularly in trials where the expected value difference was moderate. Figure 2 (right) shows behavioral data from 28 participants.

Detailed Analysis of Behavioral Data

To understand the factors influencing risk-taking, we built a behavioral model incorporating:
  1. Monetary conditions,
  2. Facial feedback conditions, and
  3. The outcome of the previous trial.
The model assumed that participants evaluate the value of each option and choose the one with the higher perceived value. Using the full model and comparing based on information criteria (AIC, BIC) , we identified the most explanatory model.

Full Model Equations:
Value of Risky Option = βr * Risk-related monetary component + βf * Facial feedback component
+ βp * Previous trial outcome
Value of Safe Option = βs * Safe-related monetary component
Each β coefficient reflects how much weight an individual assigns to each factor.
Model components:
  • Risk-related monetary component: Reward amount if successful, reward variance, and forgone safe reward
  • Facial feedback component: Joy from praise, aversion to disdain, and uncertainty about which expression will appear (quantified using Shannon entropy)
  • Previous trial outcome: Whether the previous risk choice was a success or failure
  • Safe-related monetary component: Reward amount for choosing the safe option

The best model was identified by using the maximum likelihood estimation and included the following components.
Best Model Equations:
Value of Risky Option = βr * Reward for successful risky choice + βfu * Uncertainty of facial feedback
Value of Safe Option = βs * Safe reward
This model required estimating separate βfu (weight for facial feedback uncertainty) values for the avatar and real-face conditions. Importantly, the difference in βfu across the two conditions significantly correlated with the difference in risk-taking rates (see Figure 4). In other words, when participants perceived higher value of uncertainty in avatar facial feedback, they tended to take more risks.
Figure 4 Correlation between valuation of uncertainty and risk-taking rate
For each participant, the difference in valuation of uncertainty (x-axis) is plotted against the difference in risk-taking rate (y-axis). Each dot represents one participant.

Psychological Traits That Explain Increased Risk-Taking

Given the observed difference in valuation of facial feedback uncertainty between avatar and real-face conditions, we investigated whether this difference could be explained by personality traits. We focused on traits related to anxiety and interpersonal sensitivity, using the following psychological scales:
Figure 5 Relationship between personality traits and valuation of uncertainty
Using questionnaire data, we plotted participants' scores for interpersonal sensitivity, social anxiety, and general anxiety against the difference in their valuation of uncertainty. Only the Interpersonal Reactivity Index (IRI) showed a significant correlation with the uncertainty valuation difference.
Results showed that only the IRI scores significantly correlated with differences in facial feedback uncertainty valuation (see Figure 5, left). No significant correlations were found with STAI or LSAS (see Figure 5, center and right). This suggests that the degree of concern or empathy for others plays a key role in how facial feedback uncertainty is perceived, particularly when comparing avatars to real faces.

fMRI Analysis of Brain Activity

To examine the neural basis of increased risk-taking in the avatar condition, we focused on brain activity at the moment participants were presented with the choice options. We used the difference in βfu (valuation of facial feedback uncertainty) as a behavioral index, reflecting how much a participant's valuation differed between avatar and real-face conditions.
We hypothesized that individuals with a larger βfu difference would show corresponding differences in brain regions encoding uncertainty. Results revealed that those who valued facial feedback uncertainty higher in the avatar condition exhibited reduced amygdala activity in response to uncertainty during the avatar condition. This implies that increased risk-taking under avatars may stem from decreased amygdala response to uncertainty.

To further test this, we divided participants into:
  • Group A: βfu (Avatar) > βfu (Human)
  • Group H: βfu (Human) > βfu (Avatar)
Group A was larger than Group H. In Group A, amygdala activity was negative in response to facial feedback uncertainty during the avatar condition, but positive during the real-face condition. In Group H, amygdala activity was negative during the real-face condition (see Figure 6). This pattern suggests that lower amygdala activity in response to uncertainty is associated with greater risk-taking behavior.
Figure 6 Amygdala activity reflects individual differences in valuation of facial feedback uncertainty
In the amygdala, the difference in brain activity between the avatar and human face conditions in response to feedback uncertainty correlated with the behavioral difference in uncertainty valuation (left). A scatter plot of behavioral uncertainty valuation differences (x-axis) and corresponding amygdala activity differences (y-axis) shows this relationship (middle). When participants were grouped by whether their valuation of uncertainty was greater for avatars (Group A) or human faces (Group H), both groups showed negative amygdala responses in the condition where uncertainty valuation was higher (right).

Other Brain Regions Involved

We extended the analysis to other brain areas and found that:
also exhibited activity differences between the avatar and real-face conditions that correlated with βfu differences.
However, unlike the amygdala, these regions did not show significant condition-dependent activity differences in Group A alone. These results suggest that while the amygdala plays a pivotal role in driving increased risk-taking via the valuation of uncertainty in the avatar condition, the ventral striatum and vACC are more generally involved in decision-making under risk and may interact closely with the amygdala.
Figure 7. Other brain regions showing correlation with facial feedback uncertainty valuation
Similar to Figure 6, we identified brain regions where the difference in uncertainty response between the avatar and human conditions correlated with the individual differences in valuation of facial feedback uncertainty. These regions included the ventral striatum (top) and the ventral anterior cingulate cortex (vACC; bottom). The scatter plots (middle) show the correlation between behavioral uncertainty valuation difference (x-axis) and neural activity difference. When focusing on Group A (those who valued avatar uncertainty higher), no clear activity difference between avatar and human conditions was found in these regions.

