Simon CatonAssistant Professor of Computer Science at University College Dublin (UCD) Are we being fair about fairness in machine learning?There has been a surge in approaches seeking to make machine learning models "fair(er)". The majority of these approaches can be viewed as interventions applied before, during, and/or after the use of the machine learning method(s). A lot of this work is solid research, with very tangible impacts and benefits. However, are we being fair in our exposition of fairness in machine learning, or AI in general? This talk will give a brief high-level overview of what fairness in machine learning currently looks like, but seeks to explore some of the gaps in the literature where there is a need for more focussed effort. The idea here is not to criticise, but rather advocate that to be fair about fairness research we need to go beyond existing mathematical definitions and approaches of what it means to be fair. Specifically, moving towards a more holistic view of seeing fairness embedded into a complex system of different actors, goals, and (un-)achievable trade-offs. The talk aims to give a comprehensive synopsis of currently unsolved problems and under explored topics in the fairness literature. |
Zaneta Lucka-TomczykWomen in AI (#WAI) Poland Ambassador #breakthebias - towards an inclusive and equitable AI ecosystemOur world is made up of biases. Despite our best efforts, we will still be biased human beings. but AI doesn't have to reflect that. We simply need to act on multiple levels - by raising awareness, educating, promoting diversity (gender, age, religion, etc.), and re-evaluating how the AI ecosystem is designed and used for the benefit of a global society. So let's analyze some data and examples incl. very fundamental issues, and discuss what more we can do to ensure an inclusive and equitable AI ecosystem. |