However, LSRs require constant screening of new studies, which is time-consuming and labour-intensive. A recent study from GALENOS researchers, published in BMJ Mental Health, explores whether Artificial Intelligence (AI), specifically a Large Language Model (LLM) called GPT-4o, can help lighten the load.
What Is a Large Language Model?
A large language model is a type of AI trained on vast amounts of text to understand and generate human-like language. GPT-4o is one of the most advanced models available today. In this study, researchers tested whether GPT-4o could accurately screen article titles and abstracts to decide if they should be included in a systematic review.
The Study Setup
The researchers used a real-world example: a living systematic review on interventions for anhedonia. (Anhedonia is the inability to feel pleasure, often seen in depression and other mental health conditions.) They created and refined prompts, specific instructions given to the AI, to guide its decisions. These prompts were tested on thousands of records and compared to decisions made by human reviewers.
Key Findings
The results were promising. The AI achieved 100% sensitivity for studies that were ultimately included after full-text screening. This means it didn’t miss any of the most important studies. It also showed high consistency in its decisions and could potentially reduce the number of articles humans needed to screen by 65–85%. “Refined GPT-4o prompts demonstrated high sensitivity and moderate specificity while reducing human workload. This approach shows potential for integrating LLMs into systematic review workflows to enhance efficiency.”
What Do Sensitivity and Specificity Mean?
• Sensitivity refers to the AI’s ability to correctly identify relevant studies. High sensitivity means it catches almost everything important. • Specificity refers to its ability to correctly exclude irrelevant studies. Moderate specificity means it sometimes includes studies that aren’t needed, but that’s safer than missing key ones.
Why This Matters
This study shows that AI can be a powerful assistant in research workflows. It won’t replace human reviewers, but it can help them work faster and more efficiently. That’s especially important in mental health, where timely access to evidence can improve care and save lives. As AI tools become more sophisticated, their role in research will likely grow. This study is a step toward making mental health research more agile and responsive. With careful design and oversight, AI could help researchers keep pace with the growing volume of scientific literature and ultimately, help more people get the care they need.
Read more about GALENOS’s ‘Horizon scanning’ where we identify trending topics in research.