Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market
14,341 subjects were presented with unknown songs and the ability to rate on a 5 star scale and to download. Some subjects were not presented with popularity info in the form of download counts, others were, divided into 8 different universes, each with separate download count. Download count availability increased inequality of popularity and unpredictability of popularity (across 8 different universes). 2 experiments were run, one in which popularity information presented more strongly (songs listed in single column in descending download count rank) resulted in much greater inequality.
Theoretical and practical relevance:
Authors note that social influence in the real world is much stronger than in the experiment, eg via marketing.