Feedback Design in Content Platforms
42 Pages Posted: 26 Aug 2024
Date Written: August 01, 2024
Abstract
Feedback systems are widely used in content platforms. This paper investigates how feedback designs affect content contribution from heterogeneous and competitive contributors, using Stack Overflow as an empirical context. We first show that contributors react to positive (upvotes), negative (downvotes), and exclusive (acceptance) feedback, leveraging variations in answer timing and competition in high-frequency data. We then develop a novel empirical model that characterizes the relationship between feedback design and equilibrium content contribution. Through counterfactual simulations, we assess the efficacy of the current feedback design and explore alternatives. We find that (1) a lenient upvote policy should pair with a harsh downvote policy to maintain content quality; (2) content quality is non-monotonic in downvote harshness, and Stack Overflow is near the optimal level of downvotes; (3) a better outside option can reduce both the quantity and quality of content on the platform.
Keywords: Feedback system, Positive feedback, Negative feedback, Platform design, Knowledgesharing platforms, Competition, Oblivious equilibrium, UGC
Suggested Citation: Suggested Citation