This paper proposes a cognitive model of magic based on prediction and prediction error, rather than belief formation and collapse. Spectator cognition is conceptualized as prediction grounded in conviction, and its error processing is categorized into update, collapse, and meta-collapse. Preliminary observations suggest that non-verbal presentation promotes prediction convergence and produces stronger experiences. This model may contribute to a unified understanding of cognitive change beyond magic.
Takuya Yamamoto (2026) studied this question.