Alessandro Silacci | Searcharxiv

arXiv2026

When Workout Buddies Are Virtual: AI Agents and Human Peers in a Longitudinal Physical Activity Study

Physical inactivity remains a critical global health issue, yet scalable strategies for sustained motivation are scarce. Conversational agents designed as simulated exercising peers (SEPs) represent a promising alternative, but their long-term impact is unclear. We report a six-month randomized controlled trial (N=280) comparing individuals exercising alone, with a human peer, or with a large language model-driven SEP. Results revealed a partnership paradox: human peers evoked stronger social presence, while AI peers provided steadier encouragement and more reliable working alliances. Humans motivated through authentic comparison and accountability, whereas AI peers fostered consistent, low-stakes support. These complementary strengths suggest that AI agents should not mimic human authenticity but augment it with reliability. Our findings advance human-agent interaction research and point to hybrid designs where human presence and AI consistency jointly sustain physical activity.

cs.HC↗

arXiv2025

When motivation can be more than a message: designing agents to boost physical activity

Virtual agents are commonly used in physical activity interventions to support behavior change, often taking the role of coaches that deliver encouragement and feedback. While effective for compliance, this role typically lacks relational depth. This pilot study explores how such agents might be perceived not just as instructors, but as co-participants: entities that appear to exert effort alongside users. Drawing on thematic analysis of semi-structured interviews with 12 participants from a prior physical activity intervention, we examine how users interpret and evaluate agent effort in social comparison contexts. Our findings reveal a recurring tension between perceived performance and authenticity. Participants valued social features when they believed others were genuinely trying. In contrast, ambiguous or implausible activity levels undermined trust and motivation. Many participants expressed skepticism toward virtual agents unless their actions reflected visible effort or were grounded in relatable human benchmarks. Based on these insights, we propose early design directions for fostering co-experienced exertion in agents, including behavioral cues, narrative grounding, and personalized performance. These insights contribute to the design of more engaging, socially resonant agents capable of supporting co-experienced physical activity.

cs.HC↗