This article is taken from the "Expert" section of Omalius magazine #36 (March 2025).

In early February 2025, the AI Act, the world's first general legislation on AI that frames its use and development came into force in Europe. As a specialist in human-machine interaction, does this new framework reassure you?

At UNamur, within my research group, we focus our work on the user side and his interaction with technology. When it comes to AI, we are particularly focused on this notion of transparency, which is reflected in the principle of the AI Act. How does AI make decisions? What data is it based on? What are its operating processes? Can it explain them? This need for AI transparency is of paramount importance to the user. For the time being, however, this is blocked from a purely technical point of view, essentially because of the gargantuan amount of data that AI uses to function, to train itself. At present, only experts are really capable of understanding how AI works. However, since AI is often a tool for the citizen, the need for transparency must also, and above all, be accessible to the citizen. At UNamur, a great deal of research is being carried out along these lines.

For example, you're working with doctors on the degree of trust they have in AI as part of their profession: what's it all about?

It's about an AI system that should, in particular, enable doctors to help them identify tumors on medical images. The challenge? For the doctor to know whether the answer provided by the AI is reliable, and how reliable it is. We are developing and testing this process with doctors. A process that will enable the AI to give them its degree of certainty. Early feedback shows that this transparency will be fundamental.

Today, are you confident in the way citizens are appropriating AI?

I'm fascinated by these emerging and multiple uses. Now, whether we're writing a greetings card, summarizing a text, organizing a meeting, making a cake recipe or writing an e-mail, we're turning to AI. I don't think we have any bad habits, but I'm more worried about the lack of awareness of the need for transparency in the way AI works. There's a need for information, awareness and education. We're working on this, including at UNamur. With this in mind, 24 colleagues and I have launched a course on the challenges and opportunities of AI, accessible to all university students, whatever their discipline. But this is very clearly an area that needs to be strengthened and accelerated so that it progresses at the same pace as the development of technology.

Another technology that's making inroads into the everyday life of the citizen, and which you're studying closely, is augmented reality: where do we stand?

Are we going to trade in our smartphones for smart glasses? The answer is most likely yes, and in the relatively near future! So I'm studying what's going to happen to the user when there's an extra digital layer grafted onto their environment, onto what they see. We mustn't leave this control exclusively to the tech giants, who all have such prototypes in the pipeline. My job is to find out how, from a technological point of view, we can give more control to the user. How can he filter what he sees? How can he define what information he wants to see, how much, etc.? Our aim is to give him the tools to keep control over these future augmented reality systems.

What kind of tools?

For example, we're developing techniques that allow the user to filter the elements they want to see in real time. At present, existing augmented reality tools give very little power to the user. We're working to reverse this trend. We're also making sure that this presence of augmented reality is for the benefit of the user, to enable them to better understand their environment.

More generally, does the technology adapt sufficiently to the user's needs?

No, too often the user just has to endure these technological developments. My approach as a researcher is the opposite: it's up to the system to adapt to user needs. Every development must be carried out in dialogue with the user. This is why our work lies at the crossroads of computer science research and principles inherited from applied psychology. Because we must, above all, understand how the user functions before we can develop technologies that are more relevant, more effective, more legitimate and better adapted.

Whether it concerns linguistics, economics, politics, sustainable development, law, history, educational sciences, literature or translation, research in the humanities and social sciences is as eclectic as it is rich and essential for tackling society's challenges. Of UNamur's eleven research institutes, seven are directly involved in SHS research. While there is a high degree of complementarity in these areas of research, better pooling of resources, sharing and easier access to certain services, resources and support will help to sustain and strengthen the excellence of SHS research at UNamur. It is with this in mind that the SHS impulse platform has just been created.

• Axis 1 - Support for the acquisition of databases, documentary resources and software
• Axis 2 - Subsidy for cutting-edge training in the use of specialized methods
• Axis 3 - Funding access to the SMCS "Support en Méthodologie et Calcul Statistique" platform at UCLouvain, thanks to an inter-university partnership.
• Axis 4 - Setting up an SHS space, containing a laboratory for running experiments and shared work tools promoting exchanges between researchers.

This initiative, launched in January 2025, addresses the specific challenges faced by SHS researchers. The long-term aim is to sustain and expand the services. "We will also hire a researcher expert in methodological analysis in SHS who will be able to inform innovative methodologies and frame the methodological design of research projects," emphasizes Sandrine Biémar, vice-dean of UNamur's Faculty of Education and Training Sciences, a member of the IRDENA institute and the SHS Impulse management committee. "The wish is also to support networking between SHS researchers at UNamur and to be a lever for setting up interdisciplinary projects," adds Sandrine Biémar.

The platform's management team is made up of representatives of the university's various SHS institutes, and ensures efficient management of resources. The platform's impact will be assessed during its initial phase (2025-2027), enabling strategies for its sustainability and development to be defined.

This article is taken from the "The day when" section of Omalius magazine for December 2024.

Four major teaching and research disciplines have marked the Faculty's development and become its pillars over the years: economics and political and social sciences first, then management and communication. "In its early days, the Faculty of Economic, Social and Management Sciences, created by Father Camille-Jean Joset, was united around social sciences and economics," recalls Pietro Zidda, Dean of the EMCP Faculty. "Then, the various fields developed. Management took off, enrolments in political science and communications soared. We were careful to maintain a spirit of collaboration between each of our programs". Far from the usual silos, the Faculty today makes it a point of honor for its chosen disciplines to collaborate, question and nurture each other in order to develop the skills of students and researchers so that together they can contribute to the challenges of a society in transition.

