About the Project
Borderline Personality Disorder (BPD) is a serious mental health condition affecting roughly 1–2% of the population. It is associated with chronic emotional instability, self-harm, and frequent psychiatric crises. Despite the high burden of illness, access to evidence-based psychotherapy remains limited worldwide.
MentaBot is an AI-driven chatbot trained on the principles of Mentalization-Based Treatment (MBT) — a psychotherapy approach that helps people understand their own and others’ mental states. The project explores whether MentaBot can serve as a meaningful supplement to specialist treatment, potentially improving outcomes and reducing the pressure on mental health services.
What We Are Investigating
- Quality of AI therapy: Can MentaBot generate responses that are indistinguishable from those of a trained MBT therapist? Experienced clinicians will rate and compare transcripts from real therapy sessions with AI-generated responses.
- Clinical efficacy: Can adding MentaBot to standard treatment reduce self-harm and BPD symptoms, and prevent costly psychiatric readmissions? Outpatients with BPD will be randomized to treatment-as-usual or treatment-as-usual and MentaBot.
The MentaBot interface
MentaBot is currently under development and operates in three experimental modes:
MentaBot is designed to approximate the responses of a trained MBT therapist, generating replies that aim to reflect the principles of mentalization-based treatment. The degree to which these responses are clinically adherent is one of the central questions the project seeks to investigate.
BPD Patient is designed to simulate the expressed experiences and communication patterns of a patient with borderline personality disorder. This mode is intended as a research and training tool, and does not yet fully capture the complexity of real patient presentations. Auto Mode allows the two bots to interact with each other autonomously, generating simulated therapeutic exchanges. This mode is primarily a developmental tool, enabling the research team to study and refine the quality of bot-to-bot interactions before clinical testing.
Research Team & Advisory Board
Principal Investigator: Sune Bo, MSc Psychology, PhD — University of Copenhagen & Psychiatric Research Department, Region of Southern Denmark
CHI Team
- Jacob Sherson — Center for Hybrid Intelligence, Aarhus University, Denmark
- Jafner Rafner — Center for Hybrid Intelligence, Aarhus University, Denmark
- Paul-Arno Lamarque — Center for Hybrid Intelligence, Aarhus University, Denmark
- Florian De Boni — Center for Hybrid Intelligence, Aarhus University, Denmark
- Bianka Szöllösi — Center for Hybrid Intelligence, Aarhus University, Denmark
- Florent Vinchon — Center for Hybrid Intelligence, Aarhus University, Denmark
Advisory Board & Collaborators
- Anthony Bateman — Anna Freud National Centre / UCL; Honorary Professor, University of Copenhagen, UK
- Bo Bach — Center for Personality Disorder Research, University of Copenhagen, Denmark
- Carla Sharp — University of Texas, USA
- Erik Christiansen — University of Southern Denmark; Centre for Suicide Research, Denmark
- Erik Simonsen — University of Copenhagen, Denmark
- Maiken Wolderslund — Centre for Clinical AI, Odense University Hospital & SDU, Denmark
- Majse Lind — Aalborg University, Denmark
- Marie Paldam Folker — Region of Southern Denmark, Denmark
- Mickey Kongerslev — University of Southern Denmark; Region Zealand, Denmark
- Mie Sedoc Jørgensen — Psychiatric Center Ballerup, Capital Region of Denmark, Denmark
- Morten Ellegaard Hell — Psychiatric Unit Aabenraa, Denmark
- Tanja Maria Sheldrick-Michel — Region of Southern Denmark, Denmark
Participating Institutions
MentaBot is a collaboration between clinical, academic, and technological partners:
