Scientific Integrity in Artificial Intelligence: Evidence from a Horizon Europe Research Project

Tables from a corpus-level risk-classification analysis of the publications of an EU-funded research programme on artificial intelligence in the political-communication domain. The classification scheme — four axes (topic, venue, geography, disciplinarity), per-axis operationalisation, and risk-gradient thresholds — is documented in the companion framework documentation (see relatedIdentifiers). In order to ensure the analysis stays at institutional level (avoiding personal attribution to researchers), the tables refer to individual publications by an opaque identifier (pub_uid, P001–P0NN) and contain no bibliographic information (titles, author names, venue names, DOIs). Aggregations report counts and percentages across the 75-publication corpus. Contents (in derived-tables/): a per-publication enriched table (year, venue type, data source, classification levels, risk score) plus 27 aggregation tables along axes and cross-tabulations (per-axis distributions; risk by venue, by year, by discipline; channel-mismatch; ethics-authorisation and DPIA counts; year-trend; country and topic risk; LLM use; data-source and output-category breakdowns). A subsequent version (2.0.0) will add the R analysis scripts, the rendered summary tables, the figures, and the summary statistics that consume these tables to reproduce the empirical reporting of the corresponding manuscript.