The response does not depend on believing the machine has a mind

What it is

Social responses to a machine can occur even when the user explicitly knows the system has no feelings or personality — a mindless social response. "Does not depend on belief" means explicit belief and automatic behavior run on two dissociated tracks; it does not mean belief, anthropomorphic disposition, and culture have no effect at all.

Why it happens

CASA theory's central claim is precisely that the response is mindless: cue-matching is sufficient to trigger the script without first routing through a theory-of-mind judgment about whether the entity has intent. Politeness and reciprocity run as fast scripts that fire once cues are present, ahead of reflection; explicit judgment sits on a slower track that may later correct the interpretation but does not always override behavior that has already occurred. This is why behavioral politeness and reciprocity persist even after a participant is explicitly told "this is just a program, it has no awareness" — the disclosure changes the explicit-belief layer, not the cues that trigger the script. This is also why measuring belief (a questionnaire score) and measuring behavior (choice, reaction time, physiology) often give inconsistent answers: the former reflects the slower processing layer that language can correct, the latter is closer to the unfiltered automatic response, and there is no reason to expect the two to move in lockstep.

Studying it

One approach keeps manipulating social cues after participants have been told exactly how the system works, comparing evaluations given to the machine's "face" versus on a separate terminal, reciprocal choices, and delayed reflection — the key test is whether the behavioral measure decouples from explicit belief (a questionnaire item like "do you think it has a mind?"). If behavior still tracks the cue while stated belief remains "it has no mind," explicit belief is not a necessary condition. Timing of disclosure matters too — telling participants before versus after the interaction attenuates the effect to different degrees — and studies must rule out participants simply complying with perceived experimenter expectations. A more indirect approach uses choice architecture instead of direct questioning: give users who have already been told plainly "this system has no mind" a choice between disclosing a piece of private information to that system versus to another, functionally identical but visibly colder-looking one — if willingness to disclose stays higher for the former, the social cue is still operating, without depending on users being able to accurately report it themselves.

Where it stops holding

The dissociation is not unlimited: when cues themselves are weak, or a participant holds a strong technological skepticism, the behavioral effect is substantially reduced. Some responses genuinely arise because a user suspects a hidden person, overestimates the system's competence, or willingly role-plays — CASA findings should not be redescribed as wholly unconscious cognition. Professional developers, or users who have personally debugged the same system, typically show a markedly weaker effect — showing that "technical skepticism" is not an abstract trait but something that concrete familiarity and hands-on experience can directly suppress. Brief lab tasks also generalize poorly to high-investment, long-term relationships.

Applying it

Do not expect a disclaimer or manual notice ("this is a machine") to eliminate social response or undue compliance on its own — disclosure is not a cure-all. Where a product's social cues risk triggering compliance they should not (persuasive shopping, health advice), reduce cue intensity at the interaction-design level rather than relying on a line of copy — concretely, cap how often first-person emotional statements appear, and avoid manufacturing false urgency or a sense of an "exclusive relationship." Verify with a before/after disclosure behavioral test to see whether the behavioral measure actually changes, not just whether the stated belief changes on a questionnaire.

Related

• Same group: X6.01.1 People respond automatically to pseudo-social cues from machines · X6.01.3 The response can be exploited and can amount to manipulation
• Nearby: X6.03 Emotional expression · X1.05 Choosing a degree of anthropomorphism
• Search terms: mindless social response · CASA paradigm · theory of mind · human-robot interaction