Revizyon Paketi — Örnek Çıktı

Power Asymmetries in AI-Assisted Academic Writing and Publishing: A Critical Conceptual Framework

[Author Name] (anonymised for double-blind review)

Author details, ORCID and affiliation are provided on a separate title page.

Abstract

Purpose – This study develops an integrative conceptual framework that maps the power asymmetries produced by the rapid incorporation of generative artificial intelligence (GenAI) into academic writing and publishing. The existing literature frames GenAI either as a leveller that lowers the language barrier or as a threat to research integrity. An integrated analysis that connects these two narratives through the redistribution of power is still missing.

Design/methodology/approach – The study adopts a critical interpretive conceptual review with a theory-synthesis design (Jaakkola, 2020; Snyder, 2019). It thematically synthesises theory on power and knowledge (Foucault, Fricker, coloniality and Southern theory), recent empirical work on GenAI in publishing, and international policy documents (ICMJE, COPE, WAME, UNESCO).

Findings – The asymmetries cluster along six axes: geopolitical and infrastructural, linguistic, economic and access-related, epistemic, governance and surveillance, and labour and accountability. GenAI works at the same time as a leveller and as a concentrator of power. Its net effect therefore depends on governance choices rather than on the technology itself.

Originality/value – By consolidating fragmented debates into a single power-asymmetry lens, the study proposesWhereas recent work documents these asymmetries largely in isolation, this study consolidates them into a single six-axis framework and proposes an equity-centred policy agenda. It draws particular implications for scholars at the periphery, including those in the Global South and those who are not native speakers of English.

Keywords: generative artificial intelligence; scholarly publishing; power asymmetry; epistemic justice; linguistic inequality; publication ethics.

1. Introduction

Since ChatGPT became widely available in November 2022, generative artificial intelligence (GenAI) has entered almost every stage of academic knowledge production at remarkable speed. Large language models (LLMs) are now used across a wide range of tasks, from literature searching and drafting to language editing and peer review (Stokel-Walker and Van Noorden, 2023; van Dis et al., 2023). This shift is described asMany observers describe this shift as a lasting structural change, and it is reshaping the institutional norms of scholarly communication (Dwivedi et al., 2023; Lund et al., 2023).

The current literature addresses this transformation largely through a two-sided discourse. On one side, GenAI is presented as a leveller that lowers the language barrier, especially for researchers who are not native speakers of English (Belcher, 2024; Riazi, 2025). On the other side, scholars stress ethical threats such as fabricated references, data bias, and problems of authorship and accountability (Thorp, 2023; Nature, 2023). However, an integrated analysis that connects these two discourses through the question of how power is redistributed among researchers, institutions, publishers, and technology providers is largely missing.

This study addresses that gap. It asks a central question: along which axes does AI-assisted academic writing and publishing transform existing power asymmetries, and along which axes does it reproduce them? Three sub-questions follow from this central question. First, along which analytical axes do the asymmetries cluster? Second, how can the tension between the levelling promise of GenAI and its concentrating effect be resolved? Third, which principles can guide an equity-centred form of governance?

The contribution of the study is threefold. First, it consolidates scattered empirical and normative debates into a single power-asymmetry framework. Recent scholarship tends to document these asymmetries one at a time, for example the publishing oligopoly's turn to AI (Pooley, 2024), linguistic diversity in AI-assisted writing (Ugwuanyi et al., 2025), and the broader trends in GenAI-assisted academic writing mapped by a recent review in this journal (Wu, 2025). The novelty here is to integrate these separate strands into one structure rather than to treat them as discrete concerns. Second, it transfers established theory on the relationship between power and knowledge (Foucault, 1980; Fricker, 2007) to the GenAI context and thereby builds an analytical bridge. Third, it derives equity-centred policy implications, particularly for researchers at the periphery. The remainder of the paper proceeds as follows. Section 2 presents the theoretical framework and Section 3 the method. Section 4 discusses the findings across six axes of asymmetry. Section 5 develops policy proposals, Section 6 the limitations, and Section 7 concludes.

