🔎 Unchecked AI in Academic Publishing: The AACR’s Warning Sign The American Association for Cancer Research (AACR) recently published sobering findings: in 2024, 23% of abstracts and 5% of peer reviews submitted to its journals contained text likely generated by large language models (LLMs), yet less than a quarter of authors disclosed that AI was used, even though disclosure is mandatory. What tools did they use? A detector created by Pangram Labs, trained on millions of human-written documents, plus synthetic “AI mirror” texts. It achieves ~99.85% accuracy (very low false positive rates). Why this matters: • Transparency isn’t just academic correctness—it’s essential for trust, reproducibility, and accountability. • Misused or undisclosed AI can introduce errors, especially when methods are paraphrased or rephrased without care. • There are equity dimensions: non-native English researchers are using LLMs more often. Proper support + disclosure matters. What we should do moving forward: 1. Enforce disclosure rules rigorously (journals, conferences, reviewers). 2. Integrate AI-detection tools in the submission/review pipeline. 3. Educate researchers on what AI can’t do: respecting nuance, domain-specific correctness, context. 4. Standardize policies across publishers so the expectations are clear. 🔍 My take: If we don’t adapt oversight and norms, we risk eroding core academic values. AI is a tool, not a substitute—but we need to treat it with full transparency. 😶🌫️ What do you think: Should journals issue retroactive notices or corrections when undisclosed AI use is found? How should authors safely use LLMs without risking misrepresentation? —- Ali Fatemi, Ph.D., MCCPM, DABMP Director of Medical Physics, Merit Health (CHS) Southeast Professor of Physics, Jackson State University, USA Adjunct Professor, Dept of Physics, Université Laval, Canada Founder & CEO, SpinTecx http://www.spintecx.com #AcademicIntegrity #AIethics #Publishing #LLMs #ResearchPolicy #MedicalPhysics #AIinHealthcare #MedTech #HealthTech #DigitalHealth #SmartHealthcare #FutureOfRadiology #ClinicalAI #ClinicalInnovation #MedicalImaging #Radiology #VendorNeutral #GlobalHealthTech #Standardization #PrecisionMedicine #HealthcareLeadership #PatientSafety #StartWithWhy #HealthcareStartups #VentureCapital #AngelInvesting #VCFunding #StartupFunding #TechStartups #HealthcareAI #AcademicEntrepreneur #ClinicalEntrepreneurship #ClinicalTranslation #AIInRadiology
this is an important conversation. i think transparency is key in maintaining trust, especially in research. enforcing disclosure and using detection tools sounds like a solid start. as for retroactive corrections, that might depend on the severity of the impact. it’s a tricky balance but we need clarity in how we integrate AI. how do you see authors navigating this landscape without crossing ethical lines?
💠In the name of super efficient God Medical physicist | Radiotherapist | Al & Python developer | Review & systematic researcher
1w👏👏