Our lab uses computational models to explain human learning, planning and how they are altered in psychopathology. Located in Bar-Ilan’s brain center, we’re investigating how trait anxiety biases planning content, meta-control, and efficiency. To do so we use online tasks, advanced virtual reality, eye-tracking, and MEG.

A world model inside the brain Inside an outline of a brain: the words world model at the top; beneath them three parts — a person looking in a mirror for causal knowledge about the self, the Earth for causal knowledge about the world, and activities with the points they are worth, good or bad, for intrinsic goals — which feed down into three task representations drawn as small state-transition diagrams. Beside each arrow into a task, a Shepard-style generalization gradient — inheritance decaying with dissimilarity — marks how much of the priors that task inherits, from most on the left to least on the right. Each part can be clicked for an explanation. world model🌍🏃+9🎰−6🤝+8++−−−−

Using world models to explain human variability and psychopathology

Our recent work models meta-control over when and how to plan, and how worriers remove actions correlated with but not causing threat from task representations used to plan. These studies are motivated by our theoretical account of worry as arising from biased threat-avoidant planning. Ongoing work treats anxiety as a property of a person’s world model, which comprises domain-specific beliefs about threat, control and the self. We are in the process of designing personalized tasks to measure how those idiosyncratic beliefs shape task representations used to learn and plan in biased ways.