Murray S. Bennett

Postdoctoral Scholar

The Ohio State University

I develop mathematical tools to understand and improve how humans and AI systems make decisions together. My work uses formal cognitive models, Bayesian inference, and machine learning to study the mechanisms of perception, workload, and metacognition — and to put that understanding to work in applied settings like medical diagnosis, team performance, and human-AI collaboration.

My current work equips neural networks with cognitively-grounded uncertainty signals — drawing on how humans use familiarity to judge their own confidence — and uses real-time cognitive modelling to let AI systems predict and adapt to human behaviour as it unfolds. The goal throughout is AI that is not just accurate, but interpretable and genuinely useful to the people working alongside it.

Explore my research projects, try a hands-on experiment demo, or hunt for hidden bonuses scattered across the page.

Email | ORCID | GitHub

Research Highlights

Human-AI Teams

How does AI uncertainty information shape human trust and decision quality? What cognitive costs arise when humans coordinate with AI agents — and how can we reduce them? My work addresses these questions with controlled experiments and quantitative models.

Cognitive Modeling

Using Systems Factorial Technology, General Recognition Theory, and simulation-based inference to extract the latent mechanisms behind human behaviour — in real time, from continuous data, and in settings where standard methods fall short.

Perception & Vision

Measuring what humans actually perceive — not just what they report — using pairwise judgement paradigms and bias-free psychophysical models. Applied to melanoma diagnosis, where human and AI perception are complementary rather than redundant.