How competitors become collaborators-Bridging the gap(s) between machine learning algorithms and clinicians.

Area of Science:

• Medical imaging analysis
• Clinical decision-making
• Artificial intelligence in healthcare

Background:

• Machine learning (ML) algorithms demonstrate high diagnostic accuracy in medical image analysis.
• Successful integration of ML tools into clinical practice remains a significant challenge.
• Ethical concerns may arise from the interaction between ML algorithms and clinicians.

Purpose of the Study:

• To investigate epistemic and normative factors contributing to algorithmic overreliance in clinical settings.
• To identify key challenges in fostering effective collaboration between clinicians and ML algorithms.
• To propose desiderata for bridging the gap between ML capabilities and clinical practice.

Main Methods:

• Analysis of epistemic factors: false expectations and miscalibration of uncertainties.
• Examination of normative factors: non-explainability and socio-technical context.
• Exploration of the collaborative dynamic between clinicians and ML algorithms.

Main Results:

• Algorithmic overreliance is influenced by factors such as unrealistic expectations and poor uncertainty assessment.
• The non-explainable nature of some ML models and the surrounding socio-technical environment pose significant hurdles.
• A dialectical relationship exists, requiring adaptation from both clinicians and algorithms for successful integration.

Conclusions:

• Overcoming challenges in ML-clinician collaboration requires addressing issues beyond algorithmic accuracy.
• Improving the understanding and management of ML uncertainties is crucial for safe clinical adoption.
• Successful implementation necessitates a dual approach, involving both algorithmic refinement and clinician adaptation.