What does it really take to build artificial intelligence that people can trust?
This question brought researchers, engineers, ethicists and policy experts to the Alpine town of Martigny, Switzerland, for the ELOQUENCE Plenary Meeting. For two days, participants gathered at the Idiap Research Institute to reflect on progress, exchange ideas and shape the final phase of a European project dedicated to building conversational AI systems that are fair, inclusive, and aligned with societal values.
A meeting place for ideas and people
Nestled in the Rhône Valley and surrounded by dramatic mountain landscapes, Martigny may seem like an unexpected place to discuss the future of artificial intelligence. Yet the city has long been a crossroads – historically connecting routes across the Alps and today serving as a meeting point for science, culture, and innovation.
This spirit of connection made Martigny an ideal host for the ELOQUENCE consortium. Beyond the meeting rooms, informal conversations continued over coffee breaks and evening walks through the town, reminding participants that meaningful collaboration often happens between sessions as much as during them.
The meeting took place at Idiap Research Institute, a globally recognized center for research in artificial intelligence, machine learning and human-centered technologies. Idiap’s work focuses not only on advancing algorithms, but also on understanding how technology interacts with people, a perspective that closely aligns with ELOQUENCE’s mission.
Moving beyond “smart” AI
From the very first sessions, it was clear that ELOQUENCE is not just about making AI systems smarter. Discussions focused on coordination, ethics and long-term responsibility – the often invisible foundations of trustworthy technology.
Project partners reviewed progress, upcoming milestones and challenges, while also reflecting on how to ensure transparency, accountability, and compliance with European values. Ethics and data protection were discussed as living processes, continuously revisited as systems evolve.
Fairness, languages and real people
One of the central challenges addressed during the meeting was bias – particularly in multilingual and multicultural settings. Conversational AI systems are increasingly used in professional and high-stakes contexts, from public services to customer support, where misunderstandings or biased behavior can have serious consequences.
Rather than treating bias as a purely technical problem, participants examined it from multiple angles: language, culture, usability, and human oversight. The question was not simply “Does the system work?”, but “Who does it work for and who might it leave behind?”
Learning from pilots and real-world use
Pilot studies played an important role throughout the meeting, grounding abstract discussions in practical experience. Partners shared insights from testing conversational AI systems in realistic environments, highlighting both successes and limitations.
Importantly, the conversation did not shy away from failure. Understanding where systems struggle – whether due to language complexity, user expectations, or contextual ambiguity – was seen as essential for improvement. In ELOQUENCE, learning from limitations is considered a strength, not a weakness.
From bias to fairness: asking better questions
A highlight of the second day was a dedicated session with the project’s Community of Experts, framed around a deceptively simple question: Are we asking the right questions when we talk about fairness in AI?
Discussions explored how trustworthiness can be assessed beyond standard benchmarks, emphasizing expert judgment, human feedback, and real-world evaluation. The session underscored an important insight: building trustworthy AI is not about finding a single perfect metric, but about maintaining an ongoing dialogue between technology and society.
Looking ahead
As the meeting drew to a close, attention turned to the future. The final phase of ELOQUENCE will focus on integration, demonstration and sharing results with wider audiences – from policymakers and industry to the general public.
The days in Martigny served as a reminder that the future of AI is shaped not only by code, but by collaboration and reflection… Against the backdrop of the Alps, the ELOQUENCE consortium reaffirmed a shared commitment: to build conversational AI systems that people can understand, question and ultimately trust.
