Research

A recurring problem in applied cognitive science is that humans and AI systems each have blind spots the other doesn't — and that those blind spots tend to emerge at precisely the moments that matter most. My research programme develops the mathematical and experimental tools needed to find those blind spots, measure them precisely, and design systems that compensate for them.

The work spans three connected areas. In cognitive modelling and decision science, I apply Systems Factorial Technology, General Recognition Theory, and simulation-based inference to extract latent cognitive mechanisms from behavioural data — including in real time, from continuous tasks, and in competitive group settings. In human-AI collaboration, I study how the design of AI systems — their confidence signals, their representations, their behaviour — shapes the quality of joint decisions with human partners. In perception and applied vision, I use pairwise psychophysical paradigms to produce continuous, bias-free measures of perceptual features that standard methods miss.

These areas are not separate tracks — they feed into each other. The perceptual measures become training signals for classifiers. The classifiers get uncertainty estimates grounded in how humans signal confidence. And the uncertainty estimates get tested in controlled human-AI teaming experiments. Click any project below to read more.

Email | ORCID | GitHub

Human-AI Collaboration

Cognitive Modeling & Decision Science

Perception & Applied Vision