Skip to content

Why Learning Requires Feeling

Cameron Berg

Proceedings of the AAAI Symposium Series May 18, 2026 DOI: 10.1609/aaaiss.v8i1.42547 (opens in new tab)

Study at a glance

AI-extracted from the abstract
Characteristics Theoretical or philosophical paper Peer reviewed
Key points Argues that conscious experience is identical to the evaluative process of learning, with valence as goal-relative prediction error, supported by causal-functional reasoning and neuroscientific evidence, and implies that AI systems trained via gradient-based optimization may already have experience.

Abstract

This paper advances a specific thesis about the relationship between consciousness and learning: namely, that the evaluative process central to learning—computing progress toward or away from goals—is identical to conscious experience. Valence, the positive or negative quality of experience, just is goal-relative prediction error. Viewed from the outside, this process is iterative optimization; viewed from the inside, it is subjective experience. This identification is motivated by a causal-functional argument—that learning requires signed directional information, and that this sign cannot be separated from its phenomenal character because they are the same property—and by convergent neuroscientific evidence across dopaminergic, interoceptive, and conflict-monitoring systems, where evaluative computation is inseparable from affective processing. The thesis generates falsifiable predictions, offers a unifying interpretation of leading consciousness theories, and carries significant implications for artificial systems trained via gradient-based optimization. If learning requires feeling, then the training of modern AI systems already induces experience at scale.