Skip to Content

Tracking the Brain’s Ability to Bluff

Brain-imaging experiments show how bargaining games could provide insights into psychiatric disorders.
November 2, 2010

The ability to infer another person’s state of mind—including his or her perception of you—is an integral component of human interaction. To make business deals, maintain good relationships, and to play a winning hand of poker, we must be able to surmise to some extent what a client, spouse, or friend is thinking. But that skill can go awry in a number of psychiatric disorders, including autism and borderline personality disorder.

Bluffing brain: People who choose to bluff when playing a simple bargaining game show more activation in part of the brain called the dorsal prefrontal cortex (yellow).

Read Montague, a neuroscientist at Baylor College of Medicine, uses a combination of brain imaging and interactive games to explore this skill, with the long-term goal of developing new diagnostic tests for psychiatric disorders.

“This is an extremely promising approach to identifying the mechanisms that underpin these disorders,” says Peter Fonagy, a psychiatrist at University College London who has collaborated with Montague in the past. “Psychiatry is the last medical specialty where the symptoms are equivalent to a diagnosis.”

In a study published today, Montague and collaborators found that people take one of three strategies when playing a simple economics game, and that specific parts of the brain seem to be more active in people who choose to bluff. A second paper published last month shows how the strategies chosen by healthy people playing a similar game change depending on the mental status of their opponent. Researchers ultimately hope to create an automated version of this approach and use it to diagnose disease.

“The capacity that breaks down the most in mental illness is ‘social software,’ such as the ability to pick up signals from people in groups and to collaborate,” says Montague. “We only poorly understand how the brain implements these things, or how it can break.”

Montague and others in the field of neuroeconomics—a relatively young branch of neuroscience that explores how the brain makes decisions—assess how people make choices by asking them to play interactive games. While early studies looked at how factors such as risk and reward contribute to decision-making, more recent research has focused on how people have different approaches to decision making, says Scott Huettel, a neuroscientist at Duke University, who not involved in the research. “You can see that different folks will systematically approach different problems in different ways,” he says.

Better understanding these differences could have a variety of applications. “If you’re trying to sell a product, you don’t want to assume everyone is the same,” says Huettel. “You might want to produce information in different ways based on the idea that people will respond to it differently.” A clearer picture of the spectrum of ways in which healthy people act will also help illuminate how these decision-making skills fall apart in mental illness.

In a paper published today in the journal Proceedings of the National Academy of Sciences, Montague and his collaborators asked people to play a simplified bargaining game. One person, the buyer, is told her own private value of a hypothetical object. The buyer suggests a price to the second player, the seller, who responds by naming a sale price. If the seller’s price is lower than the buyer’s, the trade goes forward. Just like in the real marketplace, the seller wants to charge the highest price but risks losing the sale if the price is too high, while the buyer wants the lowest price. The exchange repeats for a number of sessions, with neither player getting feedback on the outcomes.

Montague and collaborators grouped players according to three broad strategies. “Incrementalists” generally gave the seller a value that was proportional to their private value, suggesting they trusted the other player. “Conservatives” tried to maximize their gains by suggesting mid-range values regardless of the actual value. “Strategists” attempted to deceive the other player, suggesting a value that was inversely related to the given value.

The strategists’ approach required the greatest mentalizing—trying to get into the mind of the other player. To their opponent, their choices resembled those of an incrementalist, suggesting a trustworthy series of suggested prices. But the reality was the opposite; they suggested relatively high prices when the actual value was low, likely stopping that sale. But they surmised that this would increase their credibility in the other rounds of the game; thus, when the private value was high, they could give low suggested prices.

“Bluffing is a specific feature of theory of mind,” explains Montague. “It touches on the capacity to model other people; included in your model of me is your model of my model of you.” He notes that while none of the groups had significant differences in IQ, having an above-average IQ was necessary to be a strategist.

The researchers recorded brain activity as people played the game and found specific neural signatures for strategists. These people—who made up only about 10 percent of the group—showed greater activation in the dorsal prefrontal cortex, a region involved in cognitive control, as well as in the Brodman area 10, a region known to be involved in thinking about other people, says Montague. In addition, a region of the right temporal lobe “varied with expected payoffs in strategic deceivers but was not in other types of players,” he says.

Hans Breiter, a neuroscientist and psychiatrist at the Martinos Center for Biomedical Imaging, applauds the research for applying sophisticated mathematical techniques to decision-making. Rather than looking at simple statistical relationships, “they are looking for variables that will be seen across different stimuli and different populations,” says Breiter. The ideal variables, he explains, can be easily distinguished from noise and can also be seen at another level, like in the brain.

The ability to quantify these characteristics will be crucial in applying the techniques to psychiatry. Breiter and others hope that the approach, dubbed quantitative psychiatry, will help make diagnosis of mental illness more closely resemble that of infectious disease or high blood pressure. Both of these have multiple potential causes that will each respond best to specific types of treatment.

In a second paper, published last month in PLoS Computational Biology, Montague and colleagues demonstrated how this type of game could be applied to psychiatry. Researchers had pairs of people play a simple trust game in which one player, the “investor,” gave a certain amount of money to the investee. The amount was tripled in plain sight en route to the investee, who then chose how much to give back to the investor. In this case, one player was healthy while the second had one of several neuropsychiatric disorders: borderline personality disorder, major depression, or attention deficit hyperactivity disorder (ADHD.) Borderline personality disorder is characterized by difficulty with relationships and lack of trust, qualities that could be picked up in social interactions.

The pairs played multiple rounds of the game, in order to build a model of their partner. Scientists then created a mathematical model to track how the healthy participants played over time. They found that their characteristics segregated into four different clusters, roughly correlated to the type of psychiatric disorder their partner had. The researchers then showed they could automate the system, employing a computer model, created in a previous study, capable of playing the role of the healthy player. Scientists could again predict which cluster the human player fell into, using only the choices made by the computer model.

“What [Montague] has done is shown proof-of-principle to applying a more sophisticated approach to diagnosis that potentially could be automated,” says Breiter. “I suspect in combination with one or two other quantitative tests, this could be exquisitely sensitive.” For example, he notes, a major depressive disorder is likely made up of a number of different illnesses, “but we don’t have a quantitative metric to explicitly diagnose major depression or to subtype these illnesses.”

Montague, in collaboration with Fonagy, now plans to test this approach on a much greater number of people, with the intention of refining the ability to classify them as either healthy or suffering from a psychiatric disorder. “We hope to use these games to assign quantitative parameters that could be tracked during and post-therapy and suggest the neural systems that could be involved,” says Montague.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.