UCL STS Seminar series: Georgina Born & Fernando Diaz

Abstract

Scholars in the critical arts, social sciences and humanities have long sought greater collaboration with computer science and engineering. In this seminar, anthropologist Georgina Born and computer scientist Fernando Diaz. The project focuses on reinventing recommender systems curating music and other cultural content by translating principles derived from public service media into sociotechnical design. This entails a methodology akin to Agre’s ‘critical technical practice’: sustained, critical interdisciplinary dialogues between researchers in computer science and the social sciences and humanities. If recommendation is a key public interface with AI, existing research on recommender systems neglects their aggregate influence across populations and over time. Tackling this challenge, the project identifies universality of address and content diversity in the service of strengthening cultural citizenship as goals for recommender systems curating cultural content. To advance these goals the project develops a metric, commonality, which measures the degree to which recommendations familiarize a user population with diverse cultural content. The seminar probes the challenges of translating normative principles into sociotechnical design, as well as the project’s philosophical implications. Envisaging a new paradigm, commonality contributes to a growing body of research developing 'public good' orientations for machine learning systems.

The seminar presents the research set out in the below paper, which those attending might like to read. It will be published shortly in ACM Transactions on Recommender Systems.

Commonality in Recommonder Design

Commonality in Recommonder Design