Feasibility of physician-developed expert systems

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Expert systems in medicine offer decision support.
• Direct physician implementation of these systems requires user-friendly tools.

Purpose of the Study:

• To evaluate physician usability of a domain-independent expert system generator.
• To determine if physicians can create medical expert systems without a knowledge engineer.

Main Methods:

• A four-year study involving 70 medical students.
• Students were trained to use an expert system generator.
• Students built two small medical expert systems and were assessed.

Main Results:

• No significant correlation between prior computer experience and performance.
• Most students found the generator easy to use (32/47) and a valuable teaching tool (43/47).
• Students successfully developed medical expert systems.

Conclusions:

• Physicians can learn to directly implement medical expert systems using domain-independent software.
• The need for a knowledge engineer intermediary may be reduced.
• User-friendly software facilitates physician adoption of AI tools in healthcare.