How Does Generative AI Affect Demand for User-Generated Content? Evidence From Stack Exchange

Recent advancements in generative artificial intelligence (AI) are starting to make profound changes in many aspects of our lives, particularly with the emergence of large language models (LLMs) (e.g., GPT3, BERT) that are capable of processing and generating both human and programming languages. With the increasing popularity and rapid advancement of generative AI, more and more consumers have started turning to these LLMs rather than other humans when seeking knowledge about products or services. Nevertheless, the extent to which such LLMs may substitute content generated by other humans might differ considerably across topics, questions, and user types. Furthermore, under specific scenarios, LLMs may even stimulate user-generated content (UGC) by offering intermediate responses that humans can build upon subsequently. Within this context, we examine whether and how these LLMs substitute or complement UGC in the programming community using data from StackOverflow, the go-to question-and-answer site for programmers. One unique aspect of our data is that we can characterize the demand for UGC within this community before vs. after the introduction of ChatGPT. Specifically, we study how the launch of ChatGPT influences demand for UGC within different sub-communities (e.g., different programming languages and topics), for different kinds of questions (e.g., complicated vs. straight-forward questions), and among different types of users (e.g., high-status and low-status users within the community). Our research contributes to the rapidly emerging stream of research investigating the impact of generative AI on human intelligence and efforts