Public Statement
Serious accusations of “fraud” have been raised against the article “Authorship Verification of the Disputed Pauline Letters through Deep Learning” by Evy Beijen and Dr. R. de Heide, published in HIPHIL Novum in 2025.
These accusations have been thoroughly investigated by the editorial board of HIPHIL Novum, assisted by an external expert on AI and the New Testament. No evidence of fraud has been found, and we see no reason to retract the article.
Accusations raised
- The programming script was AI-generated.
- The deep learning model uses 100-word chunks, which is claimed to be insufficient for a BiLSTM, and other alleged methodological flaws have been mentioned.
- The code provided in the GitHub repository is not reproducible.
Response
HIPHIL Novum does not prohibit the use of AI-generated code. On the contrary, authors are encouraged to use the best tools available to further research in the area of biblical linguistics. Authors remain fully accountable for all stages of their research and for all parts of the manuscripts they publish, including any AI-generated content. See our Ethics and AI policy for further information: https://tidsskrift.dk/hiphilnovum/about
Papers submitted to Hiphil Novum undergo rigorous external double-blind peer review to ensure their scholarly quality. In this particular case, an additional external specialist was invited to evaluate both the article and the written, unpublished accusations. While some objections to the paper are indeed valid and warrant further discussion, most of the objections and the conclusions drawn from them were judged to be unsupported or disputed. The critic was invited to submit a formal article in response, and HIPHIL Novum stands ready to publish it once he has addressed the issues identified in the external expert’s review.
The most serious charge concerns the reproducibility of the results. For the sake of transparency, it is essential that all data be made available in a way that ensures they can be found, accessed, and reproduced. We acknowledge that we tested only the findability and accessibility of the data and scripts, but not their reproducibility. The authors have now been requested to update their GitHub repository accordingly.
In conclusion, we do not see any evidence of fraud, and we consider the accusations far-fetched and insufficiently sensitive to the clearly experimental nature of the paper. We welcome expert reviews of this article and encourage further debate on the use of AI and the application of statistical models to long-standing issues in biblical linguistics.
Christian Canu Højgaard
General Editor of HIPHIL Novum
