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

Version 2.1.0 adds the results annex (separata.pdf): a self-contained, skimmable PDF presenting the results of the evaluation (figures, tables, and the reproduction instructions), extracted from the parent manuscript Evaluating ethical compliance: artificial intelligence research in the European Union. The annex names the analysed project; the interpretation, remediation pathway and reform discussion are carried by the parent manuscript. R analysis scripts, pre-rendered summary tables, figures, and summary statistics for 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); the tables that these scripts consume are in the companion data record (v1.0.0 of this Zenodo record, see relatedIdentifiers). In order to ensure the analysis stays at institutional level (avoiding personal attribution to researchers), individual publications are referenced only by an opaque identifier (pub_uid, P001–P0NN). Bibliographic information (titles, author names, venue names, DOIs) is not in any file in this record. Contents: scripts/analysis.R + scripts/plots.R (R scripts; tidyverse-based; consume the tables to produce rendered tables and figures), tables/*.md (11 pre-rendered summary tables: per-axis distributions, channel-mismatch, year-trend, risk-collapsed, etc.), figures/ (rendered plots), summary-stats.md (long-form summary statistics). Reproducibility: with the tables (v1.0.0 record) downloaded into the same directory level, running Rscript scripts/analysis.R and Rscript scripts/plots.R regenerates all rendered outputs. The R scripts use only standard tidyverse packages and require no network access.