Computational Modelling in Psychology

Dr. Ven Popov

The group’s work is driven by a simple but demanding question: What would it take for psychology to have theories that are as precise, predictive, and cumulative as those in the physical sciences? Much of contemporary psychology relies on verbal frameworks and loosely constrained models that can be fit to almost any dataset but rarely make risky, quantitative predictions. We approach cognition—especially human memory—as a testbed for something more ambitious: building mechanistic, mathematical theories that specify how latent cognitive processes generate observable behavior, and that can be falsified, compared, and improved in a principled way. This perspective treats computational models not as curve-fitting tools, but as formal embodiments of theoretical commitments about representation, learning, and limited cognitive resources.

To make such theories empirically meaningful, we work at the intersection of modeling, measurement, and data. A central focus is on developing and evaluating measurement models that link noisy behavioral data—such as response times, errors, or confidence judgments—to psychologically interpretable latent variables, using hierarchical Bayesian methods and open, reusable software tools. In parallel, we study how theoretical models can be tested against neural data without over-interpreting what prediction accuracy alone can tell us about underlying representations. Across projects, the goal is the same: to move psychology toward a culture of sharper inference, better-calibrated models, and cumulative knowledge, where theories earn their credibility by making precise predictions about individual people in new tasks and new situations.