Legal Affiliates’ Views on Algorithmic Decision Making

Abstract

Procedural fairness in algorithmic decision support systems has been extensively examined across various disciplines. This study investigates how future legal practitioners perceive fairness and accountability in such systems, and how their views diverge from those of computer scientists. In a replication of earlier research, we recruited 150 legal-affiliated crowdworkers via Prolific to: (a) rate their agreement with statements related to six fairness constructs across three scenarios, (b) define algorithmic fairness, (c) identify causes of unfairness, and (d) express their views on accountability. Our findings show that legal affiliates’ perceptions of algorithmic fairness differ from those in previous studies. Participants often defined fairness as the absence of discrimination, emphasizing the quality of the system’s output. They cited ‘sensitive attributes’ as a primary source of unfairness and held the ‘company using or owning the system’ accountable when unfairness occurred.