Why Machine Learning Is a Misnomer

This paper problematises the notion of machine learning (ML) and challenges its applicability to contemporary algorithms. Drawing on theoretical insights from learning sciences, praxeological, and social-cognitive perspectives of learning and agency, the paper critically analyses the so-called agentic and learning features attributed to current ML algorithms comparing these against genuine, human agentic learning capabilities. Furthermore, the paper explores the implications of co-opting uniquely human nomenclature to describe mechanistic phenomena. The main contribution of this paper consists in exposing fundamental misalignments between genuine agentic learning processes and the operations of current machine algorithms misnomered as learning. The paper offers a more nuanced understanding of the nature of ML/AI, urging a reassessment of the dominant narratives surrounding these technologies. It will be of interest to interdisciplinary scholars of ML/AI, policymakers, educators, as well as computer scientists and technology developers interested in critical social sciences perspectives on algorithmic systems.