2. Theoretical Framework

2.1. Power, Knowledge, and the Concept of Asymmetry

This study draws on a Foucauldian (1980) understanding of power and knowledge, which conceptualises power not only as repressive constraint but also as a productive relation that determines what counts as legitimate knowledge and who is authorised to speak. Within this frame, power asymmetry is defined as the unequal distribution, among parties, of control over the infrastructure, language, resources, rules, and oversight mechanisms of academic knowledge production. In the GenAI context, this control tends to concentrate in favour of the technology companies that build the models, the commercial actors that dominate the publishing market, and the norm-setting editorial bodies.

2.2. Epistemic Injustice

Fricker’s (2007) theory of epistemic injustice provides the second analytical anchor. Testimonial injustice refers to a situation in which a knower’s credibility is deflated because of prejudice about their identity. Hermeneutical injustice refers to a situation in which certain groups lack the conceptual resources needed to express their experiences. Because GenAI systems are trained mostly on English-language and Western-centred data, they may treat the expressive forms of peripheral knowers as low-probability, and they may therefore reproduce both testimonial and hermeneutical injustice at scale (Barnawi and R’boul, 2023).

2.3. Centre–Periphery Relations and the Coloniality of Knowledge

The third anchor consists of critical approaches that read the global division of knowledge production along a centre–periphery axis. Southern theory (Connell, 2007), the coloniality of knowledge (Mignolo, 2009), and epistemologies that resist epistemicide (de Sousa Santos, 2014) argue that theoretical authority is located systematically in the North and that peripheral knowledge systems are devalued. The geopolitical character of academic writing (Canagarajah, 2002) gains a new layer with GenAI, and calls for a decolonial artificial intelligence (Mohamed et al., 2020) require a critical reading of this layer. Together, these three anchors ground the six-axis power-asymmetry framework that the following analysis develops.

3. Method

This study is a conceptual review. It does not claim to be a systematic review of the PRISMA type. Instead, it adopts a critical interpretive approach that thematically synthesises theory and empirical findings. In the terms proposed by Jaakkola (2020), it is a conceptual article that follows a theory-synthesis approach, and it applies the narrative review guidelines set out by Snyder (2019). The source pool was built in the Web of Science, Scopus, and Google Scholar databases, using combinations of the keywords “generative AI”, “ChatGPT/large language models”, “academic and scholarly writing”, “publishing”, “power, inequality, and justice”, and “Global South and linguistic”. The time window covers mainly the period from 2022 to 2026. It also includes the foundational theoretical works that ground the framework (for example, Foucault, 1980; Fricker, 2007), key empirical studies in the field (for example, Larivière et al., 2015), and international policy documents (ICMJE, 2023; COPE, 2023; UNESCO, 2021).

The analysis has two stages. In the first stage, the sources were coded according to the type of asymmetry they revealed. In the second stage, these codes were clustered into six analytical axes. The axes were derived inductively from recurring themes in the corpus and consolidated through constant comparison, rather than imposed in advance. Because the method is interpretive, the selection and synthesis process is not independent of the author’s position. This limitation is discussed in Section 6.

4. Findings and Discussion: Six Axes of Power Asymmetry

4.1. Geopolitical and Infrastructural Asymmetry

The first and most fundamental asymmetry is the concentration of the material infrastructure that makes GenAI possible. The development of foundation models requires large datasets, high computing power, and capital. For this reason, it remains under the control of a small number of companies and countries (Bommasani et al., 2021; Crawford, 2021). The environmental and financial cost of this concentration is also distributed unequally, because the training of large models is associated with high energy consumption (Strubell et al., 2019; Bender et al., 2021). As a result, peripheral researchers occupy the position of users of an infrastructure they cannot control, and of sources of its training data.

This infrastructural concentration overlaps with the existing oligopolistic structure of the publishing market. It has long been documented that five large commercial publishers control more than half of scientific output and operate with high profit margins (Larivière et al., 2015). With the transition to open access, these actors have sustained their revenue models through article processing charges (APCs) (Butler et al., 2023). Building on this, Pooley (2024) argues that the same publishers increasingly treat scholars' work and behavioural data as proprietary training assets, licensing them to technology firms while building their own AI products. The integration of GenAI tools into the production and oversight chain by the same publishers creates the risk that power consolidates further in actors that already occupy the centre.

