From Competition to Collaboration: Designing Sustainable Mechanisms Between LLMs and Online Forums

Niv Fono, Yftah Ziser, Omer Ben-Porat

Keywords: Generative AI, Strategic Behavior, Collaboration

Abstract: As users increasingly turn to large language models (LLMs) instead of traditional Question-and-Answer (Q\&A) forums, these forums face a steep decline in engagement. This creates a paradox: while Generative AI (GenAI) systems draw users away from forums, they also depend on the very data those forums produce to improve their performance. We propose a strategic game modeling the interaction between Generative AI systems and Q\&A forums, in which GenAI agents propose questions and forums decide which to publish, capturing the interdependent yet partially conflicting objectives of both parties. To ensure real-world applicability, we design the game with principled collaboration choices, including non-monetary exchanges, asymmetric information, and incentive-aware mechanisms that respect forum autonomy and community norms. We bring the game to life through comprehensive, data-driven simulations using real Stack Exchange data and commonly used LLMs from the LLaMA and Pythia families to represent GenAI, demonstrating that the interaction exhibits genuine strategic tension: GenAI learning and forum engagement are far from being aligned. Crucially, even under asymmetric information, our mechanism achieves 56--66\% of forum utility and 46--52\% of GenAI utility relative to the ideal full-information scenario. These results highlight the potential for sustainable, strategic collaboration that preserves effective knowledge sharing between AI systems and human knowledge platforms.

Area: Generative and Agentic AI (GAAI)

Generative A I: I acknowledge that I have read and will follow this policy.

Submission Number: 519