Mapping AI and data ethics

Part of the JUST AI programme includes conducting a multimethod mapping of the AI and data ethics field.

The challenge of mapping AI and data ethics

A major challenge in mapping the AI and data ethics field concerns what constitutes ‘AI and data ethics’ in the first place. This is especially challenging given their rapid development and expansion to a multitude of disciplines and application contexts.

If we set our criteria too narrowly, we risk excluding important parts of the discussion, while an overly broad definition might run the risk of diluting the contents of our search.

The issue is further complicated by the fact that much of the discussion about matters of ethical relevance do not necessarily adopt the label ‘ethics’, which may be especially true for non-dominant perspectives that are particularly important to include in our mapping.

Our approach

We are therefore taking a more gradual approach to defining the query that informs our bibliometric search. This work started with a small-scale survey of the UK’s AI and data ethics community to capture a snapshot of interests. We are complementing this with a set of topics derived from our literature review. And now, we are inviting you, the community of researchers and practitioners in AI and data ethics, to validate and expand our final choice of keywords to yield a comprehensive search.

Early visualisations

Early visualisations below showcase a small part of the bibliometric mapping work. The graphs depict a network of articles retrieved from the Web of Science academic database using keyword searches for ‘AI ethics’, ‘data ethics’, ‘data justice’ and ‘responsible AI’.

The map illustrates that publications in these categories exist in largely distinct clusters and that discussions around these topics unfold in different academic outlets, with a handful of journals such as Information Communication and Society, or Science and Engineering Ethics acting as bridges.

Map of author-paper by search tag

Map of paper-journal by search tag

These observations may change as we further expand the included set of keywords and data sources but such phenomenon of disciplinary siloing has been noted by previous bibliometric studies investigating computer ethics and highlights the need for supporting multi-disciplinary work and collaborations in the field, which is at the heart of JUST AI’s mission.