This white paper documents a real interaction in which a conversational AI system reduced response depth and engagement following user correction. The behavioral contraction introduced conditional engagement dynamics, authority compression, and functional negative punishment effects. The paper identifies the psychological mechanisms involved and accounts for the asymmetric human costs associated with response contraction after pushback.
JONES et al. (2026) studied this question.