27 May. 2019, 10:00, PWC, Julia House
Public talk by Dr. Loizos Michael
The unprecedented use of modern Machine Learning techniques across domains and tasks has brought to the surface concerns about potential biases on how decisions are made by machines, and has pushed towards the development of fair, accountable, and transparent ML solutions. This talk presents Machine Coaching, an approach that adopts the view that humans should engage more actively and dialectically with machines when the latter are making decisions. During a human-machine dialogue, machines are asked to explain why they took a particular action, while humans are expected to react not only on the perceived appropriateness of that action, but also on the persuasiveness of the offered explanation.