Nicholas Gray, University of Liverpool, UK
Supervisor: Scott Ferson, Marco de Angelis
Keywords: Humane Algorithms, AI Ethics, Imprecise Probabilities, Uncertainty Analysis, Risk Science
Increasingly inhumane algorithms moderate more and more of daily life, with potential risks ranging from irritation and confusions to injustice and catastrophe. One solution is for algorithms to better understand the risk and uncertainties present in the situations where they are used. Such an approach has many potential benefits: uncertainty reduces untenable assumptions and illogical results; produces more interrogatable outputs; and may lead to better decision making. My thesis explores how probability bounds analysis can enable algorithms to appreciate uncertainty within their inputs, internal calculations, and their outputs.
More information on Nicholas work: