Should generative AI be responsible to users? Developing and validating the user-generative AI psychological contract scale

This is the authors' abstract. We don't add key points for this paper.

As GenAI becomes increasingly embedded in AI-enabled service interactions, users' evaluations of AI are extending from functional performance to expectations regarding responsibility and obligation. Drawing on Computers Are Social Actors (CASA) theory and psychological contract theory, this study defines the user–generative AI psychological contract (AIPC) as users' beliefs about the responsibilities and obligations GenAI should fulfill toward them during ongoing interaction and develops a corresponding scale. Study 1 used purposively recruited interviews (N = 26) to identify the construct content domain. Study 2 used two nonprobability online-panel samples of active GenAI users (N = 507 and N = 506) to develop the items and examine the factor structure, reliability, and construct validity. Study 3 used the same 506-case sample employed for structural validation in Study 2 to provide an initial test of the nomological network among AIPC, AI anthropomorphism, and co-creation intention. Results indicate that AIPC consists of three dimensions—the functional support contract, relational collaboration contract, and ethical safety contract—and forms a 16-item scale. Users reported the strongest expectations regarding ethical safety. Anthropomorphism was indirectly associated with co-creation intention through functional support and relational collaboration, but not through ethical safety. The findings extend psychological contract theory to user–AI relationships and provide a measurement tool for studying responsibility expectations toward GenAI.