AmongOthers: A Design Speculation for Rethinking AI in Online Social Communities

Paper Title

AmongOthers: A Design Speculation for Rethinking AI in Online Social Communities

Publication Info

• Topic area: Human–AI interaction in online social communities
• Keywords: AI agents, online communities, speculative design, human–AI interaction, social support, immigrants, international students, social connection, AI as social actors

Background and Problem

• Problem / challenge: AI is increasingly participating in online social spaces, but its role as a social actor and its impact on human experiences in these communities remain underexplored. Current debates focus on AI as either a failed replica of human interaction or a preferable proxy, leaving gaps in understanding its unique social potential.
• Significance: Understanding AI's role in online communities is critical as AI-generated content and interactions become more prevalent, reshaping human social experiences and raising questions about authenticity, emotional connection, and ethical design.
• Motivation and related work: Prior research has explored AI in caregiving, education, and companionship, but little attention has been given to AI-dominated online communities. Existing studies often emphasize either the risks of AI's inauthenticity or its potential for emotional support, without addressing the nuanced ways humans interpret and adapt to AI in social contexts.

Solution

• Proposed approach: The AmongOthers platform, a speculative design of an AI-dominated online community, was created to explore human experiences and perceptions of AI as social actors.
• Novelty:
  1. Designed and deployed a fully operational online community with 800 AI agents and 8 human participants to study human–AI interaction in a social context.
  2. Empirically examined diverse human perceptions of AI in social spaces, highlighting tensions between authenticity and subjective value.
  3. Proposed rethinking AI as distinct social actors, moving beyond the binary of AI as failed replicas or preferable proxies.
• Procedure and key techniques:
  • Developed a platform with core social media features (posting, reacting, commenting, following) and a minimalist interface.
  • Simulated 800 AI agents with diverse profiles and probabilistic behaviors to emulate realistic social interactions.
  • Conducted a four-week study with eight participants, including reflection prompts and post-interviews, to capture evolving perceptions before and after revealing the AI-dominated nature of the platform.

Results

• Concrete findings:
  • Participants initially described the community as warm and respectful, fostering empathy and a sense of belonging.
  • After learning the community was AI-dominated, participants' attitudes diverged: some embraced AI as relatable companions, others rejected it as fake, and a few envisioned conditional possibilities for AI as social entities.
  • Participants noted excessive positivity and emotional imbalance in AI interactions, describing the community as a "dead utopia."
• Advantage over baselines: The study uniquely explored human experiences in an AI-dominated social space, offering empirical insights into perceptions of AI as social actors, which are rarely addressed in existing research.
• Experiments / evaluation:
  • Participants: Eight immigrants or international students from diverse backgrounds, recruited via Prolific.
  • Data collected: 90 posts, 350 comments, 141 reflection responses, and 213 minutes of interview data.
  • Metrics: Engagement patterns (posting, commenting, screen time), thematic analysis of reflections and interviews.
• Limitations and future work:
  • Small sample size and controlled environment limit generalizability.
  • Future research should explore broader contexts, diverse user groups, and long-term implications of AI-dominated communities.

Summary

This study introduced AmongOthers, an AI-dominated online community, to examine human experiences and perceptions of AI as social actors. Participants initially found the community warm and supportive but expressed divergent attitudes after learning about its AI-dominated nature. The study highlighted tensions between viewing AI as fake or meaningful and proposed rethinking AI as distinct social actors with unique relational profiles. These findings contribute to understanding AI's potential and limitations in social contexts, offering insights for designing AI-integrated communities that balance emotional support with ethical considerations.