I discuss why diamond open access is necessary to reduce perverse incentives in science, and to improve the outputs of Gen AI.
An editorial based on this blog post now appears in the Bulletin of Computer and Data Sciences 7(1):1-9. https://doi.org/10.71448/bcds2671-1
Over the last two decades awareness of a Reproducibility Crisis penetrated all scientific disciplines. Studies recently showed that 90% in STEM and 95% in psychology were aware of this Crisis1,2. At its core this is a crisis of trust. The findings scientists publish and tout as facts, are often not replicable by others. Moreover, a great many are not even computationally reproducible using the original data3,4. The Open Science Movement developed partly in response to this Crisis5,6. Especially in the last two decades, researchers organized grassroots movements to make science more transparent, reliable and ethical.
One of the many tenets of the Open Science (OS) Movement is open access. Published research results must be accessible to everyone. This is not a new idea. In the 1940s Robert K. Merton developed normative goals necessary to ensure integrity in science7. One was that all scientific findings should be public property. He argued that effective and efficient progress of science depends on open public access. In data science this Movement led to the concept of Open Science by Design – a strategy where scientists plan in advance how they will make all metadata, data and algorithms available and easily re-usable by anyone8.
The OS Movement has been successful. Open access journals and articles increased rapidly. They outpaced the global increase in publications in other formats in the last two decades9. One problem is that these open access articles are overwhelmingly funded by the authors of the studies via Author Processing Charges (APCs). It costs over 10 thousand US$ to publish a study open access in the journal Nature, and most other journals charge at least two thousand. Except for elite institutes and projects with generous third-party fundings, these fees are not usually covered or coverable by universities. A naïve outsider might ask why scientists would pay astronomical fees to make their own hard work publicly available, when they can simply share it online in a free repository like the Open Science Framework, Github or any number of preprint servers based on the ArXiv model?
The answer is competition. The strongest norm governing the practice of science today is publish-or-perish. In every field and science at large, scientists are judged based on their publication record. Ask any academic what the top journals in their area are, and you will get relatively consistent answers by discipline, sub-field and science in general10. Look at the faculty of any ‘top’ university, and the faculty in any discipline will have one or more publications in these ‘top’ journals. Without publications in these journals, a scientist has no chance at a career in science. This is a collective cultural fact. One deeply institutionalized in the organizations, rules, norms, expectations and behaviors of scientists11,12.
Thus, the founding of new open access journals with lower APCs has little impact on the scientific enterprise because they cannot compete with institutionalized legacy statuses of existing journals. The greatest success story is the non-profit publisher PLOS, which rose swiftly in the rankings, but could not crack into the very top tier. Although far cheaper than Nature, it is still not ‘cheap’, with APCs in their family of journals ranging from 2.5 to 3.2 thousand $US[1].
The most successful open access models are within existing paywalled journals where authors have the option to publish “gold” open access, rather than publish for free. The payment of somewhere between two and 10 thousand US$, gets authors the right to have their single article published open access inside of a closed access journal. This means that despite a massive shift toward open access, the fundamental structures of scientific publishing have not changed.
Ethical Implications
Competition in science supports motivation and innovation, but the publish-or-perish norm is a toxic externality. Scientists willingly prioritize subjective journal ranking, impact factor and increasing their citation counts to get ahead within the competitive scientific enterprise. They do this despite widespread skepticism and evidence that rankings and citations do not correlate strongly with the quality and reliability of published studies14–16. They do this because they must, or at least perceive that they must, in order to follow their scientific career aspirations. The pressure to publish thus motivates rent-seeking behaviors designed to increase publication chances, i.e., career chances. Conducting higher quality science is one of these behaviors. But there are many methods to increase publication chances that are science orthogonal, what are known today as questionable research practices (QRPs)17,18.
Some QRPs are unconscious and learned from supervisors in the process of converting research into a publishable paper. For example, researchers routinely run many statistical models but report only those that show the strongest support of their claims. Researchers who believe that their claim is true in the first place will gravitate toward models that support it, convincing themselves intrinsically that these models are the best tests of their claim.
