This summary contains an analysis of the recorded workshop on “Generative AI and scientific development - STOA Workshop”, Tuesday, 29 April 2025, 13:00-16:00 CET, European Parliament, Brussels. Room Spinelli 5E2 and webstreamed. The summary is twofold; first, it shows the synthesised Key challenges and underlying tensions systematically organised following the STEEPED matrix. Second, the main statements and described challenges of each panellist are listed.
Sources of this analysis are the recorded video (here) and the booklet (here).
1. Challenges of Open Science & Generative AI in the Scientific Exchange - Analysis of the STOA Workshop Generative AI and scientific development
The analysis collected all single statements of the panellists and synthesised them below the following STEEPED categories:
- Social – research cultures, collaboration norms, trust
- Technological – reliability, transparency, interoperability
- Economic – costs, market concentration, business models
- Environmental – energy use, computational infrastructures
- Political/Legal – EU law coherence, sovereignty, governance frameworks
- Ethical – integrity, fairness, disclosure obligations
- Demographic – inequalities across regions, disciplines, career stages
S – Social
Research cultures, collaboration norms, trust
Key challenges
- Erosion of trust in scientific communication due to AI-generated or AI-assisted outputs whose provenance is unclear (Bouter, IJsselsteijn, Belongie).
- Acceleration of publication and communication cycles, reducing time for deliberation, critique and meaningful peer interaction.
- Homogenisation of scientific language and argumentation, leading to loss of plurality, creativity and dissent in scientific debate (Ventura, IJsselsteijn).
- Shift from dialogical exchange to mediated exchange, where researchers increasingly interact through AI summaries rather than direct engagement with peers.
- Changing norms of collaboration, including uncertainty over acceptable AI use in writing, reviewing, and grant preparation.
Underlying tension