Critical Thinking in AI-Assisted Deontologically-Governed Professional Decision-Making: When and How Explainability, Reliability, and Transparency Matter
Experiments demonstrate reliability and explainability influence critical thinking in AI-assisted professional roles, highlighting design implications.
Key Points
The aim is to explore how explainability, reliability, and transparency in AI support critical thinking in professional decision-making.
Conducted three experiments with professionals in AI-assisted scenarios.
Experiment 1 assessed different levels of AI-generated explanations under consistent reliability.
Experiment 2 analyzed the effects of variable reliability on decision-making.
Experiment 3 evaluated the impact of model confidence scores on transparency.
Minimal reconstructive explanations enhance analytical reasoning when reliability is consistent.
Epistemic uncertainty encourages critical engagement when reliability fluctuates.
Transparency has a limited effect on restoring the benefits of explainability.
Cite This Study
Korosec-Serfaty et al. (2026) studied this question.