Decisions based on confirming intrinsic beliefs or window dressing for peer reviewers reduce the replicability of science. They narrow down the multiverse of potential findings into a highly selected set of results. This selectivity is independent of the process that generated the data in the first place, in other words, it is science orthogonal. A classic example is a study published suggesting that hurricanes with feminine names cause more damage than those with masculine names. It turns out that using the available data, the original researchers selected a model that produced regression coefficients that were extremely far away from the central tendency among all other plausible models’ regression coefficients19. We do not know whether this was a conscious decision but their reporting hides the truth and simultaneously increases publication chances.
There are of course conscious and highly unethical behaviors leading to an entirely false representation of reality. Science is filled with scandals of hacking and data-faking. Rent-seeking alone can explain unethical learned unconscious and conscious behaviors of scientists. They seek status and money and job security, or in some cases seek results that support a particular worldview or policy outcome20,21. But rent-seeking is unambiguously the main reason22,23.
If we as a scientific community and science-interested public, want to eliminate the perverse incentive structures that bound and inform rent-seeking behaviors of scientists we need radical change. Status should not be assigned based on journal metrics that often have little to do with the quality of the research being conducted or published and more to do with legacy and embedded norms. The most radical proposal is to eliminate journals altogether. But this is an extremely unlikely outcome no matter how powerful the OS Movement becomes.
Journals became standard in science hundreds of years ago as a means for communicating scientific discoveries across time and space24. They enabled scientists to acquire knowledge without travelling to faraway universities. Publishing was not cheap, and with the dawn of digital media, it became even more expensive as publishers raced to provide their journal both in-print and online. The costs associated with publishing gave publishing firms a great deal of power over time.
The result is that companies like Springer Nature and Elsevier have enough power to dictate to scientists how they perform their research and communicate their results25. They shape academic careers, institutional priorities, governments’ science policies, and perceptions of journals and the publishing enterprise among scientists26. Their primary legal interest is their shareholders. Profit is their priority, and only second is to provide a service to science as their product. A simple mathematical proof confirms that profit is their priority: If they cannot make a profit they will no longer provide the scientific services but if they can make a product without scientific services they have no reason to stop.
Although big publishing firms have shown many draconian practices in their ‘service’ to science27–30, scientists still need a means to communicate their results with each other and the public. This can be done without for-profit publishing but probably not without a journal publication format, or something very similar. For example, we now have a plethora of ‘green’ open access preprint servers where scholars can deposit working papers, or prior versions of their published articles. These are a viable means to communicate science without the perverse incentives generated by big publishing or institutionalized journal rankings. This all still involves the journal article as the standard unit of science production.
Diamond Open Access
If we are ‘stuck’ with journals in science as our primary communication medium, then they should be as free from perversely incentivized bias as much as possible. The first step is thus to remove for-profit publishing from the equation. It is fine to use the services of for-profit publishers. They have shown the capacity to provide print on demand, marketing and scientific journalism. But when they control and direct the scientific enterprise when have an ethical conflict of interest – namely profit versus robust science.
The costs of publishing a journal are the lowest in history. There are publication kits that help associations, institutions and stand-alone journals to take publication into their own hands31. With minimal costs and self-governance, journals have no need to push institutions to purchase journal subscriptions and no need to charge authors astronomical publication fees. Thus, they themselves are not perversely incentivized to perpetually increase their status to make their product profitable.
The optimal existing solution is diamond open access, whereby a journal charges no APCs and is freely readable and downloadable online32. It is the most ethical and equitable by design, and is perceived as the ideal model by most academics33. The OS Movement is overwhelmingly in favor of diamond open access34, as are governance bodies – at least those free from the influence of big publishing like UNESCO35,36. Despite 13 thousand journals indexed in the Directory of Open Access Journals (DOAJ) with no fees as of February 16th, 2026, these journals are not on the radar of most indexing services, and do not belong to the mainstream of journals published by major scientific societies37. This means that successful diamond open access journals are very rare.
Because of such a saturated scientific ‘market’ for publication outlets, starting new diamond open access journals has had little impact on producing a more ethical science. They do not gain reputation. The reason PLOS was so successful, despite failing to break into the highest echelon of science, was because it has a huge cash flow and can use it for branding and promotion. From an ethical science perspective it is valuable, like gold, but it is not as valuable as diamond.
