Human Trust in AI-Assisted Project Decision Systems: A Mixed-Methods Analysis of Explainability, Adoption, and Managerial Decision Quality

Abstract

This study investigated human trust in AI-assisted project decision systems by examining the relationships among explainability, perceived reliability, AI adoption, managerial reliance, and objective managerial decision quality. A mixed-methods analytical framework was applied, with the principal quantitative analysis based on 527 participants from 46 project-oriented organizations representing multiple managerial roles, industries, organizational structures, and project environments. Mean professional experience was 11.8 years, 72.9% of participants had previous experience with AI-assisted project systems, and 64.7% had received AI or analytics training. Human trust averaged 3.71, explainability 3.66, perceived reliability 3.74, behavioral intention 3.83, and managerial reliance 3.47 on five-point scales. Explainability was strongly associated with trust (r = .64, p < .001), while perceived reliability demonstrated an even stronger relationship (r = .68, p < .001). Explainability remained an independent predictor of trust (β = .36, p < .001), and adoption significantly predicted managerial reliance (β = .41, p < .001). Experimental comparisons showed that explainable AI-assisted decision making achieved the highest accuracy at 86.4%, compared with 82.1% for conventional AI assistance, 79.3% for AI-only recommendations, and 73.8% for human-only decisions. The corresponding error rate declined to 13.6% under explainable AI assistance. Explainable AI users accepted 90.6% of correct recommendations but only 13.2% of erroneous recommendations, demonstrating stronger calibrated reliance and reduced automation bias. The overall experimental effect was substantial (partial η² = .239), while explainable AI exceeded human-only performance with Cohen’s d = 1.50. The integrated multivariable model explained 49.9% of variance in managerial decision quality, with calibrated reliance emerging as the strongest independent predictor (β = .29, p < .001). Overall, the findings demonstrated that explainability strengthened trust and adoption, but superior managerial performance depended primarily on calibrated reliance, verification, organizational governance, digital readiness, and continued human oversight. Explainable human–AI collaboration therefore provided the strongest combination of decision accuracy, interpretability, accountability, and managerial control.