Does CASA theory still shape how people treat computers?

Evelien Heyselaar's 2023 Scientific Reports paper reports a direct replication of the original Reeves and Nass CASA (Computers Are Social Actors) politeness-to-computer study, and finds no CASA effect on desktop computers in 2022. The original 1990s study found that participants rated a desktop computer's performance more positively when evaluated by "the same computer" that had tutored and tested them than by a "different" computer, read as evidence that people unconsciously apply human politeness norms to machines. Heyselaar's replication recruited 132 participants (75.7% female, mean age 22.7) from Radboud University, following the original's experimenter scripts and materials as closely as possible — the computer referred to itself as "this computer" rather than "I," and never implied a "self" or "emotions." The result: "Participants did not feel the unconscious, automatic belief that the desktop computer would benefit from them being 'dishonest' by being more polite in reviewing the computer's performance." No CASA effect appeared.

The paper's explanation is novelty, not timeless psychology. Reeves and Nass had argued the opposite — that CASA reflects "belief, not disbelief" that is automatic, independent of "age, knowledge, distraction, or convenience," and holds for "every group." Heyselaar argues the accumulated replication record contradicts this: prior indirect replications of CASA have been mixed, and the mixed results track how "emergent" the tested technology was. Novel technology draws positive, human-like responses that "diminish... after a short period of time," as shown in robot-tutor studies (children initially learned more from a robot tutor than from traditional teaching, with the advantage gone after three weeks) and a three-week chatbot-friendship study (intimacy, believability, and likability ratings fell with each session). On this account, the desktop computer used in the 1990s study was itself emergent technology, so the original CASA finding may have captured a novelty response rather than a stable, evolution-rooted reaction to social cues; three decades later, desktop computers are no longer novel, and the automatic politeness disappears.

This sits in tension with Do humans apply human-human scripts to AI interactions?, which treats apparent departures from classic CASA as evidence of a second, media-specific script system that people apply just as mindlessly as human-human scripts — not evidence that automatic social responding has disappeared outright. Heyselaar's data support a plainer reading for desktop computers specifically: no substitute mindless script fired either; participants were simply not more polite to "the same computer." The finding is more consistent with Do chatbot relationships lose their appeal as novelty wears off? read upstream: that note documents novelty decay within a single longitudinal study, over weeks; this paper argues the same decay operates at the level of an entire technology category, over decades, so a technology can age out of evoking CASA-like responses altogether.

The excerpt is a single replication of one CASA sub-study (politeness), on one technology (desktop computers), with one population (Dutch university participants recruited in 2022); the authors note confounds such as not using the original computer model and modern assumptions about cloud connectivity between "different" computers, and they call for direct replications of other CASA phenomena before generalizing. It does not show that CASA fails for emergent technologies such as conversational AI or robots — the authors' own claim is the opposite, that CASA should still hold for technology that is genuinely novel to its users. The implication is that designers citing CASA to justify anthropomorphic design choices should ask whether their technology is still novel enough to the target users for the mechanism to apply, rather than treating the effect as a fixed property of any device with sufficient social cues.