We had the opportunity to chat with @nuglobalnews.bsky.social about our latest meta-analysis examining how well #ML / #AI predicts mental health outcomes.
Some promising advances (as the title suggests), but we must also proceed with caution 👇
#psychscisky

news.northeastern.edu/2025/09/17/n...

As we look towards the future of ML in mental health research, we hope this paper provides a basic, birds-eye view of existing findings in the literature. 📖🔭
4/4

That being said, several limitations affect our current use of ML for predicting treatment response that would benefit from consideration in future studies (small sample sizes, class imbalances in response outcomes, lack of simple baseline algorithms for benchmarking, etc.).
3/4

Interestingly, ML does a decent job of predicting treatment response (AUC=0.80 & Accuracy=0.76). Studies using biological features (e.g., fMRI) and robust cross-validation strategies reported better predictive performance.
2/4

Paper alert for people interested in machine learning for #mentalhealth 📣

We’ve put together a new meta-analysis examining how well we can predict treatment response for emotional disorders, now out in Clinical Psych Review 🎉

www.sciencedirect.com/science/arti...
1/4

Yay!!! Celebration time!