DeepSeek Role Enactment
Description
This data package accompanies the manuscript When AI Talks Like a Friend: Building Automated Social Presence Through Relational Role Enactment. It contains data from three studies investigating how users perceive AI role enactment and how these perceptions influence automated social presence, trust, and continued use intention. Study 1 (N = 95) developed and validated a new measure of perceived instrumental and relational role enactment. Participants identified a large language model they had recently used and completed an eight-item scale assessing whether they perceived the AI primarily as a functional tool or as a relational partner. Exploratory factor analysis supported a two-factor structure corresponding to instrumental and relational role enactment. Studies 2 (N = 143) and 3 (N = 145) were randomized online experiments conducted in financial planning and medical report interpretation contexts, respectively. Participants instructed DeepSeek to communicate either as a professional tool or as a trusted friend, after which they completed measures of perceived role enactment, automated social presence, cognitive trust, affective trust, and continued use intention. The datasets include participant-level responses, experimental condition assignments, and all survey measures used in the manuscript. They can be used to reproduce the reported analyses, including scale development, reliability and validity assessment, experimental manipulation checks, confirmatory factor analysis, measurement invariance testing, structural equation modeling, and mediation analyses, or for secondary research on human–AI communication, automated social presence, trust, and generative AI.