CASA Habituation: Social Responses to Computers Diminish with Familiarity

The CASA effect no longer replicates for people who grew up with desktop computers. Heyselaar (2023) attempted to replicate the original Nass & Moon CASA experiments with participants habituated to computer interaction. The automatic social responses that defined CASA — politeness, reciprocity, stereotype application — were significantly attenuated or absent. The implication is that CASA’s social-response effect is tied to technology novelty, not a permanent feature of human cognition.

This raises the strongest counterargument to relational instability theories of AI fatigue: if social responses to computers habituate, perhaps the relational oscillation users experience with AI agents will also fade with exposure. But there is a structural reason it may not. Desktop computers presented minimal, consistent social cues that users could learn to categorize and dismiss. AI agents present rich, inconsistent social cues — deferring like a subordinate, asserting like a peer, explaining like a teacher — that resist stable categorization. Habituation to a fixed stimulus is well-documented; habituation to a stimulus that changes its categorical signals on every interaction is not. See categorical-ambiguity-of-ai.