The Comparative Agendas Project (CAP) held its 2026 annual conference on 17–19 June in Paris, hosted by Emiliano Grossman and the CDSP (Sciences Po / CNRS). On the opening day, poltextLAB presented its joint project with Syracuse University's Maxwell School: an AI-coded dataset of more than 15 million New York Times articles spanning 1851–2025. The dataset was presented by Nick D'Amico, with questions from Rens Vliegenthart and the panel audience.
There was considerable interest at the conference in using the babelmachine.org classification tool for coding policy agendas. Alison Craig (University of Texas at Austin) presented a group project with Sean Theriault and colleagues that builds on the Babel Machine and improves its accuracy through a combined human–machine coding design. Sebők described this human-in-the-loop approach as the most realistic direction, noting that general and domain-specific models alone can only go so far in addressing the specific research questions of individual studies.
At the closing plenary, Sebők joined Frank Baumgartner and Shaun Bevan in a session moderated by Grossman. The discussion centred on their co-authored article in the journal Communication and Change, which provides an overview of how agenda studies can be conducted in the age of AI. The panel addressed the role of human researchers in the analytical process and the adoption of these tools by startup projects without compromising quality. The panellists noted that first-hand coding experience remains an essential part of training new content-analysis researchers, regardless of advances in AI.
The plenary also reviewed the current state of the Comparative Agendas community and included presentations of new projects. David Levi-Faur delivered the conference keynote on the previous day.