Flipping or Starting Over?
To have the highest ranking and most well-known journals diamond open access, societies need to cancel their contracts with for-profit publishers38. One problem with this is that many academic societies are themselves run like for-profit businesses. They seek to generate as much revenue as possible, and diamond open access would threaten this model. They have embedded relationships with publishers and agree to renew contracts together, mostly independent of the scientists they serve. I witnessed this first hand with the American Sociological Association and Sage39,40. Contracting a for-profit publisher provides a non-profit academic organization a scapegoat for amassing capital via subscription fees.
There are incredible exceptions; however, and they offer model success stories. Computational Linguistics is one of the earliest journals considered to be among the top in its field, to flip to diamond open access. The first step was a move by the Association for Computational Linguistics in 2002 to create the CL Anthology which made all articles published by association journals open access online after an embargo period. Then in 2009 the journal Computational Linguistics flipped to diamond open access. The journal Demography of the Population Association of America is another example.
The embeddedness of big publishing in science, and the capital that associations can raise through their journals when run by big publishers, are major barriers to flipping. Another major barrier is that some big publishers coerce scientific societies into signing away the rights to the titles of their journals in their publishing contracts. This tactic is most intensively deployed by Elsevier. They own the rights to the titles of nearly all the journals they publish. Therefore, when these societies want to change publishers, they cannot. They are trapped. They would have to start a new journal with a different title – what happened for example with the Journal of Infometrics41. There is otherwise no way around this problem because of copyright law.
Therefore, collective efforts to build the popularity of new diamond open access journals would greatly increase the movement toward a more ethical and effective scientific enterprise. If these are journals from societies, which are essentially the same journal but with a new (not copyrighted by Elsevier) name, it requires authors to support this journal and immediately abandon the other. Supporting new diamond open access journals, whether completely new or newly named, requires established scholars to put their status-seeking egos aside in the name of scientific progress and ethics. It is precisely those who have built major scientific reputations who need to engage in this change, because they can give them most clout to new journals by touting them and publishing in them.
Scholars and societies alone cannot carry the burden. Hiring committees need to reward these behaviors. Rather than seeing a publication in a new ‘unranked’ or ‘low ranked’ diamond open access journal as a sign that the article is ‘not high quality enough for top journals’, committees should judge publications only on their content. Then, if two publications are seen as equally scientifically rigorous and high quality, the one in a diamond open access journal should get a greater weight. Scientist who consistently publish high quality research in diamond open access journals should be favorites of hiring committees, all else equal.
A hiring committee would be unlikely to reward an applicant who shows sociopathic behaviors. Following this logic, they should not reward scientists who show behaviors that go against ethical science by practicing closed and profit-incentivized science. I need to be very clear here that I personally am still publishing regularly in journals published by for-profit publishers. I am not that famous, and I do not have tenure. This is a perfect example of why we need sweeping changes at the institutional level, and from the top of the scientific hierarchy.
With efforts to flip both existing journals and efforts to reward new, ethical journals, we give science its greatest future chances for improvement and sustainability. There are many efforts underway and these should serve as guides, for example the Diamond Open Access Fund from the Dutch Research Council and MIT’s shift+OPEN initiative.
Diamond AI
Diamond open access has a second, equally important role for the future of science. It is necessary to inform Generative Artificial Intelligence (Gen AI). The LLMs that power popular Gen AI are trained heavily on corpora assembled from what is available via the Internet. Paywalled literature is missing from training corpora, not because it is unimportant, but because it is not accessible or legally usable42. Gen AI outputs are therefore based on a highly restricted sample of all scientific knowledge. Open access for all of science would ensure that anyone using Gen AI, would get the best possible information based on all that we know as humans. As essentially everyone is using Gen AI today43, this would mean that the public would be optimally informed.
