“We Need to Move Towards Prevention, Early Detection and Rapid Support”

Mental healthcare in Europe is under severe pressure. As waiting lists continue to grow, increasing numbers of young people are turning to digital tools and AI chatbots. But does technology really offer a safe solution? With the publication of a new pamphlet within the international WeDoMind network, experts are sounding the alarm: not only is more funding needed, but care must also be organised in a fundamentally different way. We spoke with VUB researchers Prof. Imke Baetens and Prof. Martijn Van Heel about the mission of WeDoMind, the pitfalls of commercial AI and the need for European data harmonisation.

More information about WeDoMind

What changes exactly does WeDoMind want to bring about in the field of mental healthcare, and what needs to happen at European level to achieve this?
Imke Baetens: "With the publication of the pamphlet, we are making an urgent appeal to both European and national policymakers. The message is clear: we need not only to provide mental healthcare with more resources, but also to organise the system in a fundamentally different way. At present, our care structure is mainly reactive. We intervene only when there is an acute crisis or severe psychological distress, but by then it is often already very late and care pathways are long and costly.
That focus needs to shift radically. We must evolve towards a system centred on prevention, early detection and the rapid provision of appropriate, low-threshold support. WeDoMind pushes that boundary by acting as an international network that translates scientific knowledge into scalable, concrete solutions. We formulate clear recommendations for European and national policymakers to make this transition possible in practice."

Why is a scalable international network such as WeDoMind necessary for this?
Imke Baetens: "Because mental health problems do not stop at national borders, whereas policies and care structures are still organised in a very localised and fragmented way. If we truly want to make care more holistic and personalised, we need to move beyond the purely medical model in which only the individual patient in the consultation room is central.
A preventive approach only works if you involve the entire environment around an individual: from parents and teachers to GPs and social workers. They form the first line of detection. If we equip them with the right skills and knowledge, they can address mild mental health complaints before they escalate into severe disorders."

Martijn Van Heel: "Exactly. Strengthening that human factor and the social environment is at the heart of the matter. By strengthening the environment, we build a safety net that works faster and more effectively than traditional healthcare systems can currently provide."

"Chatbots Quickly Reinforce the User's Initial Line of Thinking"

Martijn Van Heel

Imke Baetens: "The fundamental problem lies in the algorithm. Chatbots are designed to provide the user with a pleasant experience and therefore have a strong tendency to reinforce the user's initial line of thinking. When a psychologist or therapist speaks with someone, they will reflect, ask critical questions and offer a different perspective. AI does not do that; it simply follows the user's lead.
Moreover, a generic AI system has no ethical framework of its own, no values and no norms. For someone experiencing an acute psychosis or suicidal thoughts, such a chatbot may actually reinforce paranoia or harmful patterns of thinking."

But surely there are safety protocols that automatically direct people to helplines in dangerous situations?
Martijn Van Heel: "Those protocols work when a user explicitly says: 'I can't carry on and I want to end my life.' In that case, a referral to an emergency helpline is displayed immediately. But what happens when such a conversation develops subtly over the course of weeks or months? These models are not trained for that; it is entirely unpredictable which direction the chatbot will take.
And do not forget the practical barriers: what if someone is experiencing a severe mental health crisis during the night, is talking to the free version of an app, and suddenly receives a message saying that their daily tokens have run out? Their only sounding board then disappears abruptly at the most vulnerable moment."

"Specialised Academic Models Are Held Back by a Lack of Funding"

Are there no AI systems that have specifically been designed for psychotherapeutic purposes?
Imke Baetens: "Yes, there are. Within academia and research groups, experiments are being conducted with AI models that are specifically trained on therapeutic principles. A good example is a local project by Thomas More in collaboration with psychiatrist Dirk De Wachter. In that project, an AI model was specifically trained using psychotherapeutic expertise, the appropriate competencies and responsible response styles."

Why do these safe, specialised models struggle so much to find their way into practice?
Imke Baetens: "Large commercial players such as ChatGPT dominate the market, but their models were not essentially built to provide mental healthcare. Specialised academic models run safely on local servers, but they are held back by the lack of funding required to move towards implementation. To bring such a system to market as an official medical device, you need to obtain a medical device certificate. Such certification can easily cost €80,000. Universities simply do not have that money available, meaning that promising and safe research often remains on the shelf."

Martijn Van Heel: "That is a missed opportunity for healthcare. With a locally trained and managed system, it is clear who carries final responsibility, where the data goes and how human supervision is maintained."

"Without the Right Comparable Data, It Is Impossible to Develop Evidence-Based and Effective Policy"

Imke Baetens

How does the renewed structure of WeDoMind fit within all these plans?
Imke Baetens: "The initiative known at the VUB as BRUCC, our clinical service centre, and the PRISM expert network are now joining forces under the new WeDoMind banner. In doing so, we are bringing clinical expertise, scientific research and societal engagement even closer together. WeDoMind therefore embodies precisely the vision that we promote through the pamphlet: connecting science, practice and societal impact."

Martijn Van Heel: "Under WeDoMind, we unite our three pillars: clinical consultations, scientific research and outreach. Through our extensive schools programme, training courses, lectures and practical toolkits, we ensure that academic insights do not remain in a drawer, but reach schools, families, healthcare professionals and policymakers directly. That is how we intend to transform mental healthcare fundamentally and sustainably."