Evaluating Artificial Intelligence's Role in Developing Research Questions in Head and Neck Reconstruction

Area of Science:

• Medical research methodology
• Artificial intelligence in healthcare
• Reconstructive surgery

Background:

• Generative artificial intelligence (AI) large language models (LLMs) like ChatGPT and Gemini are emerging technologies.
• While clinical applications of AI are discussed, its role in research development is less evaluated.
• The potential of AI in generating novel research ideas for systematic reviews (SRs) remains underexplored.

Purpose of the Study:

• To assess the role of ChatGPT and Gemini in generating novel and clinically relevant research ideas (RIs) for SRs.
• To evaluate the novelty of AI-generated RIs in the field of head and neck reconstruction.
• To compare the effectiveness of ChatGPT and Gemini in this research ideation process.

Main Methods:

• ChatGPT and Gemini were prompted to generate 10 RIs each for SRs in head and neck reconstruction domains.
• A comprehensive literature search was conducted in MEDLINE, Cochrane Library, and Embase to assess RI novelty.
• AI-generated RIs were screened against existing SRs to determine novelty and clinical relevance.

Main Results:

• A total of 60 RIs were generated by both AI models.
• Literature search identified 50 studies partially addressing the AI-generated RIs.
• Out of 60 RIs, 42 were found to be novel (not previously studied).
• No statistically significant difference in performance was observed between ChatGPT and Gemini.
• AI-generated RIs were generally broad but clinically relevant.

Conclusions:

• Both ChatGPT and Gemini can effectively generate novel and clinically relevant research ideas for SRs.
• AI has the potential to assist researchers in identifying new research avenues, particularly in complex fields like head and neck reconstruction.
• Further research may explore refining AI prompts for more specific and targeted research questions.