Search results

Large-scale data analytics have raised a number of ethical concerns. Many of these were introduced in a seminal paper by boyd and Crawford and have been developed since by others (boyd & Crawford, 2012; Lagoze, 2014; Martin, 2015; Mittelstadt, Allo, Taddeo, Wachter, & Floridi, 2016). One such concern which is frequently recognised but under-analysed is the focus on correlation of data rather than on the causative relationship between data and results. Advocates of this approach dismiss the need for an understanding of causation, holding instead that the correlation of data is sufficient to meet our needs. In crude terms, this position holds that we no longer need to know why X+Y=Z. Merely acknowledging that the pattern exists is enough.

In this chapter, the author explores the ethical implications and challenges surrounding a focus on correlation over causation. In particular, the author focusses on questions of legitimacy of data collection, the embedding of persistent bias, and the implications of future predictions. Such concerns are vital for understanding the ethical implications of, for example, the collection and use of ‘big data’ or the covert access to ‘secondary’ information ostensibly ‘publicly available’. The author’s conclusion is that by failing to consider causation, the short-term benefits of speed and cost may be countered by ethically problematic scenarios in both the short and long term.