Explaining AI for Health Solutions and the Black Box Concept

• The concept of explainability is a very hot topic in artificial intelligence (AI).
• Explainability is the concept of how well you understand how a model makes decisions
  • This is often referred as XAI, and it is an attempt to make the black box of machine learning algorithms a little more transparent, and maybe move from the black box concept to a glass box concept.

• In medicine and health, clinical explainability is of paramount importance:
  • AI remains a pure technical tool and it integrates in a workflow.
    • We need to be able to explain the utilization of an AI intervention to providers, so that we understand how decisions can be made or are made with the model predictions made by the algorithm.
    • We also need to be explain this to patients such as explaining how it impacts their care delivery.

• The answers to these questions are critical for building trust, and trust is critical to building an open mindset to adoption and adaptation into workflows.

Note that the presenter has used the concept of Augmented Intelligence instead of Artificial Intelligence. Please refer to the introductory article of this term.

Ivy Lee, MD. Augmented Artificial Intelligence in Dermatology. Core concepts and clinical decision making. 8th World Congress of Teledermatology, Skin Imaging and AI in Skin diseases – November 2020