Neuroforecasting: New research shows brain activity can predict crowd preferences

A new study published in PNAS Nexus provides evidence that brain activity can more accurately predict broader market behavior than self-reported preferences or observed choices—especially when research participants are not representative of the general population. The researchers found that neural responses tied to emotional reactions generalized across individuals and provided more consistent forecasts of what products or content would succeed in the market.

Most market predictions rely on behavioral data, assuming that what a sample of people chooses or reports will reflect the broader population. But this approach can fail if the sample is too small or not demographically matched to the public. In this new study, researchers explored whether signals from the brain—specifically those associated with emotional reactions—might be more reliable across different people. Their results showed that brain activity, especially in a region called the nucleus accumbens, consistently predicted market-level choices, even when traditional behavioral forecasts faltered.

The research builds on previous findings in a field known as neuroforecasting, which has shown that brain activity can sometimes outperform behavior and self-report in predicting real-world outcomes like music sales, advertising success, and even which news stories go viral. But these earlier studies mostly demonstrated that neuroforecasting could work—they didn’t explain why or under what conditions. This new research sought to uncover the underlying processes that make brain-based forecasting more generalizable than behavioral prediction, especially when sample demographics vary.

“Forecasting what people will choose at a large scale is critical in business and public policy. However, traditional tools like surveys or small-sample behavior often don’t predict real-world outcomes well. We wanted to know if brain activity can offer improved forecasts. And more importantly, when and why it might work better,” said study author Alexander Genevsky, an associate professor of marketing and consumer neuroscience at Erasmus University.

The researchers grounded their approach in a decision-making model called the Affect Integration Motivation (AIM) framework. This model describes how people first respond to stimuli with quick emotional reactions, which are then processed more deliberately to arrive at a final decision. These early emotional responses are often shared across individuals and show up in specific areas of the brain. In contrast, the more reflective stages of decision-making—those that weigh personal memories or context—tend to be more individual and varied. The team hypothesized that these shared emotional components would offer more reliable signals for predicting broader market behavior.

To test their theory, the researchers conducted two experiments that combined neuroimaging and behavioral data with large-scale internet-based preference studies. In the first experiment, 32 participants viewed descriptions of crowdfunding projects while undergoing brain scans. In the second, 33 participants viewed short video clips. In both cases, the participants made real choices—whether to fund a project or continue watching a video. Their neural activity was measured while they made these decisions.

Separately, large internet samples of nearly 3,000 participants for the crowdfunding study and about 1,000 for the video-viewing study were asked to evaluate the same content. The researchers then used brain and behavioral data from the small laboratory samples to predict what the larger internet-based markets would prefer. They also divided the internet participants into groups that were more or less demographically similar to the lab samples to test how well forecasts generalized across different populations.

The key finding was that activity in the nucleus accumbens—a brain region associated with positive anticipation and emotional engagement—consistently predicted which projects or videos would be more popular in the broader samples. This predictive power held even when the lab sample was not demographically similar to the internet sample.

“There is information in the brain that reveals preferences, not just of the individual, but of larger groups,” Genevsky told PsyPost. “We found that certain patterns of brain activity, especially those linked to emotional reactions, can predict what large groups of people will choose, even when their background or demographics are quite different.”

In contrast, behavioral predictions based on participants’ choices or self-reports only worked well when the lab and internet samples were closely matched. Predictions based on another brain area, the medial prefrontal cortex—which is linked to more reflective and individualized thinking—did not consistently forecast market preferences.

The strength of the nucleus accumbens signal in predicting market choices was supported by further analyses. In both experiments, neural forecasts remained significant across all demographic quartiles of the internet sample, whereas behavioral forecasts declined as representativeness decreased.

Bootstrapped simulations showed that this pattern was robust in over 96% of analytic iterations. Moreover, the researchers found that brain activity in the nucleus accumbens was more consistent across individuals in the lab than activity in the medial prefrontal cortex, further supporting its role as a shared, generalizable signal.

Another key takeaway was that strong neural forecasts could be achieved with relatively small sample sizes. In both studies, brain activity from just 20 to 25 participants was enough to produce reliable predictions of market behavior.

In contrast, behavioral predictions remained inconsistent even as the sample size increased. This has practical implications, as neuroimaging studies are often considered too expensive or resource-intensive for widespread use. The study shows that relatively small and inexpensive neural datasets can still add significant forecasting value.

“We expected brain data to help, but we were surprised by how consistently it worked, even when our lab participants were quite different from the larger population,” Genevsky said. “In contrast, behavior and survey responses worked well only when our samples closely matched the broader market. It was striking that the brain signals held up across different groups.”

While these findings highlight the unique predictive power of neural signals, the study does have limitations. The two experiments focused on entertainment-related content—crowdfunding appeals and online videos—that likely evoke strong emotional responses. It is unclear whether the same brain-based forecasting methods would apply to different kinds of decisions, such as those involving financial risk or ethical trade-offs. The researchers note that future studies should test how different types of brain activity relate to various kinds of markets, such as those based on fear, long-term planning, or social influence.

“Brain imaging is still relatively expensive and time-consuming, and it’s not a silver bullet,” Genevsky noted. “Our findings are most relevant for decisions involving emotional reactions. Additional research is exploring if the same holds for more rational or complex choices. Also, we tested only two types of markets, so more research is needed to see how broadly these findings apply.”

Another open question is whether certain individuals consistently generate more predictive brain signals than others. If so, identifying these “neural superforecasters” could further reduce the cost and complexity of neuroforecasting.

“We want to better understand which types of brain signals matter for which kinds of decisions and how to make these tools more scalable,” Genevsky explained. “We’re also interested in partnering with organizations to apply these methods in new settings, like health messaging or sustainability campaigns.

“One encouraging takeaway is that you don’t need a massive neuroscience lab to apply these ideas. Our results suggest that even relatively small brain samples (i.e., under 40 people) can provide meaningful insight into what will appeal to a much larger audience.”

The study, “Neuroforecasting reveals generalizable components of choice,” was authored by Alexander Genevsky, Lester C. Tong, and Brian Knutson