Why People Reject Good Algorithms
Understanding algorithm aversion in the workplace — and how explainability, control, and trust design overcome resistance to AI recommendations.
Abstract
Research on algorithm aversion consistently finds that people reject algorithmic recommendations that would improve their outcomes — particularly after witnessing a single AI mistake. In workforce contexts, this rejection carries significant costs: managers ignore useful coaching prompts, HR teams distrust team health signals, and employees disengage from growth recommendations. This paper examines the psychological mechanisms behind algorithm aversion and the design interventions — explainability, human control, confidence transparency — that build sustainable AI adoption.
Key Findings
Algorithm aversion is significantly reduced when users can observe the AI's reasoning process and data sources.
Confidence levels and uncertainty communication — not just recommendations — increase appropriate trust calibration in AI users.
Users who feel in control of AI recommendations (can modify, reject, or request review) show higher long-term adoption and trust.
Single AI errors create lasting trust damage unless the system provides transparent explanation and correction mechanisms.
Human override quality improves when workers regularly practice independent judgment before reviewing AI suggestions.
How this connects to RUDY
Leadership Quality OS
Help managers prepare for better 1:1s, recognition moments, difficult conversations, and team interventions.
AI Trust & Governance
Make workforce AI explainable, reviewable, auditable, and safe for sensitive people-related decisions.
Human Expertise Preservation Engine™
Prevent AI from quietly weakening judgment, deep expertise, independent problem-solving, and human mastery.
Apply this research
See RUDY in action
RUDY's platform is grounded in this research. Explore the live demo to see how these principles are built in.
Access DemoResearch-backed from day one.
Every RUDY AI module is grounded in peer-reviewed research and evidence-based frameworks.
No surveillance. No black-box scoring. Human review where it matters.