About
The AlgoBias Toolkit is a collection of resources created to assist organisations in thinking through issues relating to algorithmic bias.
In the past few years, we’ve seen an influx in discussion, research, and guidelines relating to algorithmic bias and how best to mitigate against the potential harms posed by algorithmic technologies. Evidence from research on public opinion around data practices suggests the UK public are particularly concerned about how fair data-driven technologies are to marginalised groups (Ditchfield et al., 2022). Additionally, researchers numerous agree that algorithmic bias needs to be approached from a socio-technical perspective (that is, considering how social and technical systems work together, and rather than in isolation) (Balayn and Gürses, 2021; O’Neil, 2017; Selbst et al., 2018; Eubanks, 2018). However, for mitigation methods to be most effective, these issues need to be approached in a way which considers the whole organisation and the environment its embedded within (Beresford, forthcoming). This toolkit aims to give organisations the tools to develop tangible steps in making their data-driven, algorithmic and AI technologies fairer, and reduce the risk of algorithmic bias.
Creation of this toolkit
This toolkit was created based on research in collaboration between the University of Sheffield and the Department of Work and Pensions (DWP). The process for this involved distilling the insights on this research, as well as insights from other prominent algorithmic bias mitigation work, and presenting these to participants at DWP as part of a series of educational workshops.
The resources used in these workshops can be found [link here]. For a longer discussion about the process, I used to create the workshop materials, and convert the insights gained in these workshops into the materials in the AlgoBias Toolkit, a publication on the process is forthcoming.