I serve on the advisory board of Language Science Press, a high quality open access publisher in linguistics. I alo co-edit a book series with the press. A recent query to the board concerned AI in submissions. I share my answer here because it may be useful beyond the LangSciPress community. Lightly edited to make more sense as a standalone post.
I support the Press taking a clear stance in this matter. I think synthetic text (AI slop, i.e. the output of LLMs like ChatGPT) is fundamentally incompatible with scientific and scholarly writing, and with core values of research integrity.
A widely used code of conduct for research integrity in the Netherlands defines five values: honesty (reporting processes accurately and being open about margins of uncertainty), scrupulousness (being precise and thoughtful), transparency (showing one’s process and allowing others to build on it), independence (being impartial and unswayed by commercial or political interests), and responsibility (accepting accountability for the statements made).
Synthetic text undermines every single one of these values. By design, synthetic text is
- not honest: it is text tuned to be maximally likeable and credible
- not scrupulous: it is frequently inaccurate and the opposite of thoughtful
- not transparent: it erases authorship and is essentially a form of automated plagiarism
- not independent: it reproduces and amplifies biases and represents no original thinking
- not responsible: it is not produced by a human author with agency and accountability
(If anybody is interested in the scientific sources to back up these statements, they are available here, alongside more general guidelines for GenAI and research integrity. The phrasing above draws on my earlier post about science blogging. The term ‘automated plagiarism’ is credited to Iris van Rooij.)
In our own series, we have seen the occasional GenAI-enhanced submission and rejected these, not on the grounds of it being GenAI-produced (which can be hard to prove conclusively), but simply on the grounds of scholarly quality.
We have not articulated an AI policy for our series yet. We believe it is covered by general principles of research integrity, as above, but for the benefit of authors it would be helpful to spell it out.
Therefore, I think it would make sense for the Press to adopt a clear and unequivocal policy that discourages authors from submitting synthetically generated text and that invites them to uphold basic standards and core values of research integrity.