4.2. Linguistic Asymmetry

The position of English as the common language of science imposes a multidimensional cost on researchers who are not native speakers. Amano et al. (2023) show empirically that these researchers spend markedly more time and effort on activities such as reading, writing, publishing, and attending conferences. Similar disadvantages have been identified in single-country cases (Ramírez-Castañeda, 2020). This inequality is consistent with the established literature on the structural disadvantage of the peripheral scholar (Flowerdew, 2019), on the publishing difficulties of developing countries (Salager-Meyer, 2008), and on the global politics of publishing in English (Lillis and Curry, 2010; Canagarajah, 2002).

At precisely this point, GenAI is presented as a levelling hope. Through its language-correction and fluency functions, it is argued to flatten the publishing playing field (Belcher, 2024; Riazi, 2025). However, this promise must be met with a two-part critique. First, the debate between the thesis that linguistic injustice is a myth (Hyland, 2016) and the response to it (Politzer-Ahles et al., 2016) shows that the problem is too structural to be solved by surface-level language correction. Second, the tendency of the models to homogenise toward standard or Western English risks suppressing World Englishes and peripheral rhetorical forms, and it therefore risks flattening voice (Riazi, 2025; Ugwuanyi et al., 2025; Higgins, 2024; Barnawi and R’boul, 2023). In a multi-author dialogue on World Englishes, Ugwuanyi et al. (2025) reach a parallel conclusion: GenAI can democratise the writing process, yet it tends to marginalise minoritised varieties and flatten nuance unless its design and governance are made inclusive. As a result, the language barrier is partly eased, while a new dependence and a standardisation of expression emerge.

4.3. Economic and Access Asymmetry

The third axis concerns the cost of access. A second layer is added in the GenAI era to the APC-based inequalities created by open access (Suber, 2012; Tennant et al., 2016): the most capable models, the institutional licences, and the fast computing resources are mostly paid services. Peripheral researchers therefore face a double paywall, in which they must pay an APC in order to publish and a premium tool fee in order to produce competitive text. Wright (2025) names this emerging stratification an AI-enabled scholarship divide, in which differential access to capable tools reshapes who can realistically compete for publication. This situation is consistent with broader findings on how digital inequality reproduces disadvantage for marginalised groups (Eubanks, 2018), and with international calls for equity in the ethics of artificial intelligence (UNESCO, 2021).

4.4. Epistemic Asymmetry

The fourth axis concerns knowledge itself. LLMs can reproduce and reinforce the biases in their training data (Bender et al., 2021; Noble, 2018; Birhane, 2021), because these systems generate probabilistic patterns rather than grasp meaning (Floridi and Chiriatti, 2020). One of the most concrete manifestations of this is the fabrication of non-existent sources, known as hallucination, and the production of authoritative-looking but inaccurate text (Thorp, 2023). It has been shown that fabricated abstracts can mislead even experts (Else, 2023; Gao et al., 2023). At a deeper level, the homogenisation of argument and style risks reducing the visibility of peripheral epistemologies, and it therefore risks deepening epistemic injustice at scale (Mohamed et al., 2020; de Sousa Santos, 2014; Weidinger et al., 2021). A recent systematic review in this journal reaches a convergent conclusion, identifying authorship ambiguity and the quality of scientific communication among the central ethical concerns of GenAI-assisted writing (Wu, 2025).

4.5. Governance and Surveillance Asymmetry

The fifth axis concerns who sets the rules and at whom oversight is directed. The global norms for GenAI use have been shaped largely by North-centred editorial bodies. The principle that artificial intelligence cannot be an author and that its use must be disclosed transparently was set out by the ICMJE (2023), COPE (2023), and WAME (Zielinski et al., 2023), and leading journals have adopted this line (Thorp, 2023; Nature, 2023). The necessity of these norms is not in dispute. However, the concentration of norm-setting authority at the centre reinforces the rule-taking position of peripheral actors.

The asymmetry of surveillance is sharper. It has been shown that GenAI detectors classify the texts of non-native English authors as AI-generated at a high false-positive rate (Liang et al., 2023). Lower lexical variety and more predictable language use increase the risk that these authors are falsely accused, and punitive oversight is therefore directed disproportionately at the periphery. If this bias is not taken into account, the design of disclosure regimes (Hosseini et al., 2023) can deepen inequality.

