Counterfactual explanations and algorithm aversion

Description

This dataset supports research on job applicant reactions to hiring rejection by human, AI, and hybrid (human with the support of AI) evaluators, and whether counterfactual explanations improve these reactions. The data come from two online experiments in which participants read hypothetical rejection scenarios and then reported their responses. The dataset includes measures such as fairness, trust, organisational perceptions, and behavioural intentions. It can be used to examine how evaluator type and rejection message type shape applicant reactions, and how these effects operate through cognitive and affective perceptions.

Steps to reproduce

Participants were recruited via Prolific, and the study was conducted in Qualtrics, which was used to present the scenarios, assign conditions, and collect responses. Participants read hypothetical hiring rejection scenarios varying by evaluator type (human, hybrid, AI), and in Study 2 also by rejection message type (no message, outcome-only, counterfactual explanation). Participants then completed the study measures. Data were screened before analysis, and statistical analyses were conducted in IBM SPSS Statistics using ANOVA, ANCOVA, and Hayes’ PROCESS macro.

Institutions

• University of Liverpool

  England, Liverpool

Categories

Licence