Performance of Generative Artificial Intelligence in Dental Licensing Examinations

Area of Science:

• Dental Education
• Artificial Intelligence in Healthcare
• Professional Licensure Examinations

Background:

• Generative artificial intelligence (GenAI), including large language models (LLMs), presents potential applications in healthcare and education.
• The proficiency of LLMs in interpreting dental-specific information and providing accurate answers for dental licensing examinations remains largely unexplored.

Purpose of the Study:

• To investigate the accuracy of GenAI in answering multiple-choice questions from US and UK dental licensing examinations.
• To compare the performance of different versions of GenAI (ChatGPT 3.5 and 4.0) on these examinations.

Main Methods:

• 1461 multiple-choice questions from US and UK dental licensing exams were used.
• Two versions of ChatGPT (3.5 and 4.0) were tested.
• Performance was analyzed across examinations and dental subjects.

Main Results:

• ChatGPT 4.0 achieved scores of 80.7% (US) and 62.7% (UK), passing both exams.
• ChatGPT 3.5 scored 68.3% (US) and 43.3% (UK), failing both exams.
• ChatGPT 4.0 demonstrated superior performance, answering more questions correctly than ChatGPT 3.5.

Conclusions:

• The latest GenAI versions exhibit notable proficiency in answering dental licensing examination questions.
• While promising, GenAI requires further refinement for consistent performance in all dental contexts.
• GenAI holds significant potential to impact dental professional training and patient communication.