Tackling Algorithmic Transparency in Communal Energy Accounting through Participatory Design

Abstract

Algorithmic transparency presents a significant challenge to system developers and users of algorithmic systems alike. Framing theproblem as a ‘wicked’ one, this study tackles the issue of transparency in the EnerCoach energy accounting tool through presentinga situated ethnography of the algorithmic system and exploring the issues and challenges of model transparency and post-hocexplainability therein. By engaging stakeholders through participatory design methodologies, both a conceptual understanding ofthe problem and material solutions thereof are developed and evaluated. The findings show the promising potential of participatorydesign methodologies to elevate users to a ‘critical audience’, and the solutions co-created by the study participants for the challengeof algorithmic transparency. The results also highlight the complexity of the problem: transparency of algorithmic systems must beunderstood as a multi-facetted and highly contextual, ‘wicked’ problem that requires diverse methodological interventions to reach‘satisficing’ solutions.

Publication
ACM Communities and Technologies Conference 2021