{"data":{"pictureId":"394DBB97-B3E0-4AC3-A67E-3A85F48CC9A3","playId":"","activityName":"","appversion":"3.4.1","os":"ios","stickerId":"","filterId":"","infoStickerId":"","imageEffectId":"","product":"retouch"},"source_type":"hypic","tiktok_developers_3p_anchor_params":"{\"filter_id\":[],\"template_id\":\"\",\"client_key\":\"awgvo7gzpeas2ho6\"}"}
The University of Michigan has unveiled a groundbreaking artificial intelligence model, Be.FM, designed to predict and understand human behavior. Developed in collaboration with Stanford University and MobLab, this AI system stands apart from general-purpose models like GPT or LLaMA by focusing specifically on human actions. Be.FM is trained on a vast dataset drawn from behavioral science, including controlled experiments and academic studies, making it one of the first of its kind.
As technology continues to evolve, the need for machines to interpret human behavior becomes increasingly critical. Consider a self-driving car navigating busy city streets or an investment algorithm predicting market trends. Both scenarios require a deep understanding of human actions and reactions, a capability that Be.FM aims to provide.
Revolutionizing Behavioral Predictions
Unlike traditional AI models that rely on generic text corpuses, Be.FM is enriched with data from over 68,000 experimental subjects, 20,000 survey respondents, and numerous scientific studies. This specialized training enables Be.FM to excel where other AIs falter, particularly in recognizing minority behaviors and complex social cues.
“We’re not feeding it Wikipedia,” explained Yutong Xie, a doctoral student in information science at U-M and the study’s lead author. “We built a behavioral dataset to help the model reason about why people act the way they do.”
Be.FM’s capabilities extend across four key application areas. Its most visible strength lies in predicting human behavior in real-world scenarios. For instance, it can forecast which investment options individuals might prefer, aiding economic modeling, product testing, and public policy analysis.
Understanding Psychological and Demographic Insights
Beyond behavioral predictions, Be.FM can deduce psychological traits and demographic information from behavior or background data. This ability to infer traits such as extroversion or agreeableness based on demographic factors could revolutionize user segmentation, personalized interventions, and product design.
Human behavior often shifts with context, influenced by changes in timing, social norms, or environmental signals. Be.FM can identify these contextual drivers, offering insights into why behaviors change and how to respond effectively. For example, it can analyze shifts in app usage patterns, pinpointing factors like design updates or seasonal trends as potential influences.
Enhancing Research and Development
Be.FM is also a powerful tool for organizing and applying behavioral science knowledge. Built on a large language model architecture, it can generate research ideas, summarize literature, and solve applied behavioral economics problems. This makes it invaluable for scholars and practitioners planning studies or simulating scenarios before field testing.
Across these categories, Be.FM consistently outperforms commercial and open-source models like GPT-4o and LLaMA, particularly in tasks such as personality prediction and scenario simulation. Its predictions more closely reflect real-world patterns, especially at the population level.
“Behavior in health, education, even geopolitics—the goal is to make Be.FM useful wherever people make decisions,” said Qiaozhu Mei, U-M professor of information and the corresponding author of the study.
Future Prospects and Limitations
Despite its impressive capabilities, Be.FM has limitations. Its performance beyond the four key areas remains untested, and it is not yet designed to forecast large-scale political events or predict outcomes like elections or peace deals. However, the research team is already working to expand Be.FM’s domain coverage.
The Be.FM models are available upon request, and the team invites researchers and practitioners to use the model and provide feedback. This collaborative approach aims to refine and enhance Be.FM’s capabilities, ensuring it remains at the forefront of AI-driven behavioral prediction.
As AI continues to integrate into daily life, models like Be.FM represent a significant advancement in understanding and predicting human behavior, offering new possibilities for technology, research, and policy development.
