Can simple social copying heuristics explain tag popularity in a collaborative tagging system?

While research on collaborative tagging systems has largely been the purview of computer scientists, the behavior of these systems is driven by the psychology of their users. Here we explore how simple models of boundedly rational human decision making may partly account for the high-level properties of a collaborative tagging environment, in particular with respect to the distribution of tags used across the folksonomy. We discuss several plausible heuristics people might employ to decide on tags to use for a given item, and then describe methods for testing evidence of such strategies in real collaborative tagging data. Using a large dataset of annotations collected from users of the social music website Last.fm with a novel crawling methodology (approximately one millions total users), we extract the parameters for our decision-making models from the data. We then describe a set of simple multi-agent simulations that test our heuristic models, and compare their results to the extracted parameters from the tagging dataset. Results indicate that simple social copying mechanisms can generate surprisingly good fits to the empirical data, with implications for the design and study of tagging systems.