Uncertainty Awareness and Trust in Explainable AI- On Trust Calibration using Local and Global Explanations

Uncertainty information matters for trust: The study found that explaining AI system uncertainty—not just predictions—is crucial for helping users develop appropriately calibrated trust in AI systems, yet this aspect of explainability is often overlooked in current XAI research.

Global explanations deserve more attention: While most XAI research focuses on explaining individual decisions, this paper highlights that global explanations (showing how an AI system works overall) are underexplored and important for building user understanding and appropriate reliance on AI systems.

Visual explanations can improve user satisfaction: The research tested whether intuitive visual explanations could increase user satisfaction and interpretability compared to traditional approaches, suggesting that how explanations are presented—not just their content—significantly affects user experience.

Trust calibration is a key goal: The study emphasizes that effective XAI should help users develop "calibrated trust"—meaning users understand both what the AI can and cannot do reliably—rather than either blind trust or unwarranted skepticism.

XAI needs interdisciplinary guidelines: The paper notes that explainable AI requires input from computer scientists, statisticians, psychologists, and philosophers, suggesting that responsible AI development benefits from diverse expertise beyond technical fields alone.