All human actions are preceded by a chain of neural events. In principle, our actions can be predicted from neural activity. As a result, many assume that free will is an illusion of the conscious self. The argument goes something like this: "I might think I’m freely choosing my actions, but in reality, my neurons made me do it.”

But this view rests on a reductionist understanding of causation and suffers from many defects. It does not account for causal structure, and does not distinguish prediction (“what happened”) from causation (“who caused what”).

Instead of focusing on what predicts what, Larissa Albantakis’s project will focus on what causes what. It aims to develop a formal framework to distinguish actions that are freely chosen from those that are not.

The project uses simulated agents equipped with artificial neural networks to test this framework. Through these networks, the project will assess the causal requisites for a physical substrate of consciousness and the actual causes of the agents’ actions. If successful, the project will show that causation—like consciousness—cannot be reduced to a set of smaller elements, opening the way for non-reductionist accounts of free will.