Unpacking the Structure of Knowledge Diffusion in Wikipedia: Local Biases, Noble Prizes and the Wisdom of Crowds

Abstract

This paper investigates the diffusion of around 100,000 articles about literary authors in 52 versions of Wikipedia. We studied how Wiki versions replicate articles of authors belonging to a particular linguistic group and we collected findings about the potential mechanisms governing the replication process and its fairness. Results showed that diffusion of articles follows a power law, governed by strong preferences among versions, with a high number of isolated articles only present in one Wikipedia version. We found that the English Wiki has a prominent role in diffusing knowledge. However, results also showed that other Wikipedia versions were fundamental to building a rich global corpus of knowledge. Classical Greek and Latin authors resulted the most replicated set of entries. We found that geographic proximity and linguistic similarity was pivotal to explaining mutual links between Wikis. Finally, despite the presence of preference mechanisms, we found how the relative importance that each Wikipedia versions assigns to the set of authors of each language is significantly correlated with an expert-based ranking built on the outcome of various international literary awards, including the Nobel Prize. Moreover, we showed how Wikipedia exhibits a strong Wisdom of Crowds effect, with the collective opinion of all the Wikipedia versions showing a correlation with the experts higher than any individual Wikipedia version, with a value for Pearson's’ r of about 0.9.