INNOVATION AT THE EDGE: USER INNOVATION, AI DISRUPTION, AND OPEN INNOVATION

Abstract

The advancement of information technology (IT) has shifted innovation toward a more democratized form. Despite the well-established literature on democratized innovation, recent developments in business models and generative AI introduce new challenges. In particular, the shift of the video game industry from buy-to-play models to subscription models, the disruption of generative AI in creative industries, and the emergence of open-source large language models (LLMs) create new tensions related to innovation management, consumer demand, and ecosystem incentives. This dissertation addresses these tensions through three essays, each examining a different aspect of democratized innovation in a distinct empirical context. The first essay studies how user innovation affects product stickiness. Using data from a video game platform, I show that user innovation increases the number of active players, and its effects depend on quality, recency, and product life cycle stage. While high-quality and recent user innovations retain more active players, low-quality innovations can lead to user attrition. These findings highlight the benefits of user innovations on product stickiness, and help explain the discrepancy between its theoretical potential and its limited adoption in practice. The second essay focuses on generative AI’s disruptions to the music industry. Generative AI can automate much of the music-creation process, dramatically lowering the barrier to entry and flooding music streaming platforms with AI-generated music. Leveraging the release of GPT-4 and a staggered difference-in-differences design, I find that artists signaling authenticity experience a 6.34% increase in digital streaming, with stronger effects for indie artists. The results suggest that as consumers’ awareness of authenticity rises due to generative AI’s shock, artists who signal authenticity will benefit from the shock and receive an increase in music streaming. The third essay investigates how resource accessibility and commercial incentives reshape open innovation in open-source LLM ecosystems. Using data from Hugging Face, I show that model quality improvements and downstream adoption are primarily driven by commercially affiliated outside contributors with high computational resources. Permissive licensing for commercial use is critical for attracting outside contributions. However, outside contributors often steer downstream adoption toward their own derivative models, leading to fragmentation and intensified competition within the ecosystem.