Mathematical framework could improve transparency of AI in clinical settings
MedicineHealth AdministrationData Science
THE AI ANGLE
making clinical predictions while concealing hidden reasoning through information leakageResearchers at King's College London have developed a mathematical framework to identify and quantify 'information leakage' in concept-based AI systems used in healthcare. While these models are designed to use clinically meaningful concepts to appear interpretable, hidden and unintended information can still drive predictions behind the scenes, mimicking a black box. This framework introduces measures for concepts-task and interconcept leakage to ensure clinical AI models provide genuine transparency and support meaningful human oversight.
THE TEACHING ANGLE
Instructors can explore the critical tension between apparent and actual model interpretability, examining how an AI system can present clinically intuitive explanations while secretly relying on hidden, unreviewed data.Read the original at techxplore.com Generate teaching or study materials
More in Medicine
- AI’s best coding agent fails 60% of the time — and the data backs it upThe New Stack · September 15, 2026
- Report covers advances in surgical robots, deploying warehouse automation in the real worldThe Robot Report · September 15, 2026
- Observing Earth from orbit is unpredictable and messy: Could 'liquid' AI clear the view?Phys.org — Technology · September 15, 2026
- AI helps pathologists find signs of preeclampsia, advancing diagnosis and treatmentMedical Xpress · September 15, 2026
- Patient identity is the missing control in healthcare AIHealthcare Dive · September 14, 2026