This study examines whether declared user-agent labels are associated with differences in access to popular domain home pages and in representations after access succeeds. A paired cross-sectional audit requested the HTTPS home pages of 20,000 Tranco-ranked domains from one fixed research host using Chrome, Googlebot, OAI-SearchBot, GPTBot and ChatGPT-User labels. The Chrome label received HTTP 2xx responses from 60.10 percent of domains, compared with 54.97 percent for Googlebot and between 49.34 and 51.34 percent for the OpenAI labels. Most dual-success pairs had matching extracted-text hashes, although between 4.08 and 6.92 percent of pairs with nonempty sampled text fell below 0.80 Jaccard similarity. Googlebot combined the broadest access among crawler labels with the highest divergence on several representation measures, whereas OAI-SearchBot had the lowest divergence in the conditional samples. Shorter text windows and a common successful-access sample preserved this contrast. The findings describe differentiated acquisition at prominent home pages, with substantial access differences and representation differences concentrated in a minority of successful pairs.
This study examines how defects in Chinese municipal transportation and education/public-service open data affect AI-mediated answer reliability. The two-stage experiment evaluated 24 datasets, 144 questions, and 4032 trials. Stage 1 tested five theory-informed defect families at ordered perturbation levels with two model configurations; Stage 2 used new datasets and questions with four configurations. A subsequent 1152-trial harmonized validation tested field-name masking, record loss, and unlabeled scale shifts across all 24 datasets with the two model families common to both stages. In Stage 2, all three defects reduced strict-pass performance across all four configurations. In the harmonized validation, all six model-by-defect effects were negative in both data-setting groups; evidence was strongest for field-name masking and record loss, while scale shifts were more heterogeneous. The findings connect published-data properties to answer reliability and demonstrate a controlled approach for studying the robustness of AI-mediated access to public data.
University presses face converging pressures from scholarly quality, peer-review integrity, reviewer scarcity, Open Science, artificial intelligence, international visibility, and economic sustainability. This article proposes Epigenetic Editorial Intelligence, a bioinspired conceptual framework that represents stable institutional rules as Editorial DNA and models operational adaptation through epigenetic regulation, an Artificial Immune System, organizational memory, human–AI collaboration, and multiobjective optimization. Nine operational indicators are formalized and aggregated into a Global Editorial Intelligence Index (GEII). Internal model behavior is examined through paired synthetic ablations that disable one mechanism at a time, using 200 Monte Carlo repetitions, common random-number streams, 95% confidence intervals, alternative weighting schemes and 1000 Dirichlet weight perturbations. Results are interpreted as structural-consistency diagnostics rather than empirical validation. This work provides a reproducible conceptual and mathematical foundation for subsequent implementation and testing with real editorial records.
Vietnam's public library standards bind every level of local government equally, yet provision remains uneven. This article examines that paradox through the separation of standard-setting authority from funding authority, formalised as the binding fiscal bridge (B): whether an instrument attaches to a standard an expenditure obligation on the budget holder, fixed by a rate, ratio or minimum. A provision-level analysis of 39 instruments (2000 to 2026) finds B absent from every classifiable instrument, with none in force barring a zero allocation; whether one occurs is predicted to depend on the implementing unit's direct budget status. Twenty-one interviews are consistent with this configuration. Three plausibility probes do not isolate B as the cause of funding variation. A funding-guarantee framework was removed as standards were tightened, an asymmetry apparently underexamined in policy-dismantling research. Collection development, digital services and commune-level reading are left without verifiably guaranteed recurrent funding. The proposition is tentative.
Generative AI is increasingly used by students as a source of academic information, yet less is known about how they judge, verify, and rely on AI-generated content. Grounded in Source Credibility Theory and informed by information literacy and digital verification perspectives, this study examines associations among perceived credibility, trust, verification behaviour, and reliance. Survey data from 650 undergraduate and postgraduate students across four universities in Nepal were analysed using covariance-based structural equation modelling. Perceived credibility was positively associated with trust, while trust and verification behaviour were positively associated with reliance. The direct association between perceived credibility and reliance was not significant after trust was included, whereas the indirect association through trust was significant. Because the data are cross-sectional, this pattern is interpreted as an indirect association rather than causal mediation. The study positions generative AI as an academic information source requiring credibility judgement, verification, and cautious academic incorporation.
