Geodemographic biases in crowdsourced knowledge websites: Do neighbours fill in the blanks?
Geodemographic biases in crowdsourced knowledge websites: Do neighbours fill in the blanks?
复制标题
DOI：
10.1007/s10708-017-9778-7
复制
发表时间：
2018-06-01
期刊：
影响因子：
2.7
通讯作者：
Lee, Sumin
中科院分区：
文献类型：
--
作者：
Bright, Jonathan;De Sabbata, Stefano;Lee, Sumin
维基百科和 OpenStreetMap 等众包知识网站越来越吸引批评文献，这些文献强调了这样一个事实：这些网站的贡献者基础往往存在地理人口统计学偏见：来自更富裕、受过更好教育的人群。然而，虽然贡献者的偏见是众所周知的，但我们对这是否也会导致这些网站上的结果产生偏见知之甚少：或者确实做出贡献的部分人口是否也通过添加有关其他尚未吸引贡献者基础的不太富裕的邻近地区的知识来“填补空白”。本文探讨了这种“邻里效应”在实践中是否存在的问题。它利用英国酒精许可证数据的新颖数据集来评估志愿者地理信息网站 OpenStreetMap 完整性的变化。结果支持现有文献表明完整性与人口统计相关：财富和教育水平较高的地区通常表现出较高的完整性水平。然后，本文通过展示邻里效应存在的证据做出了新颖的贡献：拥有较富裕邻居的较贫困地区通常比周围也有较贫困邻居的较贫困地区具有更高的完整性水平。结果表明，即使用户群仍然存在片面和偏见，众包知识网站也可以追求某种完整性。
Crowdsourced knowledge websites such as Wikipedia and OpenStreetMap are increasingly attracting a critical literature which has highlighted the fact that the contributor bases of these sites are often geodemographically biased: drawn from more affluent and better educated segments of the population. However, while bias in contributors is well known, we know less about whether this also results in a bias in outcomes on these websites: or whether the partial portion of the population which does make contributions also works to "fill in the blanks", by adding knowledge about other less well-off neighbouring areas which have not attracted a contributor base. This article addresses the question of whether such "neighbourhood effects" exist in practice. It makes use of a novel dataset of alcohol license data in the UK to assess variation in the completeness of the volunteer geographic information site OpenStreetMap. The results support existing literature in showing that completeness is related to demographics: areas with higher levels of wealth and education typically exhibit higher levels of completeness. The article then makes a novel contribution by showing evidence of the existence of neighbourhood effects: poorer regions with more affluent neighbours typically having higher levels of completeness than poorer regions which are also surrounded by poorer neighbours. The results suggest that crowdsourced knowledge websites can aspire to a kind of completeness even whilst user bases remain partial and biased.