Shaping AI Through Futures Thinking
230 AI researchers, members of the Flanders AI Research Program (FAIR), gathered in Mechelen on March 10, 2025, for the Flanders AI Internal Research Day to explore where AI might take us over the next 15 years.
The participants took part in an interactive Futures Thinking workshop led by Ben Robaeyst (imec-MICT-UGent) and supported by Griet Capelle (imec). Together, we imagined Flanders in 2040, reflecting on what we hope AI will achieve and on the risks that must be prevented. We also explored the concrete actions we can take today to guide AI-development in a responsible direction.
By discussing both the expected (what the participants believe the future of AI will look like) and preferred futures (what they believe it should look like), the participants formulated ideas and suggestions for the further development of AI. The workshop served both as a first step in collective Futures Thinking and as a moment of reflection, confirming that the goals of the Flanders AI Research Program are still highly relevant and meaningful. The session focused strongly on learning, about the possibilities of AI, the importance of intentionally shaping the future of AI and on sharing the research being done by the participants. Some clear and recurring visions for AI’s future stood out, combining an optimism about the societal benefits of AI with a sober awareness of its potential pitfalls.

What the future may bring
The FAIR community has a nuanced vision of the future of AI — centered on the promotion of human well-being, fairness, and sustainability as well as on the prevention of harmful outcomes. Participants broadly agreed on AI’s potential as tool for societal improvement and economic development: it should serve people’s needs and enhance quality of life, primarily by reinforcing human decision-making and maintaining human oversight, keeping “humans in the loop.”
AI is considered a powerful instrument across various domains.
In healthcare and personal assistance, it can support faster diagnoses, suggest personalized treatments, and monitor health risks. In the Medical Imaging use case, for instance, AI-driven analysis can improve diagnostic accuracy and efficiency. Similarly, in the Single Cell Technologies use case AI helps to integrate and visualize complex biological data, aiding in personalized medicine.
AI can also boost productivity by automating mundane tasks, freeing up people to engage in more creative and meaningful work. The Human-Machine Flexible Production use case, in which AI systems collaborate with humans on manufacturing floors, enhancing flexibility and efficiency is a great example of this.
In parallel with these concrete applications, the FAIR community also envisioned a future rooted in strong ethical values, forming a synergetic relationship between humans and AI. A cross-cutting theme was the importance of “human-centric AI”: AI should be a collaborator and supporter, not a replacement.
Another recurring theme was the need for a fair AI with respect for human rights. There should be principles promoting social justice, inclusivity, and equitable access to opportunities. Participants emphasized AI’s role in supporting disadvantaged groups, distributing resources fairly, and reinforcing social welfare systems — e.g., matching unemployed individuals with jobs or providing guidance to students in skills and career paths. In the Education & Training use case, adaptive learning platforms use AI to personalize education and upskilling opportunities. Trust and transparency are essential: AI systems should be explainable and fair, opaque “black box” technologies should be rejected. Open data, auditing mechanisms, and transparent decision-making processes are advocated to keep AI inclusive and trustworthy. At the same time, participants were very aware about potential risks. If misused (e.g., AI nudging, disinformation, or militarization), AI could deepen inequality and undermine human dignity. Participants warned against algorithmic biases and stressed that AI should be governed by ethical standards — our moral standards should not be outpaced by technological development.
Finally, environmental sustainability is both an aspiration and a concern. AI can help tackle climate change—optimizing energy use, guiding precision agriculture, and predicting disasters. For example, the Smart Grid use case focuses on AI-driven decision support for low voltage distribution grids, enhancing energy efficiency. Similarly, the Robotics for Agrifood use case explores AI applications in agriculture to address challenges like climate change and resource management. But participants also warned about AI’s energy use, urging designs that minimize carbon footprints. In short: AI must help heal the planet, not harm it.

We’re not there yet…
What can we do?
If we want to build on that desired future, we need to take concrete steps today. During the workshop, the researchers outlined actions across education, research, governance, and infrastructure, sending a strong message: change must start in the classroom and through public awareness. The community advocated for AI and data ethics in school curricula, teacher training, lifelong learning, and national literacy campaigns — some even joked about an "AI driver's license". An informed public is essential for responsible AI. Research must refocus on public-interest problems via mission-driven programs tackling health, climate, and social equity. Funding should align with societal priorities—supporting both fundamental and applied research.
Governance recommendations were both regional and global. The proposals ranged from the creation of regional oversight bodies (e.g., a Flanders or European AI agency ), over independent test centers, to enforceable legal standards on transparency, data protection, and limits on intrusive AI use.
Some participants saw the need for international treaties on autonomous weapons or AI norms, arguing that AI’s risks transcend borders and demand collective action.
Open-source development and decentralization were emphasized as key strategies to democratize access to artificial intelligence. By making AI models, computational resources, and development tools publicly available and accessible, these efforts aimed to reduce barriers to entry, prevent monopolization by a few dominant actors, and foster a more inclusive and equitable AI ecosystem where individuals, communities, and smaller organizations can also participate on equal footing.
A lot of attention was paid to environmental sustainability. Proposals included investing in energy-efficient hardware, optimizing algorithms, reporting AI’s energy footprint, and using incentives like carbon credits. AI must be green by design.
Transparency and accountability were also recurring themes: certification of AI models, dataset audits, and standardized evaluations for high-stakes deployments (e.g., in hospitals or courts). Ethical checks should be embedded in every development stage.
Finally, a holistic approach is essential. Education, technology, policy, and research must reinforce each other. AI literacy, for example, fuels better regulation, and research supports sound lawmaking.
Conclusion
What emerged from the Research Day was not a single vision, but a rich spectrum of ideas — ranging from concrete applications in healthcare and education to calls for stronger governance, public engagement, and responsible innovation. The emphasis throughout was on ensuring that AI systems remain trustworthy, meaning transparent, accountable, and aligned with human values.
There was a strong sense of both opportunity and responsibility. Many spoke of AI’s potential to contribute to societal well-being, environmental sustainability, and a more equitable access to resources. At the same time, concerns were raised about risks such as bias, loss of control.
The workshop showed that technological development does not occur in isolation. In a rapidly evolving world, the relevance of today’s choices—across education, research, regulation, and infrastructure—was a recurring theme. While the workshop offered a snapshot of current thinking, it also highlighted the importance of continuous reflection, collaboration and coordination. In the coming years, the ability to stay attuned to research and new developments in AI and respond with thoughtful, well-informed action will remain essential.
The summary was based on Ben Robaeyst’s analysis of the workshop discussions and does not constitute an official position of the whole of the Flanders AI Research Program.





















