The Effort Heuristic Predicts the Evaluation of Generative Artificial Intelligence Users

Abstract

People often associate effort with excellence, assuming that great effort produces high quality. This association forms the basis of the effort heuristic, which we applied to the novel context of Generative AI (GAI) across professional settings. Specifically, we predicted that people would evaluate a service provider (e.g., doctor) more positively when the service provider fulfils professional duties (e.g., giving diagnosis to patients) without using GAI than with GAI, as the use of GAI is assumed to undermine effort. We tested the prediction experimentally with a large community sample (N = 910), randomly assigning them to either the “With-GAI” or “Without-GAI” condition. Participants evaluated the perceived effort of service providers and indicated their general attitudes towards GAI. Results showed that the use of the effort heuristic in evaluations is robust and generalizable across professional settings. Compared to the Without-GAI condition, service providers in the With-GAI condition were evaluated as less effortful and less trustworthy in completing their duties. These results could not be explained by demographic factors, such as age, income level, educational level, and familiarity with GAI. Evaluations of service providers did not predict general attitudes towards GAI, suggesting that evaluations about a specific person and general attitudes involve distinct mental processes. Together, our findings contribute to the social cognition of GAI, a technology which poses unprecedented and significant challenges to the world.