Beyond efficiency: interactional foundations of fairness, accountability and transparency in AI-supported performance evaluation

Abstract

This study examines how features of AI-supported performance evaluation relate to managers' judgements of fairness, accountability and transparency (FAT), and how these judgements relate to trust. Survey data were collected from 289 Australian managers using a standardised performance-evaluation scenario. The hypothesised model, indirect effects, moderation effects, alternative specifications and out-of-sample prediction were assessed using PLS-SEM. Interaction quality, human agency and perceived humanness were positively associated with all three FAT dimensions, whereas voice was associated only with transparency. All three FAT dimensions were positively associated with trust, with transparency showing the largest coefficient. Most specific indirect effects were significant, but task complexity and perceived uncanniness did not moderate the FAT–trust relationships. Human agency showed slightly greater unique explanatory value than voice, while the overall FAT appraisal in the exploratory collective model was strongly associated with trust. The cross-sectional research design identifies associations rather than causal effects, while the Australian, AI-experienced sample limits generalisability. Organisations should combine understandable interaction with contextual sensitivity, human review, documented override authority, traceable responsibility and credible appeal mechanisms. This study brings together human–AI interaction factors to explain trust in AI-supported evaluations. It shows that fairness, accountability and transparency make distinct contributions to trust, while their combined contribution can also be represented as an overall FAT appraisal.