Task-based value generalization correlates with positive overgeneralization and bipolar traits

Success has a way of changing how we see the world. When we achieve a goal, we often feel a surge of confidence that spills over into other areas of our lives. We might solve a difficult problem at work and suddenly feel capable of tackling challenges in our relationships or hobbies. This feeling is normal, but for some people, this confidence grows too large, expanding far beyond what the evidence supports. In its extreme form, this pattern is a hallmark of mania, a state of elevated mood and boundless energy that characterizes bipolar disorder. Scientists have long suspected that this tendency to overgeneralize success is a key piece of the puzzle in understanding why some people are vulnerable to these mood swings. For years, however, researchers could only study this phenomenon by asking people to describe their own thoughts and feelings. They relied on questionnaires where individuals reported whether they tended to expect success in new situations after a win. While these self-reports are useful, they do not reveal the actual mental machinery at work. They tell us what people believe, but not how their brains are actually processing the link between a specific reward and a new, unfamiliar situation.

To get a clearer look at this process, a team of researchers at the Icahn School of Medicine at Mount Sinai, New York University, and Monash University designed a new way to measure it. They created a simple, computer-based game called "Fossil Hunt" to observe how people learn and generalize value without needing to ask them a single question about their feelings. In this study, 163 participants played the role of paleontologists trying to identify valuable dinosaur fossils. The game presented them with a series of stick-figure shapes that looked like four-legged creatures. At the start, the players learned that one specific shape was always a "valuable" fossil that would earn them points, while a very different shape at the opposite end of the spectrum was worthless. The shapes in between were a smooth gradient, shifting gradually from the valuable one to the worthless one. After mastering this rule, the players were shown new, intermediate shapes they had never seen before and asked to guess if they were valuable. The researchers watched closely to see how far the players' confidence stretched. Did they only value shapes that looked exactly like the winning fossil, or did they extend that value to shapes that were only vaguely similar?

The results revealed a clear and measurable difference in how people's minds worked. The researchers found that some participants had a very narrow view of value; they only considered shapes very close to the winning fossil as valuable. Others had a much broader view, judging shapes that were quite different from the winner as still being valuable. This "breadth of generalization" was not just a random quirk of the game. It turned out to be a powerful predictor of how people described themselves on psychological surveys. Those who showed the broadest generalization in the game were the same people who scored highest on a scale measuring positive overgeneralization—the tendency to expect success in ambitious goals and social situations after a win. Furthermore, this behavioral measure was linked to traits associated with bipolar disorder, even in people who did not have a clinical diagnosis. The study also looked at how fast people made their decisions. Those with broader generalization made their judgments quickly, even when the shapes were quite different from the winner. In contrast, people with a narrower view took longer to decide on the intermediate shapes, hesitating as the similarity to the winner faded. This hesitation suggests that for them, the link between the reward and the new shape was weak, whereas for the overgeneralizers, the link remained strong and immediate.

To understand what might be driving these differences, the researchers ran computer simulations based on a model of how agents learn. They tested a concept called self-efficacy, which is essentially a person's belief in their ability to reach the best possible outcome. In their model, they adjusted how much weight the computer agent placed on the best potential future versus the worst potential future. When the agent was programmed to focus heavily on the best possible outcome, it learned to spread the value of a reward much more widely across different states, mimicking the broad generalization seen in the human participants. When the agent focused more on the worst outcome, its learning became conservative and narrow. This suggests that the way people anticipate the future—specifically, how much they believe they can achieve the best result—might be the engine that drives how broadly they generalize success.

The study does not claim to have solved the mystery of mania or to have found a cure. Instead, it offers a new, objective tool to measure a cognitive style that has previously been accessible only through self-report. By showing that a simple behavioral task can capture the same traits as complex questionnaires, the researchers provide a way to study these processes more directly. The findings suggest that the tendency to overgeneralize reward is a stable trait linked to the bipolar spectrum, rather than just a temporary reaction to a bad or good day. While the study was conducted on a general online sample and not on people with diagnosed bipolar disorder, the connection between broad reward generalization and subclinical traits offers a promising path for future research. It points toward a specific mechanism—how we weigh our potential for future success—that could help explain why some people's confidence expands to fill every corner of their lives, while others keep it tightly contained.