The EMCP Faculty is committed to three major missions. The first is to train responsible experts and decision-makers, through rigorous, hands-on teaching that stimulates critical thinking and openness to the world. It also aims to conduct ambitious, interdisciplinary research with a strong scientific impact, feeding into teaching and innovation. Finally, the EMCP Faculty wishes to act as a responsible player in societal development, sharing knowledge and contributing to informed decisions at regional, national and international levels.

It is therefore quite natural that the FSESG has become the EMCP Faculty, a name now incorporating communication and political sciences and reflecting the importance they have acquired in recent years. Four disciplines united to prepare students and researchers in a transdisciplinary way for the challenges of tomorrow.

The spirit of collaboration is firmly rooted at the heart of the Faculty, which strives to develop transdisciplinary approaches to meet the complex challenges of a society in transition. "To meet these challenges, a solution from a single discipline is no longer enough. We need to think more broadly, with an approach that transcends disciplines," explains Anne-Sophie Collard, Vice-Dean of the EMCP Faculty. A sentiment shared by Zora Gilet, a management engineering student: "This new name above all brings coherence to the Faculty's image and visibility for all the courses on offer. It also represents the intra-faculty diversity that we wish to promote at all levels."

This vision is also accompanied by an internal reorganization, with the creation of four thematic schools or schools: UNamur School of Economics (NSE), UNamur School of Management (UNSM), UNamur School of Social Sciences, Politics and Communication and UNamur School of Evening Studies in Economics and Management. These schools aim to strengthen synergy between disciplines, while promoting a pedagogy that integrates cross-disciplinary skills and innovative working methods. Soft skills, for example, are now systematically integrated into projects, to prepare students to respond to societal challenges in a collaborative and creative way. "I think this change helps to concretize and recognize all the large-scale projects that have been set up in recent years," explains Zora. It's an expression of a desire to develop and innovate, which is more than positive today. I consider myself lucky to be able to witness this change and proud to be part of this community."

"EMCP aims to be the catalyst lever for a future where walls are broken down and barriers between disciplines abolished as much as possible, to provide strong solutions to societal problems," concludes Pietro Zidda. This new name therefore goes beyond mere naming: it symbolizes a renewed commitment to transforming the way graduates are trained, giving them the tools they need to provide strong, coherent solutions to major contemporary challenges.

Within the Faculty, various projects and teaching experiments illustrate this EMCP vision. Examples?

Learning by doing: an approach that offers immersion in concrete projects from the first year, combining knowledge and cross-disciplinary skills to respond to real-life challenges.

Regards croisés: this project invites students and teachers to explore a topical issue from a variety of disciplinary angles, enriched by exchanges with experts in the field at a major final conference.

A little wink, the Dean of Faculty wondered how this fresco would be perceived by an outside audience: "This work is quite original compared to what we usually do. So I asked the artificial intelligence to give me its interpretation. And then, surprise, the answer was bluffing! ChatGPT perfectly identified the meaning and intentions of the project, as if it had read our initial brief", laughs the dean.

Students have also been involved in this transformation, notably through their circles. Thus, the Cercle €co became the Cercle EMCP. "We were contacted by the Dean, who suggested that our Circle should bear the same name as the Faculty, and this seemed to us to be a perfectly natural move. We had many discussions with the dean and the vice-presidents of the Cercle. We put a lot of effort into the name change, but it was an extremely rewarding experience," explains Matthieu Dupuis, President of the Cercle EMCP. "The change may have come as a surprise to some students, but this new name enriches the Faculty's image by enhancing the value of all its courses of study. It embodies strong values and, in my opinion, represents our Faculty better than the old one."

"The current generation of large language models (LLMs), like ChatGPT, learn language in a very different way," continues Paul Van Eecke. "By observing in huge quantities of text which words often appear in close proximity to each other, they learn to generate texts that are often indistinguishable from human texts. This leads to extremely powerful models in many forms of text generation, from text synthesis or translation to question answering, but which at the same time have a number of inherent limitations. They are thus prone to hallucinations and biases, often struggle with forms of human reasoning, and require huge amounts of data and energy to build and use."

The researchers propose an alternative model in which artificial agents learn language as humans do: by participating in meaningful communicative interactions in their environment. In a series of experiments, they show how these agents develop linguistic constructs directly related to their environment and sensory perceptions. This leads to linguistic models that :

• Are less prone to hallucinations and biases because their understanding of language is based on direct interaction with the world.
• Manage data and energy more efficiently, leaving a smaller ecological footprint.
• Are more grounded in meaning and intention, enabling them to understand language and context in a more human way.

"The integration of communicative and situated interactions into AI models is a crucial step in the development of the next generation of linguistic models. This research offers a promising route to linguistic technologies that approximate the way humans understand and use language," the researchers conclude.

Source: Katrien Beuls, Paul Van Eecke; Humans Learn Language from Situated Communicative Interactions. What about Machines?Computational Linguistics 2024; 50 (4): 1277-1311. doi: https://doi.org/10.1162/coli_a_00534