Beyond a Conventional Chatbot: How AI Streamers Transcend Live Streaming Experiences from Viewers’ Perspectives

Paper Title

Beyond a Conventional Chatbot: How AI Streamers Transcend Live Streaming Experiences from Viewers’ Perspectives

Publication Info

• Topic area: Human-Computer Interaction (HCI) focusing on AI systems in public social spaces.
• Keywords: AI streamers, live streaming, generative AI, Neuro-sama, social AI agents, audience engagement, ethical risks, creative content, human-AI interaction, moderation challenges.

Background and Problem

• Problem / challenge: Existing AI systems for social needs are constrained to private, structured interactions (e.g., chatbots) and lack exploration in dynamic, real-time public social spaces like live streaming. AI agents’ ability to independently perform and interact in such environments remains understudied.
• Significance: Understanding AI streamers can reveal how AI systems reshape human-AI interaction, innovate live streaming practices, and address ethical and emotional risks in public online spaces.
• Motivation and related work: Prior research has explored AI agents for emotional and social needs but focused on constrained contexts (e.g., one-on-one conversations or asynchronous interactions). AI streaming introduces AI agents as independent, proactive actors in public spaces, requiring new frameworks to address their unique opportunities and risks.

Solution

• Proposed approach: Qualitative thematic analysis of 1,891 YouTube comments on Neuro-sama, a popular AI streamer, to examine viewer perceptions of AI streaming innovations and risks.
• Novelty:
  1. Empirical investigation of AI agents in public, real-time social spaces.
  2. Identification of how AI streamers innovate live streaming through creative content creation, novel identity practices, and audience interaction mechanisms.
  3. Proposal of design principles to enhance AI streamers’ social and creative affordances while mitigating risks.
• Procedure and key techniques:
  • Data collection using YouTube Data API, filtering comments for relevance and length.
  • Stratified sampling across videos to ensure balanced representation.
  • Thematic analysis following Braun and Clarke’s six-phase guide to identify recurring themes and relationships.

Results

• Concrete findings:
  • AI streamers like Neuro-sama transcend conventional AI applications by creating unique personalities, engaging in boundary-pushing behaviors, and fostering nuanced relationships with human creators.
  • Risks include emotional damage due to lack of accountability, embedded biases from training data, and challenges in real-time moderation.
• Advantage over baselines: AI streamers offer richer, multi-modal interactions and public engagement compared to traditional chatbots and AI influencers, fostering sustained audience interest and community identity.
• Experiments / evaluation: Analysis of 1,891 comments revealed themes related to viewer perceptions of AI streamers’ innovations and risks. Sampling ensured diverse perspectives across videos.
• Limitations and future work:
  • Focus on a single AI streamer and English-based community may limit generalizability.
  • Data from fan-managed channels may reflect engaged viewers rather than casual audiences.
  • Future research should explore global perspectives, mutual dynamics between viewers and AI streamers, and experiences of other stakeholders like developers and moderators.

Summary

This study investigates Neuro-sama, an AI streamer, as a case to explore how AI systems innovate live streaming practices and introduce new risks. Findings reveal that AI streamers transcend conventional AI applications by developing unique personalities, engaging in boundary-pushing behaviors, and fostering nuanced relationships with human creators, while also posing challenges such as emotional damage, biases, and moderation difficulties. Proposed design principles aim to enhance AI streamers’ creative and social affordances while addressing ethical concerns. These insights contribute to broader discussions on designing AI systems for public, dynamic social spaces.