From Active Inference to Causal Inference?

In this talk, I will first review core tenets of active inference, a modelling approach developed by theoretical neuroscientist Karl Friston and colleagues. I will then critically examine the extent to which an active inference agent engages in causal inference. To do this, I will first outline how causal inference, as per Judea Pearl’s framework, differs from mere probabilistic inference. Finally, I will argue that, contrary to initial appearances, active inference agents may perform interventional, but not counterfactual, causal inference.