AI-mediated social support: the prospect of human–AI collaboration

Abstract

The rise of large language models (LLMs) expands opportunities for creating interpersonal messages. Building on artificial intelligent (AI)-mediated communication (AI-MC), this study examines how people use LLM-based chatbots to generate support messages and how patterns of human–AI collaboration shape message features and influence message evaluations of helpfulness and authenticity. We propose the process-adoption model, categorizing message generation into four patterns: human-only, AI-only, modified-AI, and AI-guided. Results showed that AI-only and modified-AI messages were more likely than human-only messages to include informational and emotional support, which in turn, enhanced viewers’ evaluations of message helpfulness and authenticity. AI-guided messages were more likely to provide reciprocal self-disclosure than AI-only messages, which enhanced perceived authenticity. Lastly, AI-guided messages were rated as more authentic than AI-only messages even after accounting for the mediating effects of message features. These findings provide a nuanced understanding of the AI-MC spectrum, and discussions are provided about human–AI collaboration in support provision.