Human Behavior Prediction

The field predicts individual and group behavior to support applications such as mobility planning, learning support, health monitoring, and human-robot collaboration. Approaches range from cognitive models of memory and decision-making to sequence and trajectory models, multimodal sensing, and language-model agents grounded in surveys, interviews, or behavioral histories. Evaluation spans choices in games and surveys, movement and mobility benchmarks, online learning logs, and robot interaction, with growing attention to uncertainty, calibration, and prediction at different time horizons. A central challenge is that strong individual predictions do not always reproduce population-level patterns: models can miss how people change their beliefs, compress human variation, or reflect training-data biases, while detailed behavioral histories also raise privacy risks.