4.6. Labour, Accountability, and Acceleration

The sixth axis concerns the distribution of labour and responsibility. The ghostwriter effect that appears in AI-assisted writing blurs responsibility for the text. Because artificial intelligence cannot meet the criteria for accountability, final responsibility remains with the human author (Lund et al., 2023; Dwivedi et al., 2023; Resnik and Hosseini, 2025). At the same time, the increase in production capacity can accelerate the publish-or-perish pressure and the long-observed exponential growth of scientific output (Bornmann and Mutz, 2015). The fall in production cost can also lower the barriers to integrity threats such as fake-paper factories and predatory publishing (Else and Van Noorden, 2021; Grudniewicz et al., 2019; Conroy, 2023). The question of which actors bear this burden, namely detection, correction, and reputational risk, is itself a question of asymmetry.

5. Toward an Equity-Centred Framework: Policy Proposals

The analysis above shows that the effect of GenAI depends more on governance choices than on the technology itself. Five principles are proposed for an equity-centred agenda. First, inclusive co-design and critical AI literacy: peripheral stakeholders should be included in the development of tools and policies (Riazi, 2025; Ugwuanyi et al., 2025; Higgins, 2024). Second, proportionate and non-punitive disclosure regimes; and because detector bias has been demonstrated (Liang et al., 2023), accusations based on detectors alone should be avoided. Third, subsidised or open GenAI access for researchers in low- and middle-income countries, together with support for diamond open-access models (Suber, 2012; Tennant et al., 2016; UNESCO, 2021). Fourth, investment in multilingual models and the legitimation of World Englishes. Fifth, the pluralisation of governance, which means stronger representation of the South on norm-setting bodies and the use of reflexive accountability frameworks (Hosseini et al., 2023; Resnik and Hosseini, 2025).

6. Limitations

The limitations of the study should be stated clearly. First, because of its conceptual and interpretive character, the findings have not been tested empirically. The proposed six-axis framework is open to validation through quantitative and qualitative studies. Second, and in tension with the subject of the study, the source pool consists mainly of English-language literature. This situation is a reflection of the very linguistic asymmetry under examination, and future studies should address it with multilingual sources. Third, because the field evolves rapidly, model capabilities and policy documents may quickly become outdated.

7. Conclusion

Generative artificial intelligence is at the same time a levelling and a concentrating force in academic writing and publishing. By partly lowering the language barrier, it can open breathing space for peripheral researchers. However, it also tends to consolidate control over infrastructure, capital, norm-setting, and oversight at the centre. The central argument of this study is that the net outcome is determined not by the nature of the technology but by the governance choices that surround it. Without equity-centred design and governance, GenAI risks recoding old asymmetries within a new infrastructure. By naming six interacting axes rather than a single divide, the framework is offered as a practical lens for both future empirical research and equity-centred policy design.For this reason, the proposed six-axis framework aims to offer an agenda both for future empirical research and for equity-centred policy design.

Declarations

AI use statement. Generative artificial intelligence tools were used in the preparation of this manuscript for literature mapping and language editing. All AI outputs were reviewed, verified, and corrected by the author(s). The author(s) hold final responsibility for the accuracy, integrity, and originality of the entire content. In line with the principles of the ICMJE (2023) and COPE (2023), AI tools are not listed as authors.

Conflict of interest. The author(s) declare no conflict of interest.

Funding. No external funding was received for this study.

Data availability. The study is conceptual; no dataset was generated or analysed.

References

Amano, T., Ramírez-Castañeda, V., Berdejo-Espinola, V., Borokini, I., Chowdhury, S., Golivets, M., González-Trujillo, J. D., Montaño-Centellas, F., Paudel, K., White, R. L., & Veríssimo, D. (2023). The manifold costs of being a non-native English speaker in science. PLOS Biology, 21(7), e3002184. https://doi.org/10.1371/journal.pbio.3002184

Barnawi, O. Z., & R’boul, H. (2023). Colonizability and hermeneutical injustice in applied linguistics research: (Im)possibilities for epistemological delinking. Applied Linguistics, 44(5), 865–881. https://doi.org/10.1093/applin/amad047

Belcher, D. (2024). The promising and problematic potential of generative AI as a leveler of the publishing playing field. Journal of English for Research Publication Purposes, 5(1–2), 93–105.

