Entry

Evaluating Cognitive Biases in AI-Assisted Mammography Interpretation: A Simulation Reader Study of Explainable AI Across Radiologist Experience Levels

Filippo Pesapane, Antuono Latronico, Francesca Abbate, Silvia Penco, Anna Rotili, Valeria Dominelli, Luca Nicosia, Enrico Cassano

Synopsis

Monocentric, fully crossed simulation reader study (Mar–Jun 2024) with 6 breast radiologists reviewing 200 mammograms under three conditions (unassisted, AI-assisted, AI-assisted with saliency XAI). In deliberately discordant cases without explanations, automation bias occurred in 36.1% and anchoring bias in 33.9%; XAI roughly halved both (to ~17%). Least-experienced radiologists were most susceptible.

Keywords

·mammography ·automation bias ·anchoring bias ·explainable AI ·radiologist experience

Open paper ↗ Report issue ↗

Related entries

Automation Bias in AI-Assisted Medical Decision-Making Under Time Pressure in Computational Pathology

November 1, 2024 · arXiv
The Deskilling Effect: Is Artificial Intelligence Eroding Clinical Competence?

June 2, 2026 · Annals of Internal Medicine 2026
In ChatGPT they trust: a study of students' perceptions and misuse of ChatGPT in higher education

December 1, 2025 · AI and Ethics 2025
Impact of AI recommendation correctness on diagnostic accuracy in clinical decision-making

November 19, 2025 · International Journal of Medical Informatics 2026
Human-AI experience in integrated development environments: a systematic literature review

March 8, 2025 · Empirical Software Engineering 2026
Trust and reliance on AI — An experimental study on the extent and costs of overreliance on AI

2024 · Computers in Human Behavior 2024