Brain network model can predict when people will feel surprised

Surprise is a key human emotion that is typically felt when something that we are witnessing or experiencing differs from our expectations. This natural human response to the unexpected has been the focus of numerous psychology studies, which uncovered some of its underlying neural processes.

Researchers at the University of Chicago have developed a brain network model that can predict people’s surprise. In a paper published in Nature Human Behaviour, they showed that this model generalized well across various tasks, predicting the surprise of individuals who were performing a task or watching different videos containing unexpected elements.

The study carried out by these researchers builds on previous research focusing on surprise. Earlier work found that humans experience surprise when reality clashes with their expectations in many different situations. Some of these past works discovered patterns of brain activity associated with each specific experience of surprise.

“The first goal of the paper was to look at whether we can use the brain as a common space to understand our experiences,” Ziwei Zhang, co-author of the paper, told Medical Xpress.

“If my brain shows a similar pattern when I’m surprised while I learn a task to yours, when you are surprised while you watch a basketball game, it tells us that these very different experiences share something in common and that the brain responds to these expectation violations similarly.”

After reviewing previous research that explored how the human brain responds to the unexpected, Zhang and her colleague Dr. Monica Rosenberg set out to determine whether these brain responses could be used to predict when individuals will be surprised across different contexts. Essentially, their objective was to correctly guess whether people would be surprised at a given moment by examining their brain activity.

“We used three functional magnetic resonance imaging (fMRI) datasets collected and shared by other research groups,” explained Zhang.

“In one dataset, shared by Dr. Joseph McGuire and colleagues, volunteers performed a learning task in the MRI scanner. Their goal was to learn where a cartoon bag containing coins would appear on a computer screen. The better they learned the task, the more coins they could ‘catch’ in the game.”

In this previous work by Dr. McGuire and his collaborators, participants were found to learn to predict where the bag would appear on a computer screen. In some instances, however, the bag would appear in unexpected locations and the researchers assumed that in these instances people felt surprised.

“The second dataset we analyzed was collected and shared by Dr. James Antony and colleagues,” said Zhang.

“For this study, volunteers watched basketball game videos during fMRI. Surprise is higher when belief about who is going to win the game changes, informed by scores and strength of the teams. Lastly, Dr. Shari Liu and colleagues collected and shared fMRI data collected while people watched cartoons in which characters took more or less surprising actions.

“Examples of surprising actions included when characters changed their goals (for example, heading in an unpredicted direction) or acted inefficiently (for example, making an exaggerated move when circumventing an obstacle).”

To predict the surprise of participants who took part in these three distinct studies, Zhang and Rosenberg created a new brain network model, dubbed the surprise edge-fluctuation-based predictive model (EFPM). This model was designed to track fluctuations in the interactions between different parts of the brain to predict when people experience surprise.

“If we think of the brain as a system where different regions interact with each other, we are already thinking about the brain as a network,” explained Zhang. “In this paper, we built a model that predicts how surprised someone is based on their brain network configuration at that time.”

As a first step in the development of their model, the researchers identified the brain interactions that predicted surprise during the learning task employed by Dr. McGuire and his collaborators.

Subsequently, they tried to determine whether the strength of these same brain interactions could be used to predict the surprise of participants of the other two previous works they examined, in which people were asked to watch videos of basketball games and cartoons, respectively.

“We think that there are a few implications of this work,” said Zhang.

“For one, the brain model we built was able to predict surprise in a novel person doing the same task (i.e., the learning task), and a different group of people doing something completely different (i.e., watching a basketball game). In other words, we can look at data from a totally new person doing something completely different and guess, with some degree of accuracy, how surprised they are.”

Overall, the findings gathered by Zhang and Rosenberg suggest that some parts of the human brain respond to a violation of expectations irrespective of what people are doing or what contexts they are in.

Using their model, they were able to predict the surprise felt by individuals simply using neuroimaging data. Although predictions were by no means perfect, they were more accurate than the researchers would expect by chance alone.

“We think that the implications of this model reach beyond just surprise,” said Zhang. “The method we developed can be used as a framework for predicting other experiences as well, like how attentive someone is or how happy they are feeling.”

This recent study could soon inspire new research aimed at predicting human emotions from recorded brain activity. As their analyses relied on datasets compiled by other research groups, Zhang and Rosenberg feel that it also emphasizes the benefits of open science and of making data or code available to others.

“Our work was possible because scientists like Dr. Joseph McGuire, Dr. James Antony, Dr. Shari Liu, Dr. Joshua Faskowitz and their colleagues made their datasets and scripts available for the community,” explained Zhang.

“We are thankful that we could be part of the activity of demonstrating how open data can be helpful for researchers in the community.”

In their next studies, the researchers plan to assess the ability of their model to predict surprise across an even broader range of contexts. For instance, they would like to explore its predictive abilities in instances where individuals are listening to stories, musical compositions and even interacting with others in social settings.

“So far, we [have] talked about the generalizable nature of the model, meaning that this model is sensitive to the ‘shared’ neural signatures of surprise across people and contexts,” added Zhang.

“In another line of work, we are also interested in investigating how and why people are surprised at different times, and how this affects what they remember and learn. For example, do individual differences in surprise as they listen to stories predict how people remember the stories differently? We believe that understanding both the ‘shared’ and ‘unique’ aspects of expectation violation are important.”

