Joy Paul, a graduate researcher at the University of Nottingham, was poring over thousands of anonymized mental health forum posts when she noticed a pattern—existing AI tools kept misreading despair as anger, grief as apathy. That moment sparked a collaboration that would lead to Emo-MHC, an emotionally aware artificial intelligence model capable of classifying mental health conditions with 92% accuracy. Developed by Dr. Shaily Kabir at the University of Nottingham and Dr. Sangeeta Sangeeta at Keele University, with key contributions from students Joy Paul and Zerin Jahan, Emo-MHC represents a leap forward in how technology can support mental health diagnosis. In a field where early and accurate detection can mean the difference between recovery and crisis, this model doesn’t just analyze words—it interprets emotional texture.
Traditional AI models for mental health classification often rely on structured clinical surveys or self-reported data, which can miss the subtle emotional shifts embedded in natural language. They may flag anxiety but overlook the underlying hopelessness that signals depression. Emo-MHC changes that by integrating deep learning with lexicon-based emotion detection, allowing it to identify not just what a person is saying, but how they’re feeling. By training on publicly available datasets—including doctors’ notes, social media content, and online support forums—the model learned to recognize emotional cues that earlier systems frequently misinterpreted or missed entirely.
When tested, Emo-MHC achieved a classification accuracy of 92%, outperforming standard benchmarks by 8 percentage points—a significant margin in clinical AI. The results, presented at the 2026 International Conference on Innovations in Computational Intelligence (ICICI), suggest a future where clinicians can use such tools to augment their assessments, reducing diagnostic delays and improving treatment plans. For overburdened health systems like the NHS, where mental health services face rising demand and shrinking resources, tools like Emo-MHC could streamline triage and free up clinicians to focus on care rather than paperwork.
Dr. Sangeeta Sangeeta, a data science lecturer at Keele and co-developer of the model, emphasized its broader potential: "In the era of artificial intelligence and large language models, these technologies have significant potential to support individuals experiencing mental health challenges." The team now aims to refine Emo-MHC further, exploring integration into real-world clinical workflows and expanding its ability to detect a wider range of conditions. As mental health diagnoses continue to rise globally, Emo-MHC offers more than technical innovation—it offers hope, one emotionally intelligent algorithm at a time.