Jaspreet Sahota, The Problem of Historical Bias in Supervised Machine Learning

Ethics of AI in Context: Emerging Scholars

The Problem of Historical Bias in Supervised Machine Learning

Machine learning algorithms are becoming ubiquitous in business and government. Algorithms are routinely deployed that make decisions about  wehow live: e.g. credit adjudication, parole approval, resume screening, insurance costs, etc. Training supervised algorithms on the basis of historical data has the risk of perpetuating historical biases in contemporary society. This can lead to a pernicious feedback cycle that should be avoided by eliminating bias from training data and furthering research into deep learning models.

☛ please register here

Jaspreet SahotaJaspreet Sahota
Independent Researcher
Ph.D. Physics, University of Toronto

Wed, Nov 13, 2019
04:00 PM - 05:30 PM
Centre for Ethics, University of Toronto
200 Larkin