Over the past 20 years, politics in United States has become increasingly polarized. Republicans and Democrats are more divided along ideological lines than at any point in the last two decades (see fig 1).
Is this a good thing or a bad thing? Previous studies have shown that the diversity of perspectives in the groups leads to superior team performance on complex tasks. On the other hand, strong political perspectives are associated with conflict, misinformation and a reluctance to engage with people who have different opinions.
To answer this question, Shi and colleagues asked if groups of Wikipedia editors with diverse political ideologies produced better or worst articles. First, they went through the English-language Wikipedia articles and estimated their quality using an independent machine learning method (to keep things simple, let’s assume that this rating reflects the articles’ true quality).
Then they focused on the group of editors that contributed to each article. They estimated the probability that each editor may contribute to liberal or conservative articles and called it the editor’s ideological orientation. The group’s political alignment was defined by the mean/variance of the editors’ ideological orientation. This showed how diverse the team of editors responsible for each article were.
Then they tested their key hypothesis by asking if there was any statistical relationship between the two measures defined above: article quality and editorial team’s diversity. Figure 2 shows that in the Political articles category, the more diverse the teams, the higher the quality. A similar pattern was observed in Science and Social Issues articles. Better articles had more diverse editors.
Every Wiki article has a talk page where editors can discuss the article in it. The content of these conversations can tell us interesting things about how articles were produced. For example, in ideological conflicts people may talk about different concepts and/or use different words. The authors decomposed diversity into lexical and semantic diversity. Semantic diversity captured distinct issues discussed on a talk page, whereas lexical diversity captured the number of ways in which editors discuss those issues. Fig 3 shows that compared to the less diverse groups (green line), more diverse groups (purple line) talked about less subjects (they had more focus), but used a lot of distinctive words.
Cool things about this paper are:
(*) Zahra Arjmandi is a volunteer research assistant at School of Cognitive Sciences, IPM, Tehran Iran