Data underlying the paper "Towards Effective Human Intervention in Algorithmic Decision-Making"

This project captures perceptions of ability, benevolence, integrity and fairness towards different decision-maker configurations in algorithmic decision-making for policy enforcement. It consists of two studies. In the first study, we interview 21 participants with experience renting out their properties out as short-term rentals. Through these interviews we identify the main characteristics that interviewees considered when evaluating the appropriateness of decision-maker configurations (i.e., decision-maker profile, model type, data provenance). In the second study, we run a crowdsourced user study in Prolific with 223 participants. They were shown an illegal holiday identification scenario. Each participant was assigned to a scenario with a decision-maker configuration with varying decision-maker profile, different model types, and data provenance. The dataset includes:<br>The prompts used for running the interviewsThe interview protocolPre-registration of the crowdsourced studyThe materials used for designing the crowdsourced user studyThe (anonymized) data from the crowdsourced studyThe script used to analyze the crowdsourced dataA document with all the rationales behind the statistical analysis of the crowdsourced dataAnalysis of open-ended questions of the crowdsourced study