Predictions, precision, and agentive attention.

Abstract

Ransom, Fazelpour, and Mole (this journal - 2017) raise an important puzzle for the 'prediction error minimization' account of cognitive processing. That account depicts all cognitive processing as fundamentally in the business of minimizing prediction errors concerning the evolving flow of sensory information. One of the cornerstones of these highly ambitious, would-be unifying accounts is their depiction of attention as nothing other than the process of optimizing the precision (inverse variance) of critical prediction error signals. But that story, Ransom et al. suggest, cannot accommodate voluntary shifts of attention. In this paper, I show why this challenge to the grand unifying project fails. It fails because it locates the origins of voluntary attention in complexes of unanalyzed desire rather than in changing complexes of beliefs.