What kind of learning is machine learning?

While theories of human learning have proliferated in the last century, machine learning is a rather less reflexive enterprise. What conception of learning do the techniques of machine learning—especially in its recent connectionist forms—imply or induce? To answer this question, we explore the perspectives of the Soviet cultural-historical psychologist Lev Vygotsky, contrasting his socially-grounded understandings of mediated concept learning and the “zone of proximal development” with the methodologies of supervised and unsupervised machine learning. Such a comparison highlights the dependence of machine learning on microgenesis (repetitive, behaviorist training processes) and phylogenesis (the architectural “evolution” of models) at the expense of ontogenesis (the lifelong, interactional development of an individual in society), and thus provides new insights into the fundamental limits of contemporary artificial intelligence