Analogies Between Human Learning (HL) and Machine Learning (ML)

Determine whether concrete analogies exist between the human-learning scheme described in Section III—where graduate students constitute the “training set,” novice students undergo “supervised learning,” experienced students engage in “unsupervised learning,” and “backpropagation” occurs when supervisors learn from junior researchers—and standard machine learning methodologies; if such analogies exist, identify and describe the precise correspondences between these HL components and their ML counterparts.

Background

The paper proposes a tongue-in-cheek but conceptually framed alternative to traditional machine learning, termed human learning (HL), in which research groups use human interactions to understand data. Within this framework, the authors describe roles and processes using ML-inspired terminology: students as a “training set,” supervised and unsupervised learning stages based on experience, and “backpropagation” when supervisors learn from students.

The authors explicitly acknowledge uncertainty regarding whether there are genuine analogies between HL processes and machine learning mechanisms, indicating an unresolved question about mapping HL constructs to ML concepts.