AI & Ethics

Can We Save Affective Computing? - Jonathan Herington, Assistant Professor, University of Rochester

Please register for this event here.

Abstract: 

Recently there has been a debate over the permissibility of research on automated emotion recognition (AER). On one side is the abolitionist view that this is bad science with bad consequences: that AER is a wholly illegitimate avenue of research akin to “algorithmic phrenology”. On the other, is the permissive view that it is early science with unknown consequences: that AER, like most exploratory scientific projects, is an immature field which may (or may not) turn out to have scientific value and that preventing inquiry into AER inappropriately conflates the responsibilities of scientists with those of policymakers.

In this talk, I argue that the permissive view requires us to accept a doubly-idealized vision of the epistemic and social aims of science. First, the conduct of science is idealized as: (i) untainted by non-epistemic values, (ii) epistemically co-operative, and (iii) ahistorical. By assuming science is innately self-correcting and “value-free”, the permissive view overlooks that research into AER may inadvertently undermine the epistemic aims of science.  Second, the context of science is idealized as basically just, without: (i) differences in basic capacities and needs amongst citizens, (ii) historical and ongoing social hierarchies, and (iii) bad actors who may misuse the products of science. While AER research may promote justice in the context of a basically just society, in our resolutely unjust reality, it is likely to undermine the social aims of science. I end by exploring ways in which a thicker description of these idealizations might help us identify ways to reform – rather than abolish – AER research.

Bio: 

Jon Herington is an Assistant Professor of Philosophy, and Faculty Affiliate of the Goergen Institute for Data Science, at the University of Rochester. His research focuses on the political philosophy of science, health and technology – and he has published work on the governance of dual-use research, fairness in machine-learning algorithms, resource allocation during health emergencies, the harms of climate change, and the ethics of GMO labelling.

Please register for this event here.