Defining the undefinable: the black box problem in healthcare artificial intelligence

Area of Science:

• Artificial Intelligence
• Medical Ethics
• Healthcare Technology

Background:

• The 'black box problem' in artificial intelligence (AI) creates tension in healthcare applications.
• Current definitions of AI systems are often vague, leading to miscommunication.
• This ambiguity hinders discussions on opaque versus explainable AI in medicine.

Purpose of the Study:

• To propose a clear and coherent definition of the 'black box problem' in AI.
• To facilitate more productive discussions regarding AI in healthcare.
• To address the ambiguity surrounding AI system definitions.

Main Methods:

• Synthesizing existing definitions of the 'black box problem' from literature.
• Examining criteria extrapolated from various definitions.
• Analyzing the implications for AI in healthcare.

Main Results:

• A proposed coherent definition for the 'black box problem' in AI.
• Identification of key criteria to differentiate AI system transparency.
• Clarification of terminology for AI in healthcare.

Conclusions:

• A precise definition of the 'black box problem' is crucial for AI in healthcare.
• Standardized definitions will improve communication among stakeholders.
• This work provides a foundation for future research on explainable AI.