REthinking MEntal health through Clinical and Data Intelligence
Artificial intelligence has demonstrated remarkable diagnostic capabilities in medicine, often surpassing human performance in identifying complex conditions. However, a critical gap remains between the technological potential of AI and its effective clinical implementation. Despite their high performance, many AI models function as “black boxes,” making it difficult - if not impossible - for clinicians to understand the reasoning behind a given diagnosis or therapeutic recommendation.
The REMEDI Lab was founded on the belief that the future of digital medicine lies not in automating diagnosis, but in developing intelligent systems that can meaningfully interact with clinical experts -providing not only accurate results but also clear, clinically grounded explanations.
People
Heads and co-founders
- A. Raballo
- A. Mira
- Fabrizio Barazzoni, Co-director
Scientific advisory board
- Prof. Lonneke van der Plas
Professor of Natural Language Processing
USI Università della Svizzera italiana
- Dr. Esaú Villatoro-Tello
Research Associate — Speech and Audio Processing Group
Idiap Research Institute (Switzerland)
- Strategic advisor
Davide Gai
President, Healthcapital SA
- Prof. Marina Timoteo
Professor of Comparative private law
University of Bologna and Director AlmaLaurea
Expert in Law and Digital Technologies in China
Partners
- Roberto Cianella
Direttore Generale
Federazione Cantonale Ticinese dei Servizi Autoambulanze, FCTSA
Research Collaborators
- Anish Mukherjee, SNSF PostDoc, USI
- Ziqing Dong, SNSF PostDoc, USI
- Matteo Gasparin, PostDoc, USI
- Federico Ravenda, PhD student, USI
- Volodymyr Karpenko, PhD student, USI
- Daniele Montagnani, PhD student, University of Pavia and USI
- Gualtiero Marenco Turi, Master student, USI
Objectives
Our objectives are structured around three main pillars:
- Explainable AI for medicine
We aim to design AI methods tailored to the medical context, capable of articulating their decision-making process in clinically relevant terms. The goal is not merely algorithmic transparency, but the translation of computational logic into reasoning that mirrors clinical thinking.
- Natural Language Processing in healthcare
We explore the transformative potential of advanced NLP techniques to interpret clinical documentation and to detect early diagnostic signals through the analysis of spontaneous digital communication, such as social media content.
- Methodological integration of clinical expertise and AI
We are committed to building validation frameworks that respect the rigor of medical research while addressing the specific needs of machine learning systems, thus creating a bridge between traditional clinical paradigms and cutting-edge AI.
Our methodological approach integrates diverse data sources: from structured clinical datasets collected in outpatient and psychiatric settings, to conversational data from digital platforms where individuals seek support for their health conditions. This triangulation enables us to capture both the formal clinical dimension and the more spontaneous, authentic expressions of illness experience.
We aim to develop data- and evidence-centric methods that are scalable, reproducible, interpretable, and equipped with robust uncertainty quantification, allowing for effective, sustainable, and trustworthy support to clinical experts.
Ultimately, our goal is to create decision-support tools that do not replace clinical judgment, but enhance it - providing transparent, validated computational insights that contribute to more precise, accessible, and patient-centered care.
Research
Research
-
Preventive and precision psychiatry, with Antonio Preti, Professor, Professor, Torino University and Michele Poletti, Chief Psychologist, IRCCS, Reggio Emilia, Italy
- Differential privacy, with Conor Daniel Hassan, PostDoc Aalto Universit, Kerrie Mengersen, Queensland University of Technology, Australia
- Federated learning, with Bei Jiang, Professor, University of Alberta and Jack Jewson, Senior Lecturer, Monash University
Scientific references
- Ravenda F., Preti A., Poletti M., Crestani F., Mira A., Raballo A. (2025). Transforming Social Media Text into Predictive Tools for Depression through AI. PLOS Digital Health.
- Ravenda F., Bahrainian S.A., Raballo A., Mira A., Kando N. (2025). Are LLMs Effective Psychological Assessors? Leveraging Adaptive RAG for Interpretable Mental Health Screening. In Proceedings of Association for Computational Linguistic conference, Long Paper, Main Conference 2025.
- De Grandi A, Ravenda F, Raballo A, Crestani F. The Emotional Spectrum of LLMs: Leveraging Empathy and Emotion-Based Markers for Mental Health Support. In Proceedings of CLPsych @ NAACL 2025.
- Ravenda F, Bahrainian SA, Kando N, Mira A, Raballo A, Crestani F. Tailoring adaptive-zero-shot retrieval and probabilistic modelling for psychometric data. In Proceedings of the 40th ACM/SIGAPP Symposium on Applied Computing 2025.
- Ravenda F, Kara-Isitt FZ, Swift S, Mira A, Raballo A. From Evidence Mining to Meta-Prediction: a Gradient of Methodologies for Task-Specific Challenges in Psychological Assessment. In Proceedings of CLPsych @ NAACL 2025.
- Ravenda, F., Bahrainian, S. A., Raballo, A., & Mira, A., Navigating through the hidden embedding space: steering LLMs to improve mental health assessment. In Proceedings of the 41st ACM/SIGAPP Symposium on Applied Computing 2026.
- Ravenda, F., Preti, A., Poletti, M., Mira, A., & Raballo, A., Rethinking psychometrics through LLMs: how item semantics shape measurement and prediction in psychological questionnaires. Nature Scientific Reports, 15(1), 37313 (2025).
- Di Noia, Antonio, Federico Ravenda, and Antonietta Mira. "A general framework for adaptive nonparametric dimensionality reduction." To appear in Nature Scientific Reports, 2025.
- Raballo, A., Ravenda, F., & Mira, A., Diagnosing schizophrenia spectrum disorders: Large language models (LLMs) vs. leading international psychiatrists (LIPs). Psychiatry and Clinical Neurosciences, 79(9), 599 (2025).
Media and Scientific Dissemination