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21) (pp. 610–623). Association for Computing Machinery. https://doi.org/10.1145/3442188.3445922

Birhane, A. (2021). Algorithmic injustice: A relational ethics approach. Patterns, 2(2), 100205. https://doi.org/10.1016/j.patter.2021.100205

Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., … Liang, P. (2021). On the opportunities and risks of foundation models. arXiv. https://doi.org/10.48550/arXiv.2108.07258

Bornmann, L., & Mutz, R. (2015). Growth rates of modern science: A bibliometric analysis based on the number of publications and cited references. Journal of the Association for Information Science and Technology, 66(11), 2215–2222. https://doi.org/10.1002/asi.23329

Butler, L.-A., Matthias, L., Simard, M.-A., Mongeon, P., & Haustein, S. (2023). The oligopoly’s shift to open access: How the big five academic publishers profit from article processing charges. Quantitative Science Studies, 4(4), 778–799. https://doi.org/10.1162/qss_a_00272

Canagarajah, A. S. (2002). A geopolitics of academic writing. University of Pittsburgh Press.

Committee on Publication Ethics (COPE). (2023). Authorship and AI tools: COPE position statement. https://publicationethics.org/cope-position-statements/ai-author

Connell, R. (2007). Southern theory: The global dynamics of knowledge in social science. Polity Press.

Conroy, G. (2023). Scientists used ChatGPT to generate an entire paper from scratch — but is it any good? Nature, 619(7970), 443–444. https://doi.org/10.1038/d41586-023-02218-z

de Sousa Santos, B. (2014). Epistemologies of the South: Justice against epistemicide. Paradigm Publishers.

Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., … Wright, R. (2023). Opinion paper: “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642

Else, H. (2023). Abstracts written by ChatGPT fool scientists. Nature, 613(7944), 423. https://doi.org/10.1038/d41586-023-00056-7

Else, H., & Van Noorden, R. (2021). The fight against fake-paper factories that churn out sham science. Nature, 591(7851), 516–519. https://doi.org/10.1038/d41586-021-00733-5

Eubanks, V. (2018). Automating inequality: How high-tech tools profile, police, and punish the poor. St. Martin’s Press.

Floridi, L., & Chiriatti, M. (2020). GPT-3: Its nature, scope, limits, and consequences. Minds and Machines, 30(4), 681–694. https://doi.org/10.1007/s11023-020-09548-1

Flowerdew, J. (2019). The linguistic disadvantage of scholars who write in English as an additional language: Myth or reality. Language Teaching, 52(2), 249–260. https://doi.org/10.1017/S0261444819000041

Foucault, M. (1980). Power/knowledge: Selected interviews and other writings 1972–1977 (C. Gordon, Ed.). Pantheon Books.

Fricker, M. (2007). Epistemic injustice: Power and the ethics of knowing. Oxford University Press.

Gao, C. A., Howard, F. M., Markov, N. S., Dyer, E. C., Ramesh, S., Luo, Y., & Pearson, A. T. (2023). Comparing scientific abstracts generated by ChatGPT to real abstracts with detectors and blinded human reviewers. npj Digital Medicine, 6, 75. https://doi.org/10.1038/s41746-023-00819-6

Grudniewicz, A., Moher, D., Cobey, K. D., Bryson, G. L., Cukier, S., Allen, K., … Lalu, M. M. (2019). Predatory journals: No definition, no defence. Nature, 576(7786), 210–212. https://doi.org/10.1038/d41586-019-03759-y

Higgins, C. (2024). Toward un-WEIRDing academic publishing about language. International Journal of the Sociology of Language, 2024(289–290), 27–31.

Hosseini, M., Resnik, D. B., & Holmes, K. (2023). The ethics of disclosing the use of artificial intelligence tools in writing scholarly manuscripts. Research Ethics, 19(4), 449–465. https://doi.org/10.1177/17470161231180449

Hyland, K. (2016). Academic publishing and the myth of linguistic injustice. Journal of Second Language Writing, 31, 58–69. https://doi.org/10.1016/j.jslw.2016.01.005

International Committee of Medical Journal Editors (ICMJE). (2023). Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals: Use of artificial intelligence in publishing. https://www.icmje.org/recommendations/

Larivière, V., Haustein, S., & Mongeon, P. (2015). The oligopoly of academic publishers in the digital era. PLOS ONE, 10(6), e0127502. https://doi.org/10.1371/journal.pone.0127502

Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. https://doi.org/10.1016/j.patter.2023.100779

Lillis, T., & Curry, M. J. (2010). Academic writing in a global context: The politics and practices of publishing in English. Routledge.

