"Helpless" infants are active, goal-directed agents: Reply to Cusack et al.

Why are humans born “helpless”? Cusack et al. propose an explanation for infants’ limited behavioral repertoires compared to other animals. The authors propose that infants’ helplessness has learning benefits analogous to training foundation models in machine learning. We share Cusack et al.’s enthusiasm for using computational tools to theorize about cognitive development. However, we predict that computational approaches that do not reckon with the characteristic features of infant development reviewed here – that infant learning is active from the beginning, shaped by interdependencies between motor and perceptual learning, subject to experientially- and culturally-shaped variability, and attuned to the social environment – will fall short as models of infant learning. Computational theories of infant learning must be firmly rooted in insights from developmental science, understanding infants as adapted to their own social ecological niche.