On April 23rd, 2025, the Université de Namur hosted the second ALGEPI Annual Workshop where thought-provoking debates between researchers, students and industry experts on how AI is shaping the media landscape took place. From artificial intelligence, algorithmic recommender systems and media regulation, to user-centric innovation, the day was packed with fascinating insights.

Prof. Heritiana Ranaivoson (imec-SMIT, Vrije Universiteit Brussel) kicked off the day by introducing the ALGEPI project – a collaborative effort to tackle the challenges AI poses to epistemic welfare. The project bridges disciplines to generate policy recommendations, citizen toolkits, and publish guidelines for a responsible epistemic welfare-by-design. Bruno Dumas (co-President of the Namur Digital Institute (NADI)) introduced the event’s academic host and the Namur Digital Institute (NADI), highlighting its role in ALGEPI and its interdisciplinary research on digital transformation.

After the introductory remarks, Dieter Boen (Innovation Manager at VRT), delivered a keynote that grounded the theory into practice, with a future-facing vision of the public broadcaster. For the following years, the VRT’s innovation program has at its core the post-platform media with a focus on reliable news, Flemish and European AI models and conversational AI and interactive content, among others. From AI-assisted workflows to conversational news interfaces, AI innovation at VRT is not just about automation – it is about broadening tastes, maintaining its public value, and responsibly using data.

After the keynote, Némésis Srour, VP Product at SPIDEO and Grégoire Bideau, Co-founder of Arvester, tackled the timely question of how content is discovered and who decides what we see on the screens, in a session on discoverability, user agency and explainability. Initially developed for audiovisual content, SPIDEO’s recommender systems expanded into the cultural sectors, with transparency, trust and controllability handed to its clients who can fine-tune how their content surfaces on the screens. Based on thumbnails and existing metadata, Grégoire Bideau gave a fascinating tour using the Arvester tool, of how content visibility across major video streaming services looks like, giving a food-for-thought to those studying the topic on prominence and diversity in catalogues. Both highlighted the urgent need to measure discoverability beyond click-throughs, focusing on catalogue diversity, relevance, and true discovery.

During the lunch break, Dr. Raphaël Tuor (Human-IST Institute at the University of Fribourg) and Juliette Parmentier (PhD researcher at ReSIC at ULB and NADI at UNamur) organised interactive sessions with several prototypes: a news app prototype which contained two newsfeed interfaces to uncover to what extent users understand that their own choices and browsing habits influence what they see and the ‘SwipeBooks’ prototype, whose goal is to get users to reflect about the way they receive literary recommendations by inviting them to reflect on their own ideas about how books should be recommended.

The second half of the day started with the session on filter bubbles, AI regulation and transparency. Dr. Lien Michiels (imec-SMIT, Vrije Universiteit Brussel and University of Antwerp) dissected the elusive concept of filter bubbles. The key message was that we need diverse, but standardised research methods and validated metrics that consider all the stakeholders so that we can gather robust evidence and make meaningful normative judgements. The following presentation was carried out by Lidia Dutkiewicz (PhD Researcher at CiTiP at KU Leuven), who walked us through how the AI Act intersects with media organisations. Although media is not classified as high-risk under the AI Act, transparency obligations remain critical. Her message is to stop talking about ‘AI transparency’ and start defining what we mean, to facilitate internal, external, and third-party transparency.

The last part of the day gave the audience a sneak peek into the more recent ALGEPI work. Prof. Hilde Van den Bulck (AMSoC) and Dr. Willem Buyens (AMSoC) presented the preliminary results of the user survey on news recommender systems launched in Belgium and Switzerland. With a current sample size of 2,860 respondents representing the adult population, the survey included questions on the users’ general media use, their knowledge on algorithms and their benefits and risks, news avoidance and news consumption, and users’ epistemic agency in terms of control, transparency and preferences. The preliminary findings suggest that users are increasingly aware of algorithmic curation but remain wary. Users value reliability and want more control and explanation over the content they’re shown. Measuring users’ awareness of algorithmic recommendations is a moving target. Lastly, PhD Researchers Aina Errando (imec-SMIT, Vrije Universiteit Brussel) and Michelle Kulig (AMSoC) provided a bird’s eye view of the evolving research landscape on news recommender systems via their comparative systematic literature review on the multi-stakeholder challenges and opportunities of news recommender systems in newsrooms. Among others, the presenters highlight the need for more interdisciplinary and multi-method research to address organisational/societal challenges and opportunities of news recommender systems more holistically.

Some key takeaways from the workshop are that:

• Algorithms shape what we see and what we know – thus designing them with epistemic welfare in mind is crucial
• Transparency needs unpacking – vague obligations will not build trust, so we need to define what kind of transparency and for whom
• Discoverability is a cultural metric – it links with matters of prominence, diversity, and algorithmic opacity, and is not just a business metric
• Public media can lead responsibly – with AI tailored to local values
• We need better metrics, better data, and better collaboration – between interdisciplinary researchers and with the industry

The ALGEPI workshop reminded us that AI isn’t just a tool – it is a space of contestation, possibility, and responsibility. The media sector has a unique role to play in shaping AI that informs, includes, and empowers. If you are curious to learn more, the ALGEPI project continues to evolve, and our upcoming publications, tools, and policy recommendations will be essential reading for anyone working at the intersection of media, AI, and society.