ECAT scientists win Best Paper Award at RecSys 2025 in Prague

The global stage for recommender systems research 

The ACM Conference on Recommender Systems (RecSys) is the premier international forum for presenting new research results, systems, and techniques in the broad field of recommender systems. The 2025 edition, held in Prague, also welcomed regulators such as Ireland’s Digital Services Coordinator (Coimisiún na Meán) and the UK’s Ofcom, underlining the growing policy relevance of recommender technologies. Beyond the technical programme, discussions touched on broader political and social contexts, including the challenges of researcher access to platforms and the role of recommender systems in addressing misinformation, children’s safety, and fairness. 

ECAT takes home the Best Paper Award 

Among hundreds of submissions from around the world, a JRC-led paper received the Best Paper Award, the top distinction at RecSys. 

The winning contribution, You Don’t Bring Me Flowers: Mitigating Unwanted Recommendations Through Conformal Risk Control, was co-authored by Giovanni De Toni (visiting scientist at ECAT from the University of Trento and Fondazione Bruno Kessler), together with the JRC researchers Cristian Consonni, Emilia Gómez and Erasmo Purificato. 

The work introduces a novel, model-agnostic method to limit users’ exposure to irrelevant or potentially harmful recommendations. By applying conformal risk control and leveraging user feedback and watch-time data, the method reduces unwanted content without compromising accuracy. Tests on a large-scale video platform dataset confirmed its effectiveness. 

According to the conference chairs, the paper was unanimously selected by the award committee, who praised both its methodological strength and societal impact. 

A strong ECAT presence at RecSys 2025 

The Best Paper Award crowned a week of intense participation for the European Centre for Algorithmic Transparency (ECAT) and the HUMAINT team, who made several high-profile contributions across workshops, sessions, and tutorials. 

On 22 September, Erasmo Purificato presented Inside the Frame: A Plan for Audio-Visual Feature Analysis of Video Recommendations for Children at the DaQuaMRec workshop, outlining a research approach to link the multimedia features of YouTube videos with children’s engagement. The following day, Emilia Gómez was invited to give a keynote address, “Recommender Systems: A European, Science for Policy Perspective”, where she stressed how the Digital Services Act and the AI Act are reshaping the field. 

On 24 September, visiting scientist Giovanni De Toni took to the stage to present the award-winning paper You Don’t Bring Me Flowers: Mitigating Unwanted Recommendations Through Conformal Risk Control, co-authored with Erasmo Purificato, Emilia Gómez and Cristian Consonni. The next day, Purificato returned to the programme with a late-breaking results poster entitled How Fair is Your Diffusion Recommender Model?, presenting new findings on algorithmic fairness in state-of-the-art diffusion-based models. 

The week concluded on 26 September with a four-hour tutorial led by João Vinagre and Erasmo Purificato, with contributions from Emilia Gómez and Silvia Merisio of DG CONNECT. Entitled “Data Access for Recommender Systems Research: leveraging the EU’s Digital Services Act”, the training combined a detailed explanation of Article 40 of the DSA with a practical exercise in drafting data access requests to Very Large Online Platforms. 

Looking ahead 

Winning a Best Paper Award at RecSys confirms the JRC’s ability to deliver scientific contributions that are recognised at the very highest level while directly supporting European policymaking. 

As the scientific and technical backbone of the Digital Services Act, ECAT’s research helps regulators and policymakers understand how recommender systems work, what risks they pose, and how they can be made safer and more transparent for citizens.