Purpose In scientific notation, “a.u.” denotes atomic units, but in figure labels it is also widely used to mean arbitrary units. This technically incompatible dual usage motivated this study, which examined the prevalence, disciplinary distribution, and editorial handling of the use of “a.u.” to denote arbitrary units of intensity in peer-reviewed publications. Methods A cross-sectional bibliometric content analysis was conducted using 1,915 research articles published between 2020 and 2024 in six representative high-impact journals. Articles with intensity axes labeled “a.u.” were identified and coded by field, context, and whether the intended meaning was explicitly defined. Results Of the 1,915 articles sampled, 1,140 (59.5%) contained at least one figure with “a.u.” on an intensity axis. Among these articles, only 14.3% explicitly clarified that the label denoted arbitrary rather than atomic units. Significant inter-journal variation was observed: Optics Express had the highest prevalence of “a.u.” intensity axis labeling (71.1%) and the lowest clarification rate (8.1%), whereas PLOS One had the lowest prevalence (41.9%) and the highest clarification rate (25.7%). Conclusion The conflation of “a.u.” as atomic units with “a.u.” as arbitrary units is a pervasive, editorially tolerated notational ambiguity that can undermine scientific precision and reproducibility. The unambiguous alternative “arb. units” should be adopted, and explicit editorial guidelines should be implemented.
Purpose Universities increasingly issue institutional artificial intelligence (AI) guidance, but it remains unclear whether university-linked journals translate such guidance into publicly visible and usable editorial instructions. This study audited generative AI governance in Asian university-linked journals indexed in Scopus and traceable in SCImago and examined how core policy requirements co-occurred. Methods This cross-sectional policy audit examined 75 unique university-linked Asian journals. Two trained coders independently coded the original 80-record journal-field frame. Across 11 paired pre-adjudication variables, reliability was substantial to almost perfect: Cohen κ=0.747–0.945, quadratic weighted κ=0.944 for the maturity index using 77 valid 0–4 pairs, and Krippendorff α=0.748–0.966. Journal-level analyses included latent class analysis and Firth penalized logistic regression. Results Forty-two journals (56.0%) permitted AI use with disclosure and author accountability, 15 (20.0%) had no explicit AI statement, and 1 (1.3%) had a dedicated AI policy page. Disclosure and accountability signals were common (73%–76%), whereas the six operational requirements were less visible (24%–32%). The count of operational elements per journal was strongly bimodal, and latent class analysis separated journals into operationally nonarticulated (n=49) and operationally articulated (n=26) configurations. Firth regression identified positive associations with articulated configuration membership for health sciences and West Asia, although the estimates were imprecise; the confidence intervals for the other field, region, and ranking estimates included the null. Conclusion The journals showed polarized public articulation of AI policy rather than a gradual continuum. Policy visibility did not necessarily translate into policy usability. The configurations should be interpreted as exploratory patterns, not as a maturity taxonomy.
Purpose Medical education research has expanded rapidly, yet no study has comprehensively mapped its intellectual structure across publications with available reference data. We conducted a reference co-citation analysis to identify the foundational works, intellectual communities, and structural trends within this subset of publications. Methods We analyzed 174,729 medical education publications with reference data, representing 28.1% of 622,407 records from four bibliographic databases for 2000–2025. After deduplicating reference identifiers, we constructed a co-citation network of 7,203 frequently cited references, applied Louvain community detection, tracked the citation trajectories of 15 key theoretical frameworks, and examined structural evolution across four time periods using size-invariant network metrics. Results The co-citation network was organized into 13 major intellectual communities, the largest of which were Learning Theory and Qualitative Methods, Assessment and Competency-Based Education, and Simulation and Patient Safety. Theoretical frameworks exhibited three trajectory patterns: sustained growth, peak-then-decline, and rapid policy-contemporaneous rise. Cross-community connectivity declined over 25 years, as reflected by a decrease in intercommunity edge fraction from 25% to 17% and an increase in modularity from 0.71 to 0.78; these trends were confirmed in uniform downsampling and multiseed analyses. Conclusion Among medical education publications with available reference data, the knowledge structure was organized into distinct intellectual communities that showed progressive specialization. These findings provide an empirical foundation for promoting integrative scholarship and informing curriculum design for medical education scholars.
Purpose After rejection, resubmission often requires journal-specific reformatting. We compared systematic review submission policies across high-impact ophthalmology journals and quantified policy similarity to support data-informed resubmission planning. Methods We conducted a cross-sectional policy analysis of the top 50 ophthalmology journals by 2024 SCImago Journal Rank that publish systematic reviews and are not invite-only. In January 2026, policies were extracted from public author instructions using an a priori data dictionary. Pairwise similarity was calculated on a scale from 0 to 1 using the Gower coefficient across mixed policy variables. Results Policies were heterogeneous and often unstated. Main-text word limits were stated by 29 of 50 journals (58%); among journals with numeric limits (n=23), the median was 4,000 words (interquartile range, 3,500–5,500). Compliance with the PRISMA guidelines was explicitly required by 35 journals (70%), and prospective registration was explicitly required by 6 journals (12%). Across 1,225 journal pairs, the median similarity was 0.64 (interquartile range, 0.57–0.71; range, 0.05–0.98). Similarity among the top five highest-ranking journals ranged from 0.62 to 0.90 (median, 0.75). Conclusion Systematic review submission policies vary widely across high-impact ophthalmology journals, and most journal pairs show only modest similarity. Similarity-based guidance may help authors identify policy-aligned resubmission targets and anticipate potential reformatting requirements.
Purpose This study aimed to define key research ethics and publication policy areas and assess policy disclosure among journals hosted on the Korea Institute of Science and Technology Information (KISTI) academic publishing platform. Methods We conducted a cross-sectional content analysis of journal policy documents. Fifteen policy areas were identified based on international guidelines, domestic reports, prior studies, and publisher policies related to artificial intelligence (AI) and data sharing. Policies from 181 journals were collected on July 10, 2025, and each journal was coded according to whether each policy area was disclosed. Descriptive statistics were used to summarize disclosure overall and according to subject field, access model, and indexing status. Results Journals disclosed a mean of 8.29 policy areas out of 15. Five journals disclosed none of the areas, whereas three disclosed all 15. The most frequently disclosed areas were copyright (158 of 181, 87.3%) and duplicate publication (157 of 181, 86.7%), followed by plagiarism, authorship, and sanctions for misconduct. Emerging governance policies were uncommon: only 5 journals disclosed an AI policy, and 29 disclosed a data sharing policy. When present, AI policies addressed non-authorship, disclosure requirements, scope limitations, accountability, and restrictions on reviewer use. Disclosure rates varied across disciplines, with medicine and pharmacy journals showing the highest rates overall. Open access journals generally disclosed more policy areas than subscription journals, whereas indexing status showed no consistent pattern. Conclusion Policy disclosure among journals hosted on the KISTI academic publishing platform remains incomplete and uneven, particularly with respect to AI and data sharing. To address these gaps, journals should adopt standardized policy templates and receive targeted platform-supported guidance to strengthen research ethics and publication governance in Korea.
