Generative Artificial Intelligence (GenAI)-Based Recommender Addressing Contribution Pollution and Information Cacophony on Digital Platforms
针对数字平台上用户贡献过多导致的贡献污染和信息杂音问题，提出一个基于生成式人工智能的推荐系统框架，并用三个平台的真实数据验证其效果，为平台治理提供政策指导。
Digital platforms facilitate interactions among geographically distributed users which reinforce the positive network effects. However, a disproportionate increase in the contributions of users results in contribution pollution (disproportionate increase in users’ contributions) and information cacophony (incoherence of contributions). Unaddressed, contribution pollution and information cacophony challenges have negative implications for platform efficiency and users’ humanistic outcomes. This paper highlights the limitations of existing mechanisms and proposes a framework for a generative artificial intelligence (GenAI)–based recommender system. To evaluate the efficacy of the GenAI–based recommender system, we develop an agent-based simulation model and conduct experiments using real data from three platforms (Reddit, Hacker News, and Stack Overflow). This research contributes a novel sociotechnical framework that addresses the challenges of contribution pollution and information cacophony. For the governance of platforms, this research provides actionable guidance on the policy implications of integrating GenAI-based technologies for user- and platform-level outcomes.