Is There (Still) Anybody Out There? Online Forums in the Age of Generative AI, joint with Ralf Elsas-Nicolle
Generative artificial intelligence is reshaping how individuals search for information and solve
problems. For decades, online forums, particularly in technical and programming communities,
have served as core infrastructure for decentralized knowledge production. We study how the
diffusion of generative AI affects participation, question complexity, and solution quality in
such communities, and whether governance structures moderate AI-driven substitution. Using
a decade-long panel of posts and traffic data, we compare Statalist, a real-name forum, with
pseudonymous platforms (Stack Overflow and Stack Exchange’s Cross Validated) around the
introduction of ChatGPT in November 2022. We combine activity measures with large-scale
textual analyses, including an LLM-based complexity taxonomy and a benchmark comparison
between community answers and a standardized “AI expert” proposal. We find that posting
activity declines across platforms, but less sharply in the non-anonymous setting. At the same
time, the share of high-complexity questions increases more strongly on Statalist, suggesting that
AI primarily displaces routine problem-solving while communities retain a comparative advantage
in complex, context-rich discussions. However, more recent data indicate that this resilience may
be transitional rather than permanent. Overall, our results suggest that governance shapes how
the division of labor between humans and AI evolves in digital knowledge communities.