The need of trust in predictive artificial intelligence

How this was made

The poster compares research on predictive monitoring and trustworthy AI using records from the Scopus database. It charts how many articles each topic produced per year, and maps a bipartite network that connects authors to the keywords they chose. Keywords used more often are drawn larger.

The poster was presented at the Fourth International Conference on Quantitative Ethnography (ICQE 2022), held in Copenhagen, Denmark, 15–19 October 2022 (icqe22.org).

This interactive edition, by Jorge Luis Braojos Morillas, was rebuilt from the poster image:

• Yearly article counts were read from the poster's two bar charts.
• Node positions, sizes and types were measured from the printed network: 2,261 nodes, of which 1,046 are keywords and 1,215 are authors.
• The 63 Ethics connections are the ones highlighted on the poster. The 110 Machine learning connections come from the poster's second panel, which shows the same layout rotated by about 10°, aligned back onto the main network.
• The remaining connections (about 6,000) were reconstructed from the layout and the visible lines. Treat them as approximate.
• The poster names only four keywords, so every other node carries a number instead of a name.

An illustration of the poster's argument, not a copy of the dataset. With the original network file, every node could carry its real name and exact connections.