© 2025 Science X Network

Surprise is a key human emotion that is typically felt when something that we are witnessing or experiencing differs from our expectations. This natural human response to the unexpected has been the focus of numerous psychology studies, which uncovered some of its underlying neural processes.

Researchers at the University of Chicago have developed a brain network model that can predict people’s surprise. In a paper published in Nature Human Behaviour, they showed that this model generalized well across various tasks, predicting the surprise of individuals who were performing a task or watching different videos containing unexpected elements.

The study carried out by these researchers builds on previous research focusing on surprise. Earlier work found that humans experience surprise when reality clashes with their expectations in many different situations. Some of these past works discovered patterns of brain activity associated with each specific experience of surprise.

“The first goal of the paper was to look at whether we can use the brain as a common space to understand our experiences,” Ziwei Zhang, co-author of the paper, told Medical Xpress.

“If my brain shows a similar pattern when I’m surprised while I learn a task to yours, when you are surprised while you watch a basketball game, it tells us that these very different experiences share something in common and that the brain responds to these expectation violations similarly.”

After reviewing previous research that explored how the human brain responds to the unexpected, Zhang and her colleague Dr. Monica Rosenberg set out to determine whether these brain responses could be used to predict when individuals will be surprised across different contexts. Essentially, their objective was to correctly guess whether people would be surprised at a given moment by examining their brain activity.

“We used three functional magnetic resonance imaging (fMRI) datasets collected and shared by other research groups,” explained Zhang.

“In one dataset, shared by Dr. Joseph McGuire and colleagues, volunteers performed a learning task in the MRI scanner. Their goal was to learn where a cartoon bag containing coins would appear on a computer screen. The better they learned the task, the more coins they could ‘catch’ in the game.”

In this previous work by Dr. McGuire and his collaborators, participants were found to learn to predict where the bag would appear on a computer screen. In some instances, however, the bag would appear in unexpected locations and the researchers assumed that in these instances people felt surprised.

“The second dataset we analyzed was collected and shared by Dr. James Antony and colleagues,” said Zhang.

“For this study, volunteers watched basketball game videos during fMRI. Surprise is higher when belief about who is going to win the game changes, informed by scores and strength of the teams. Lastly, Dr. Shari Liu and colleagues collected and shared fMRI data collected while people watched cartoons in which characters took more or less surprising actions.

“Examples of surprising actions included when characters changed their goals (for example, heading in an unpredicted direction) or acted inefficiently (for example, making an exaggerated move when circumventing an obstacle).”

To predict the surprise of participants who took part in these three distinct studies, Zhang and Rosenberg created a new brain network model, dubbed the surprise edge-fluctuation-based predictive model (EFPM). This model was designed to track fluctuations in the interactions between different parts of the brain to predict when people experience surprise.

“If we think of the brain as a system where different regions interact with each other, we are already thinking about the brain as a network,” explained Zhang. “In this paper, we built a model that predicts how surprised someone is based on their brain network configuration at that time.”

As a first step in the development of their model, the researchers identified the brain interactions that predicted surprise during the learning task employed by Dr. McGuire and his collaborators.

Subsequently, they tried to determine whether the strength of these same brain interactions could be used to predict the surprise of participants of the other two previous works they examined, in which people were asked to watch videos of basketball games and cartoons, respectively.

“We think that there are a few implications of this work,” said Zhang.

“For one, the brain model we built was able to predict surprise in a novel person doing the same task (i.e., the learning task), and a different group of people doing something completely different (i.e., watching a basketball game). In other words, we can look at data from a totally new person doing something completely different and guess, with some degree of accuracy, how surprised they are.”

Overall, the findings gathered by Zhang and Rosenberg suggest that some parts of the human brain respond to a violation of expectations irrespective of what people are doing or what contexts they are in.

Using their model, they were able to predict the surprise felt by individuals simply using neuroimaging data. Although predictions were by no means perfect, they were more accurate than the researchers would expect by chance alone.

“We think that the implications of this model reach beyond just surprise,” said Zhang. “The method we developed can be used as a framework for predicting other experiences as well, like how attentive someone is or how happy they are feeling.”

This recent study could soon inspire new research aimed at predicting human emotions from recorded brain activity. As their analyses relied on datasets compiled by other research groups, Zhang and Rosenberg feel that it also emphasizes the benefits of open science and of making data or code available to others.

“Our work was possible because scientists like Dr. Joseph McGuire, Dr. James Antony, Dr. Shari Liu, Dr. Joshua Faskowitz and their colleagues made their datasets and scripts available for the community,” explained Zhang.

“We are thankful that we could be part of the activity of demonstrating how open data can be helpful for researchers in the community.”

In their next studies, the researchers plan to assess the ability of their model to predict surprise across an even broader range of contexts. For instance, they would like to explore its predictive abilities in instances where individuals are listening to stories, musical compositions and even interacting with others in social settings.

“So far, we [have] talked about the generalizable nature of the model, meaning that this model is sensitive to the ‘shared’ neural signatures of surprise across people and contexts,” added Zhang.

“In another line of work, we are also interested in investigating how and why people are surprised at different times, and how this affects what they remember and learn. For example, do individual differences in surprise as they listen to stories predict how people remember the stories differently? We believe that understanding both the ‘shared’ and ‘unique’ aspects of expectation violation are important.”

© 2025 Science X Network