Interpreting and designing for ‘fairness’

Activity: Talk or presentation typesLecture and oral contribution

Peter Alexander Earls Davis - Other

The embedding of ‘fairness’ into the very design of machine learning techniques to reduce the occurrence of unfair outcomes is a significant topic in interdisciplinary research. This body of research analyses the protection of ‘sensitive attributes’ and thus targets the known discriminatory biases that may be reinforced through machine learning techniques. These attributes generally correspond to the protected grounds found in non-discrimination legislation but there has been an emerging body of interdisciplinary literature criticising the restrictiveness of this focus. There are alternative notions of ‘fairness’ across different areas of law however, including consumer law and data protection and privacy law for instance that seem to only partially overlap (if at all) with this focus on equality.
In the recent reform discussions of the Privacy Act (Cth) 1988 for instance, there is a proposal to introduce a ‘fairness and reasonableness’ test, inspired in part by the fairness principle in Article 5(1)(a) of the EU General Data Protection Regulation, that seems to go beyond the focus on bias in the machine learning literature mentioned above. There is a need for a more developed discussion of what ‘fairness’ actually means in law(s), how this relates to the questions of bias in the machine learning literature, and therefore, what this means for policymakers and enforcement authorities. The purpose of this event then is to explore different meanings of fairness in light of interdisciplinary and comparative insights and reflections.
12 Dec 2023

Event (Workshop)

TitleInterpreting and designing for ‘fairness’
Date12/12/2023 → …
LocationAustralian National University

ID: 376267621