Algorithmic Bias Sandpit
This event ran in September 2022, and subsequently two research awards were granted to cross-institutional, interdisciplinary teams.
The Research Challenge
The theme of the Sandpit is ‘Algorithmic Bias’. It is widely recognised that the use of data-driven models for classification and decision making is vulnerable both to the existence of historic prejudicial decision-making reflected in the data, and to the omission of significant communities from the data leading to blind spots in the resulting models. This has the effect both of degrading the accuracy of the classifications and decisions and to the continuation, and in some cases magnification, of injustice. In our sandpit we will co-create innovative interdisciplinary projects to explore aspects of algorithmic bias and its mitigation from technical, social, and socio-technical perspectives.
Some of the questions considered included:
- How should we define, conceptualise, identify and assess the existence and extent of algorithmic bias in statistical models?
- How can we articulate, evaluate and mitigate the harms to individuals, communities, organisations and nation states that arise from the use of statistical models that exhibit algorithmic bias?
- What role can technology, people, organisations, governments and societies play in minimising the existence of algorithmic bias in the statistical models of the future?
Sandpits are highly interactive brainstorming and networking sessions. They involve participants from different disciplines and backgrounds working together to explore and develop innovative approaches to a challenge. It is our aim that all participants will have had a chance to deepen their knowledge of the area, widen their perspectives, make new contacts and come up with new ideas that will stimulate their research and scholarship.