Glossary

Amygdala

A structure found in the limbic system of the brain, shared across many animal species. It plays a key role in processing emotions such as fear and anxiety, as well as in interpreting facial expressions. (Illustrated as the almond-shaped blue region.)


fMRI Experiment

A non-invasive imaging technique that measures brain activity by detecting changes in blood flow in specific brain regions. It is widely used in neuroscience and medical research to study the function of various brain areas in detail.


Information Criteria (AIC/BIC)

AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) are statistical measures used to evaluate and compare models. They help determine which model best explains the data.


Shannon Entropy

A measure of uncertainty in information theory. It quantifies how unpredictable events are.
Entropy = -∑ [Probability of an event × log(Probability of that event)]
Here, ∑ denotes summation over all possible events. In this study, the events are “admiration” and “contempt” facial expressions. Entropy is maximized when all events are equally likely.


Maximum Likelihood Estimation

Maximum likelihood estimation is a statistical method used to find the set of parameters in a model that makes the observed data most probable. This approach helps determine the most appropriate model based on the data.


State-Trait Anxiety Inventory (STAI)

A psychological questionnaire that assesses anxiety in two forms:
  • State anxiety: Temporary anxiety felt in a specific situation
  • Trait anxiety: A person’s general tendency to feel anxious
    It is widely used to evaluate levels of anxiety.


Liebowitz Social Anxiety Scale (LSAS)

A self-report measure used to assess the severity of social anxiety disorder. It evaluates the degree of fear and avoidance experienced in various social situations.


Interpersonal Reactivity Index (IRI)

A comprehensive questionnaire that assesses both emotional and cognitive aspects of interpersonal reactivity. It consists of four subscales:
  1. Empathic Concern - sympathy and concern for others
  2. Perspective Taking - ability to adopt others’ viewpoints
  3. Personal Distress - feelings of anxiety or discomfort when witnessing others' suffering
  4. Fantasy - tendency to imaginatively project oneself into fictional characters’ experiences
Each item is rated on a 5-point scale from “Does not describe me well” (1) to “Describes me very well” (5).

1. Empathic Concern
I often have tender, concerned feelings for people less fortunate than me.
Sometimes I don't feel very sorry for other people when they are having problems.
When I see someone being taken advantage of, I feel kind of protective towards them.
Other people's misfortunes do not usually disturb me a great deal.
When I see someone being treated unfairly, I sometimes don't feel very much pity for them.
I am often quite touched by things that I see happen.
I would describe myself as a pretty soft-hearted person.

2. Perspective Taking
I sometimes find it difficult to see things from the "other guy's" point of view.
I try to look at everybody's side of a disagreement before I make a decision.
I sometimes try to understand my friends better by imagining how things look from their perspective.
If I'm sure I'm right about something, I don't waste much time listening to other people's arguments.
I believe that there are two sides to every question and try to look at them both.
When I'm upset at someone, I usually try to "put myself in his shoes" for a while.
Before criticizing somebody, I try to imagine how I would feel if I were in their place.

3. Personal Distress
In emergency situations, I feel apprehensive and ill-at-ease.
I sometimes feel helpless when I am in the middle of a very emotional situation.
When I see someone get hurt, I tend to remain calm.
Being in a tense emotional situation scares me.
I am usually pretty effective in dealing with emergencies.
I tend to lose control during emergencies.
When I see someone who badly needs help in an emergency, I go to pieces.

4. Fantasy
I daydream and fantasize, with some regularity, about things that might happen to me.
I really get involved with the feelings of the characters in a novel.
I am usually objective when I watch a movie or play, and I don't often get completely caught up in it.
Becoming extremely involved in a good book or movie is somewhat rare for me.
After seeing a play or movie, I have felt as though I were one of the characters.
When I watch a good movie, I can very easily put myself in the place of a leading character.
When I am reading an interesting story or novel, I imagine how I would feel if the events in the story were happening to me.


Ventral Striatum

The striatum is a core component of the basal ganglia in the brain, involved in decision-making, emotional and behavioral regulation, learning, memory, and motor control. The ventral portion includes the nucleus accumbens, which is crucial for reward and emotion processing.


Ventral Anterior Cingulate Cortex (vACC)

A region on the medial surface of the frontal lobe, shaped like a collar, involved in reward anticipation, emotional regulation, empathy, and decision-making.

Technical Contact

HARUNO Masahiko
Neural Information Engineering Laboratory
Center for Information and Neural Networks
Advanced ICT Research Institute

Media Contact

Press Office
Public Relations Department