1. Alucinaciones bibliográficas y la contaminación del registro científico.
El uso de modelos de lenguaje está introduciendo referencias inexistentes y datos fabricados, erosionando la confianza en la literatura académica.
- Hallucinated citations are polluting the scientific literature.
- I never saw a journal article or a student essay with a reference to a paper that didnt exist until people started getting help from chatgpt.
- Ai-generated references commonly combine fragments of genuine publications. Joe Shockman, co-founder and chief executive of Grounded AI, calls such references frankenstein citations.
2. El colapso del sistema de revisión por pares y la "recursión de IA".
La saturación de manuscritos de baja calidad y el uso de IA para evaluar otros trabajos generados por IA amenaza con crear un sistema sin supervisión humana real.
- If ai writes the manuscripts and ai reviews them, its ai turtles all the way down. Journals need to figure out what human judgment still means in this pipeline.
- Peer review is in a death spiral, says uchicagos , and ai writing tools are speeding it up.
- Ai submissions being reviewed by ai reviewers means unfiltered slop in the publication record. Its depressing as hell.
3. La presión por publicar y la degradación de la calidad del contenido.
El modelo de "publicar o morir" se ve exacerbado por herramientas que permiten la producción masiva de artículos, priorizando la cantidad sobre el rigor.
- Ai will help researches produce more papers okay but for what purpose? no, really why? who actually believes there is a lack of publications out there?
- The current state of ai tools, amplified by existing publish-or-perish incentives, appears to be pushing the system toward an equilibrium of more rather than better research.
- Submissions up 42, writing worse, and the editors had to double their staff to cope.
4. Falta de transparencia y el desafío de la detección.
A pesar de las políticas editoriales, la mayoría de los autores no declaran el uso de IA, lo que dificulta la validación de la autoría y la originalidad.
- Only 0.1 of academic papers published since 2023 have explicitly disclosed the use of ai for writing assistance, yet textual analysis suggests that the actual rate of ai use is 40 times higher.
- Scientists are failing to disclose their use of ai despite journal mandates.
- If the legal academy were willing to treat undisclosed ai use as career-ending in the same way that plagiarism is and for the same reasons, it could stick.
5. Potencial de la IA como herramienta de apoyo técnico.
Algunos sectores de la comunidad científica defienden el uso de IA para tareas mecánicas, limpieza de datos y mejora de la legibilidad, siempre bajo supervisión humana.
- Ai can speed up research, but the citations it generates often dont hold up once you check them.
- It codes faster, filters better, and scales research. But it still cant generate true insight or decide what matters. Thats still human work.
- Do use it for formatting and copy-editing. Know your target journals guidelines.