AHFE Open AccessThe Transparency Paradox: How AI Explanations Reduce Perceived Autonomy in Organizational Decision-Making
Abstract
Artificial intelligence transparency is widely assumed to enhance user experience, yet this study reveals a paradox: detailed AI explanations reduce perceived autonomy. In a between-subjects experiment (N = 557), business students made organizational decisions with either transparent AI recommendations (detailed rationales) or basic recommendations (minimal explanation). Participants receiving detailed explanations reported significantly lower autonomy (M = 3.73) than those receiving basic recommendations (M = 3.84), d = -0.19. Personality substantially moderated effects: Openness to Experience reversed the paradox (interaction = -0.227), with intellectually curious individuals benefiting from transparency, while Extraversion amplified autonomy reduction (interaction = 0.173). Males showed twice the autonomy reduction of females (0.137 vs. 0.063), and effects disappeared by ages 23-25. Neither AI familiarity, attitudes, nor decision complexity moderated effects, suggesting fundamental psychological responses. Despite reduced autonomy, participants maintained positive AI attitudes, revealing dissociation between momentary decision control and general acceptance. Findings challenge universal transparency mandates and suggest personality-adaptive systems may better serve diverse users.
Keywords: Human–AI Integration, Autonomy Gap, Business Decision-Making, Transparency, Trust in Intelligent Systems
DOI: 10.54941/ahfe1007091
Cite this paper
More from this volume
- Artificial Intelligence Maturity Model (AIMM)
- An Experimental Study on Consensus Building with an AI Chatbot Across Two Topics
- An Agent-Based Simulation Framework for ADHD: Modeling Attention Regulation and Adaptive Therapeutic Interventions
- CRMSON: Co-Designing Adaptive and Ethical AI Systems to Address Mental Health Barriers in Aviation
- Usability Evaluation of FAIR Data Planning in the Data Stewardship Wizard
- Seeing the Invisible Load: XR + Multimodal Sensing for Cognitive Ergonomics in Industrial Training
- Conceptual Framework for Designing Domain-Specific LLM-Based Information Systems
- Shaping Conversations: Custom GPTs to Spark Reflection in Design
- Privacy at the Core: Toward Automated Detection of Privacy-Sensitive Content in an LLM-Based Care Documentation Support System
- Dynamic Difficulty Adjustment via Dynamic Scripting: An Empirical Study of Player Flow in a Brawler Game
- Sinusoidal time-based features and human error metrics: Advancing software defect prediction in safety-critical systems
- Designing an Experimental Method for Evaluating Divergent Thinking with a Color Queue under Time Constraints