How was the study conducted?

The Youth-GEMs project is a large, EU-funded research initiative aiming to improve youth mental health by better understanding of how genes, environment, and experience interact. A key output of the project is an App that aimes to monitor and predict mental health changes.

To understand how decisions were made during the app’s development, a diverse group—including psychiatrists, AI engineers, biomedical researchers, social scientists, ethicists, and, young people with lived experience of mental health challenges—came together in a multilogue: a structured, open conversation that aims to give all voices  equal weight. Together, they explored each person’s contribution, their hopes and expectations for the role of AI in the app, but also the concerns and critical questions they brought to the table.

Before this, however, the researchers examined both the original EU funding call and the YOUTH-Gems research proposal. This helped them to point out what motivated the researchers to include AI in the app’s development in the first place and what those who designed the funding calls considered important—and why.

What were the main findings?

By the end of the study, the team identified three key areas where important decisions had to be made, which directly shaped the outcomes:

1. Whether AI is seen as useful depends on one’s understanding of mental health.

If mental health is seen through a biomedical lens – as a disruption in brain function – then AI could  be helpful. It might detect patterns in behaviour, sleep, or phone use that signal risk or mental health changes and thereby might allow for early intervention.

However,  if mental health is seen as a phenomenological concept – something deeply personal, rooted in emotions, relationships, and self-understanding—then AI modelling will not be useful as people need to learn to reflect upon themselves, their emotions and behaviours.

2. Co-creation with end-users is crucial but complex

The team who developed the app worked closely with young people who had experience of mental health challenges. These young advisors helped shape the app’s look, tone, and content identifying user needs and highlighting blind spots the technical teams hadn’t considered. Their input led to changes in the wording of prompts and visuals, making the app feel more supportive and less triggering.

But co-creating the AI model itself—the coding and algorithm development (modelling, validating and re-modelling)—appeared impossible and was not done. . These technical decisions were made by AI experts and psychiatrists. While everyone valued inclusion, some areas (like early-stage coding) were too technical for meaningful joint work.

3. The goal of autonomy and empowerment was widely shared, but there are differences of what these concepts mean to  different disciplines.

The technical experts (engineers and AI-experts) mainly focused on compliance to existing regulations and judicial requirements and considered respect for autonomy as providing informed consent and ensuring data-protection.

Young people and some psychiatrists rather focused on what is actually helpful for them and saw autonomy as having the possibility to express self-reflection, activeinvolvement and agency.

What were the main implications of the study?

In this study, researchers concluded that:

• Research decisions often begin before a project starts. Innovative research is dependent on funding calls criteria, which means there is limited freedom to decide whether to use AI in the first place, or how to include it in the app development.

• How mental health is understood deeply shapes which tools are seen as useful. A biomedical concept of mental health is likely to benefit of algorithmic calculations, and more so if supported by encompassing Al-tools. A phenomenological concept of mental health is not in need of complex algorithms of behavioral, emotional, physical and social features.

• Co-creation with young people improves opp-development/design but balancing co-creation with technical complexities is challenging and not always feasible.

• Legal compliance is crucial but it doesn’t always equal meaningful empowerment. Sometimes, new ethical questions arise because of those regulations.

Most importantly, transdisciplinary multilogues enable explicit scientific self-reflection, render disciplinary differences visible, and foster constructive critical engagement.

This lay article was created in collaboration with Dorothee Horstkötter, Assistant Professor at Maastricht University and leading researcher at Youth-GEMs.  You can access the full publication here: https://www.sciencedirect.com/science/article/pii/S2666659625000150?via%3Dihub