Hace poco decía que en The Energy Journal todavía no estábamos teniendo muchos problemas con la IA. Antes hablo...el viernes, revisando un paper, me encontré con referencias claramente fabricadas por la IA. Así que rechacé el paper:
Before sending your paper to reviewers, I read it again, and found that many of the references are fabricated. They have been mixed up, clearly pointing to the use of AI. The fact that you have not acknowledged their use is in clear breach of the AI policy of our journal, and therefore I must reject your paper.
La respuesta del autor:
Thank you for explaining the basis for the rejection of XXX. You are correct, and I owe you both an acknowledgment and an apology. The bibliography for this manuscript was assembled with the help of our research assistants who used a large language model to expand a partial reference list into BibTeX. The references were not adequately cross-checked before submission, and the AI-use disclosure that the journal requires was not included.
The responsibility for both is mine, as I should have audited their work before submitting, and I did not. Most embarrassingly, my own 2020 XXX paper is among the entries with a wrong title and pages; that one alone I should have caught at a glance. After receiving your decision, I checked all 36 references against CrossRef DOI metadata and publisher pages. Seven were wrong (mixed-up authors, wrong journals or titles) and have been corrected, with DOIs added across the full bibliography. I will also apply the same audit to a parallel manuscript in progress, which I am planning to submit to EJ, so that this cannot recur.
I understand the decision on XXX stands. If you would be open to a fresh submission of the corrected manuscript at some future point, I would welcome that opportunity. The resubmission would include full AI-use disclosure, every reference DOI-verified, and our complete replication package (data, code, and DCE design files) attached so that the empirical work is independently checkable. If not, I understand entirely. Either way, thank you for catching this; it is a more useful outcome than going to peer review with a broken bibliography.
Y ahora, ¿qué hago?¿Cómo recuperamos la confianza perdida en este autor? ¿Seguro que esto es lo que queremos enseñar a los investigadores en formación?