Key Points A systematic audit of six leading journals identifies 301 documented submission requirements, none of which directly evaluates the scientific validity of a manuscript. Thirty‐six of these requirements carry conflicting mandatory values across journals, so full compliance with one journal can create measurable non‐compliance with another. This non‐scientific compliance burden is estimated to consume 24 million hours of global scientific labour annually, falling disproportionately on independent and unaffiliated researchers. Three targeted interventions—a submission transparency index, portable peer review and format‐neutral initial submission—could reduce this burden without altering editorial standards or peer review.
ABSTRACT Master's theses represent a substantial yet often underutilised form of scholarly output, containing original empirical, theoretical and methodological contributions. Despite their scientific value, access to master's theses has historically been limited. A 2019 pilot study of six universities in the SEA‐EU Alliance assessed the openness of master's theses within institutional repositories. We conducted a five‐year follow‐up analysis including three newly joined universities, resulting in a sample of nine universities. Institutions were categorised based on legal and institutional policies regulating repository deposition: (1) mandatory public deposit of all theses, (2) mandatory deposit of selected theses, (3) voluntary public deposit and (4) no regulations. Master's thesis availability in 2023 was compared with 2019 baseline data. Across most universities, no substantial improvement in open access availability was detected after 5 years. In several cases, accessibility declined. The overall reduction was largely attributable to decreased openness in Engineering and technology and in Medical and health sciences. Despite increased attention to open science, accessibility of master's theses remains stagnant. We recommend clearer student‐centred copyright policies, guidance on ethical dissemination, improved repository functionality, incentives for supervisors and departments, and enhanced methodological and scientific‐writing support to foster a culture of openness.
The Harbingers study of early career researchers (ECRs), their work life and scholarly communications, began by studying generational-Millennial-change (H1 c.2016), then pandemic change (H2 c2020) and is now investigating yet another: artificial intelligence (H3 2024-). We report here on a scoping pilot study, which looks at the impact of AI on peer review from over 150 international ECRs from all disciplines. The data were collected by open-ended, in-depth interviews, but being a convenience sample the findings need to be treated with caution. We find that ECRs are displaying a complex, shifting attitude towards the integration of AI into peer review, characterised by cautious optimism on the one hand and scepticism on the other. ECRs are already reporting specific instances of AI increasingly influencing reviewing practices: both potentially beneficial applications and actual misuse. AI is seen as a new systemic threat, potentially worsening the alleged exploitative nature of publishing if adopted without ethical guidelines.
ABSTRACT Scriptural Arabic relies on highly intentional word choices, employing apparent synonyms and near‐synonyms that convey distinct semantic values based on their specific textual placement. Historically, computational translation has struggled to reproduce these precise textual boundaries. Addressing this issue, the present investigation assesses how effectively current artificial intelligence renders these cognitive and near‐synonymous pairs into English. Anchored in an established semantic model of lexical relations, this study operationalizes its evaluation through specific diagnostic parameters: connotative divergence, contextual licensing, collocational restrictions and semantic prosody. A curated sample of verses was assembled to evaluate both Neural Machine Translation (NMT) engines and Large Language Models (LLMs), applying strict prompt constraints and overlap checks to mitigate the retrieval of memorized human translations. The selection focuses on six established pairs of terms: khawf and khashyah (fear/awe), sabil and ṭarīq (metaphorical way/physical path), jasad and jism (lifeless body/living form), ḍiyāʾ and nūr (radiance/light), maṭar and ghayth (punitive rain/merciful rain), alongside khalaqa and jaʿala (creation ex nihilo/appointment). Machine‐generated texts were subsequently compared with recognized scholarly interpretations and authoritative exegetical works (tafsīr). The findings demonstrate that modern algorithms neutralize these lexical variations. Instead of providing context‐sensitive equivalents, the programs default to statistically frequent English terms, thereby producing semantically impoverished texts and compromising semantic accuracy. By detailing the specific types of connotative data that escape computational recognition, this paper clarifies the present limitations of applying automated processing to religious texts. This analysis consequently provides a definitive foundation for evaluating future programming iterations in the field of automated scriptural rendering.
Summary Researchers in Bosnia and Herzegovina face systemic barriers to publishing, including scarce funding and weak institutional support. Limited mentorship and low research literacy increase vulnerability to predatory journals and conferences. Fragmented academic and healthcare systems hinder collaboration and international visibility. Financial constraints and APC costs divert high‐quality local research away from mainstream journals. Practical reforms should focus on research support offices, regional collaborations, and improved promotion criteria.
