Examining the Impact of EmpathicAI among Vulnerable Communities

Abstract

This research examines how empathetic AI influences emotional support exchanges within trauma recovery communities, specifically the “raised by narcissists” (RBN) subreddit. Drawing on IT-Enabled Public Goods Theory, we investigate how three empathy modalities, cognitive empathy (understanding perspectives), emotional empathy (sharing emotions), and motivational empathy (driving prosocial behaviors), impact connectivity in online communities. We present findings from a controlled survey experiment (N=736) testing AI-generated comments with different empathy modalities on user engagement likelihood. Results demonstrate that participants who read empathetic AI comments showed significantly higher engagement intentions compared to neutral responses, with cognitive empathy (β = 0.49, p < 0.05) and motivational empathy (β = 0.40, p < 0.05) showing stronger effects than emotional empathy (β = 0.34, p < 0.1). Notably, neutral AI responses performed worse than no comment at all, suggesting that emotionally flat interactions may actually discourage engagement in sensitive contexts. Our findings extend IT-enabled public goods theory by demonstrating how different empathy modalities contribute unequally to connectivity outcomes in vulnerable populations. This research advances understanding of AI’s role in trauma recovery communities by examining empathetic AI in emotionally complex, peer-support environments rather than dyadic interactions, informing development of more compassionate and contextually appropriate AI technologies for vulnerable online communities.