Ghost in the machine or monkey with a typewriter-generating titles for Christmas research articles in The BMJ using...

Area of Science:

• Medical research
• Artificial Intelligence in Medicine
• Scientific Publishing

Background:

• The British Medical Journal (The BMJ) publishes Christmas-themed research articles.
• Evaluating the quality of research titles is crucial for engagement and discoverability.

Purpose of the Study:

• To assess the plausibility and engagement of AI-generated titles for potential Christmas research articles in The BMJ.
• To compare AI-generated titles with historical human-generated titles.

Main Methods:

• An observational study utilized Generative Pre-trained Transformer 3 (GPT-3) to generate titles.
• 25 human participants evaluated the AI-generated titles against historical controls.
• Metrics included plausibility, attractiveness, enjoyability, and educational value.

Main Results:

• AI-generated titles were rated similarly to control titles for enjoyability and attractiveness.
• Human-curated AI titles showed improved scientific merit compared to uncurated AI titles.
• The most plausible AI title was related to conspiracy theories and vaccinations; the highest-rated was about coffee in emergency departments.

Conclusions:

• AI can produce plausible and engaging titles for medical research articles.
• Human intervention significantly enhances the scientific value of AI-generated titles.
• AI shows potential in assisting with scientific title generation, particularly with human oversight.