From AI Disclosure to Judgment Traceability: Addressing Research Waste in AI-Assisted Health Sciences Publishing
JMIR Metascience and Research Integrity · 19.10.2026 · 1 · e106068
DOI: 10.2196/106068 ↗Kurz gefasst
Fachartikel in der Zeitschrift JMIR Metascience and Research Integrity, online erschienen am 19.10.2026. Verfasst von Tino Hsieh und 1 weiteren Autor:innen. Die Arbeit nennt 37 Literaturangaben.
Abstract (Originalsprache)
Generative artificial intelligence (GAI) is changing how health sciences manuscripts are produced, revised, and evaluated. Publication guidance emphasizes necessary safeguards, such as disclosure, authorship, confidentiality, transparency, verification, and human responsibility, for AI-assisted content. However, disclosure alone does not establish whether consequential claims in a manuscript are grounded in accountable authorial and domain judgment. This viewpoint addresses this governance gap. We define expression-rich but judgment-thin scholarship as AI-amplified scholarly output in which linguistic fluency, structural coherence, and surface plausibility exceed the traceability of the reasoning, evidence interpretation, uncertainty, contextual limits, and responsibility behind major claims. We frame AI-induced research waste as a conceptual extension of the established research waste literature rather than as a measured prevalence claim. The concern is not that AI-assisted writing is inherently low quality; rather, GAI can make it easier to produce large volumes of polished text whose underlying judgments may be weakly supported, insufficiently limited, or difficult to defend. Building on emerging work on claim accountability, provenance, responsible AI-assisted writing, and publication ethics, we propose domain judgment traceability as a claim-centered governance approach for health sciences publishing. The framework asks whether major clinical, methodological, statistical, ethical, policy, and implementation claims can be traced to an appropriate evidence base, an identifiable interpretive step, explicit uncertainty, population and setting limits, equity and harm considerations, proportionate recommendation strength, and accountable human authorship. This approach shifts the unit of governance from the AI tool or prompt to the major scholarly claim. It also distinguishes disclosure from defensibility: disclosure describes whether and how AI contributed to manuscript development, whereas judgment traceability examines whether the resulting claim can be reconstructed, challenged, and defended by responsible authors. We propose practical implications for authors, reviewers, editors, and journals, including claim-level verification, proportionate requests for clarification, and selective piloting of traceability prompts rather than routine AI detection or exhaustive prompt surveillance. We also identify safeguards against gatekeeping, including attention to language equity, reviewer subjectivity, and the risk of conflating linguistic polish with scholarly competence. Domain judgment traceability is presented as a provisional conceptual governance aid, not a validated scoring instrument or reporting guideline. Its purpose is to complement—not replace—existing disclosure policies and reporting standards by making accountable judgment more visible where claims may influence clinical care, research interpretation, policy, implementation, or public trust. The framework creates an empirical agenda: future studies can test whether claim-centered prompts improve reviewer agreement, citation accuracy, proportionality between evidence and recommendations, author experience, and editorial usefulness without imposing an excessive documentation burden. The goal is not to reject AI-assisted scholarship or require disclosure of every prompt but to ensure that increasingly scalable scholarly expression is proportionate, defensible, context sensitive, and answerable to human judgment in health sciences publishing.
Zitieren
APA
Hsieh, T., & Chen, Y. (2026). From AI Disclosure to Judgment Traceability: Addressing Research Waste in AI-Assisted Health Sciences Publishing. JMIR Metascience and Research Integrity, 1, e106068. https://doi.org/10.2196/106068
Vancouver
Hsieh T, Chen Y. From AI Disclosure to Judgment Traceability: Addressing Research Waste in AI-Assisted Health Sciences Publishing. JMIR Metascience and Research Integrity. 2026;1:e106068. doi:10.2196/106068
BibTeX
@article{Hsieh2026,
author = {Hsieh, Tino and Chen, Ya-Ting},
title = {From AI Disclosure to Judgment Traceability: Addressing Research Waste in AI-Assisted Health Sciences Publishing},
journal = {JMIR Metascience and Research Integrity},
year = {2026},
volume = {1},
pages = {e106068},
doi = {10.2196/106068}
}
EUVIMED