Exploring Trust Factors in Designing AI-Supported Performance Appraisal System
TL;DR
The findings suggest that trustworthy AI integration requires addressing pre-existing structural failures before introducing algorithmic decision support, and offer design recommendations for AI-supported appraisal systems, with implications for HCI research and organizational practice.
Abstract
Performance appraisal is a critical organizational process that directly influences employee development, compensation, and career progression. Despite growing interest in Artificial Intelligence (AI)-assisted appraisal systems, little is understood about the factors that make such systems trustworthy from the perspective of those who use them. This study addresses this gap by investigating the factors that shape trust in AI-supported performance appraisal across three stakeholder groups: managers, employees, and Human Resource (HR) managers. Using semi-structured interviews with 12 participants across four countries and reflexive thematic analysis, the study identifies three overarching themes from the manager group. First, trust in the current appraisal process is already structurally compromised by recency bias, unacknowledged subjectivity, and opacity. Second, trust in AI is contingent upon non-negotiable boundaries, including explainability, data quality, and human oversight. Third, AI integration carries dual implications, where it can either restore fairness through longitudinal consistency or deepen distrust through bias amplification and privacy invasion. The findings suggest that trustworthy AI integration requires addressing pre-existing structural failures before introducing algorithmic decision support. These results offer design recommendations for AI-supported appraisal systems, with implications for HCI research and organizational practice.
