Garza, R.; El Hady, A.; Bahl, A.

Animals integrate information over time and maintain persistent internal representations of cues to guide decision-making. How the underlying behavioral algorithms of individual animals depend on factors such as experience, developmental stage, or genotype remains poorly understood. Drift-diffusion models provide a powerful theoretical framework to describe and predict performance metrics across a wide range of species. The stochastic nature of these models and the typical limited throughput of most experimental designs challenge the automatic inference of latent variables. Here, we combine high-throughput behavioral assays in larval zebrafish with drift-diffusion modeling, revealing that larvae progressively develop more persistent, self-reinforcing integration dynamics during early development. This effect is reduced in fish carrying mutations in genes linked to human epilepsy and schizophrenia. Our results show that behavior-based drift-diffusion modeling can offer a scalable, automated approach to generate experimentally testable hypotheses about the algorithmic implementation of sensorimotor integration in health and disease.