Description
Ruster, L. P., Daniell, K. A. (2022). Algorithmic impact assessments (AIAs): A case for moving beyond mitigating harms. Social Responsibility of Algorithms, Paris. Abstract Algorithmic Impact Assessments (AIAs) have recently emerged as a form of accountability for organisations leveraging algorithms in automated systems. Origin stories of AIAs to date focus on historical counterparts in domains such as Environmental Impact Assessments, Human Rights Impact Assessments and Privacy Impact assessments (for example see Moss et al. 2021). Associated with this narrative is an assumption (implicit and in some cases explicitly defined (Metcalf et al. 2021; Moss et al. 2021) where AIAs are reduced to being about identifying, minimising and mitigating harms. This paper explores the narratives currently sitting behind AIAs globally, but with a particular emphasis on Europe and Australia. Through a review of publicly available AIAs, it uncovers a prevalence of language associating the ‘impact’ within AIAs as equating to harms and critically questions why this is the case. Through the lens of affordances, specifically Davis’ mechanisms and conditions framework (Davis 2020), it analyses how existing AIAs currently encourage particular definitions of impact and related orientations towards algorithmic systems. It looks at other disciplines that use the concept of ‘impact’ in ways that seek to uncover both positive and negative impacts, for example in the literature of responsible business and social impact (see for example Rawhouser, Cummings, and Newbert 2019) and explore potential applications to AIAs. In doing so, it questions the fundamental assumptions underlying AIAs as focusing on harms and offers new avenues of exploration to expand the remit of algorithmic responsibility to also include positive impacts.| Period | 13 Jun 2022 → … |
|---|---|
| Event title | Conference abstract contribution -Algorithmic impact assessments (AIAs): A case for moving beyond mitigating harms |
| Event type | Other |