Lund, B. D., Wang, T., Mannuru, N. R., Nie, B., Shimray, S., & Wang, Z. (2023). ChatGPT and a new academic reality: Artificial intelligence-written research papers and the ethics of the large language models in scholarly publishing. Journal of the Association for Information Science and Technology, 74(5), 570–581. https://doi.org/10.1002/asi.24750

Mignolo, W. D. (2009). Epistemic disobedience, independent thought and decolonial freedom. Theory, Culture & Society, 26(7–8), 159–181. https://doi.org/10.1177/0263276409349275

Mohamed, S., Png, M.-T., & Isaac, W. (2020). Decolonial AI: Decolonial theory as sociotechnical foresight in artificial intelligence. Philosophy & Technology, 33(4), 659–684. https://doi.org/10.1007/s13347-020-00405-8

Nature. (2023). Tools such as ChatGPT threaten transparent science; here are our ground rules for their use [Editorial]. Nature, 613(7945), 612. https://doi.org/10.1038/d41586-023-00191-1

Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. New York University Press.

Politzer-Ahles, S., Holliday, J. J., Girolamo, T., Spychalska, M., & Berkson, K. H. (2016). Is linguistic injustice a myth? A response to Hyland (2016). Journal of Second Language Writing, 34, 3–8. https://doi.org/10.1016/j.jslw.2016.09.003

Ramírez-Castañeda, V. (2020). Disadvantages in preparing and publishing scientific papers caused by the dominance of the English language in science: The case of Colombian researchers in biological sciences. PLOS ONE, 15(9), e0238372. https://doi.org/10.1371/journal.pone.0238372

Resnik, D. B., & Hosseini, M. (2025). The ethics of using artificial intelligence in scientific research: New guidance needed for a new tool. AI and Ethics, 5(2), 1499–1521. https://doi.org/10.1007/s43681-024-00493-8

Riazi, A. M. (Ed.). (2025). Generative artificial intelligence (GenAI) and writing for scholarly publication [Special issue]. Journal of English for Research Publication Purposes, 6(2).

Salager-Meyer, F. (2008). Scientific publishing in developing countries: Challenges for the future. Journal of English for Academic Purposes, 7(2), 121–132. https://doi.org/10.1016/j.jeap.2008.03.009

Stokel-Walker, C. (2023). ChatGPT listed as author on research papers: Many scientists disapprove. Nature, 613(7945), 620–621. https://doi.org/10.1038/d41586-023-00107-z

Stokel-Walker, C., & Van Noorden, R. (2023). What ChatGPT and generative AI mean for science. Nature, 614(7947), 214–216. https://doi.org/10.1038/d41586-023-00340-6

Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (pp. 3645–3650). https://doi.org/10.18653/v1/P19-1355

Suber, P. (2012). Open access. MIT Press.

Tennant, J. P., Waldner, F., Jacques, D. C., Masuzzo, P., Collister, L. B., & Hartgerink, C. H. J. (2016). The academic, economic and societal impacts of open access: An evidence-based review. F1000Research, 5, 632. https://doi.org/10.12688/f1000research.8460.3

Thorp, H. H. (2023). ChatGPT is fun, but not an author. Science, 379(6630), 313. https://doi.org/10.1126/science.adg7879

UNESCO. (2021). Recommendation on the ethics of artificial intelligence. United Nations Educational, Scientific and Cultural Organization. https://unesdoc.unesco.org/ark:/48223/pf0000381137

van Dis, E. A. M., Bollen, J., Zuidema, W., van Rooij, R., & Bockting, C. L. (2023). ChatGPT: Five priorities for research. Nature, 614(7947), 224–226. https://doi.org/10.1038/d41586-023-00288-7

Weidinger, L., Mellor, J., Rauh, M., Griffin, C., Uesato, J., Huang, P.-S., … Gabriel, I. (2021). Ethical and social risks of harm from language models. arXiv. https://doi.org/10.48550/arXiv.2112.04359

Zielinski, C., Winker, M. A., Aggarwal, R., Ferris, L. E., Heinemann, M., Lapeña, J. F., … Citrome, L. (2023). Chatbots, generative AI, and scholarly manuscripts: WAME recommendations on chatbots and generative artificial intelligence in relation to scholarly publications. World Association of Medical Editors. https://wame.org/page3.php?id=106