Speaker: Gualtiero Piccinini

According to simple mapping accounts, a physical system implements a computation just in case there is a mapping from physical states to computational states. According to causal mapping accounts, the mapping must mirror the causal structure of the physical system. Mapping accounts are more liberal than mechanistic accounts, and this allows mapping accounts to apply in contexts where mechanistic accounts may be too restrictive. But existing mapping accounts are inadequate because the notion of mirroring at their core is insufficient. As a consequence, many have argued that even sophisticated mapping accounts, such as causal mapping accounts, cannot avoid trivializing the notion of implementation, entailing unlimited pancomputationalism—the thesis that any physical system performs many nonequivalent computations. We argue that mapping accounts can be made adequate by incorporating appropriate physical constraints. Specifically, according to the robust mapping account we propose, a mapping from physical to computational states is a legitimate basis for implementation only if it includes only physical states relevant to the computation, the physical states have enough spatiotemporal structure to map onto the structure of the computational states, and the evolving physical states bear neither more nor less information about the evolving computation than do the computational states they map onto. When these conditions are in place, a physical system can be said to implement a computation in a robust sense, which does not trivialize the notion of implementation. This robust mapping account can be fruitfully applied to important questions in the physical foundations of computation and cognitive science. 

Please note that these seminars are open to the public and in person only.