It is not necessary for Gen AI development that all articles are diamond open access, they just need to be somewhere, e.g., green or gold open access. Yet, without diamond open access the entire knowledge enterprise will continue to favor the work of those with greater resources44. Resources are of course necessary to produce higher quality science because of research costs, but when it comes to publishing and dissemination, a resource advantage reproduces the already existing Global North-South disadvantages in science. This limits human capacity to tap resources in lower income societies. These are societies filled with potential contributions to science that could benefit both 1) their own societies’ development – because it enables them to gain status, resources and build stronger, more attractive and sustainable scientific institutions, and 2) all of scientific knowledge because there are brilliant minds waiting to be tapped that might otherwise give up on science because it is for them no sustainable in their region.
If the knowledge in Gen AI remains Global North and WEIRD biased (Western, educated, industrialized, rich and democratic), the result is that these countries remain culturally and economically advantaged beyond that which exists presently. This means that without diamond open access norms, we are willingly allowing Gen AI to increase cultural hegemony and economic domination of a minority. I am not directly arguing whether this is good or bad. I am in the Global North and profit from a stronger Global North science advantage. My argument is that science should be neutral, favoring the most optimal and reliable knowledge, rather than legacies or strategies that game the scientific system.
References
1. Baker, M. 1,500 scientists lift the lid on reproducibility. Nature 533, 452–454 (2016).
2. Metskas, A. How Much Do Academic Psychologists Trust Academic Psychology, and Is There Still a Replication Crisis? Transparent Replications https://replications.clearerthinking.org/how-much-do-academic-psychologists-trust-academic-psychology-and-is-there-still-a-replication-crisis/#survey-demographics (2025).
3. Open Science Collaboration. Estimating the reproducibility of psychological science. Science 349, (2015).
4. Breznau, N. et al. The reliability of replications: a study in computational reproductions. Royal Society Open Science 12, 241038 (2025).
5. Engzell, P. & Rohrer, J. M. Improving Social Science: Lessons from the Open Science Movement. PS: Political Science & Politics 1–4 (2021) doi:10.1017/S1049096520000967.
6. Breznau, N. Legacy of Jon Tennant, “Open science is just good science”. Crowdid https://crowdid.hypotheses.org/548 (2022) doi:10.58079/ne8h.
7. Merton, R. K. The Sociology of Science: Theoretical and Empirical Investigations. (University of Chicago press, 1973).
8. Wittenburg, P. Open Science and Data Science. Data Intelligence 3, 95–105 (2021).
9. NCSES. Publication Output by Region, Country, or Economy and by Scientific Field. https://ncses.nsf.gov/pubs/nsb202333/publication-output-by-region-country-or-economy-and-by-scientific-field#utm_source=chatgpt.com (2023).
10. Serenko, A. & Bontis, N. A critical evaluation of expert survey‐based journal rankings: The role of personal research interests. Asso for Info Science & Tech 69, 749–752 (2018).
11. Mancoridis, M., Sumers, T. & Griffiths, T. Publish or Perish: Simulating the Impact of Publication Policies on Science. Proceedings of the Annual Meeting of the Cognitive Science Society 46, (2024).
12. Breznau, N. Questionable research practices from the practitioners’ perspectives. Crowdid https://crowdid.hypotheses.org/1666 (2025) doi:10.58079/14f3u.
13. Lawrence, S. Free online availability substantially increases a paper’s impact. Nature 411, 521–521 (2001).
14. Fleck, C. The Impact Factor Fetishism. European Journal of Sociology / Archives Européennes de Sociologie 54, 327–356 (2013).
15. Rushforth, A. & De Rijcke, S. Practicing responsible research assessment: Qualitative study of faculty hiring, promotion, and tenure assessments in the United States. Res Eval 33, (2024).
16. Dougherty, M. R. & Horne, Z. Citation counts and journal impact factors do not capture some indicators of research quality in the behavioural and brain sciences. R Soc Open Sci. 9, 220334 (2022).
17. Gopalakrishna, G. et al. Prevalence of questionable research practices, research misconduct and their potential explanatory factors: A survey among academic researchers in The Netherlands. PLOS ONE 17, e0263023 (2022).
18. John, L. K., Loewenstein, G. & Prelec, D. Measuring the Prevalence of Questionable Research Practices With Incentives for Truth Telling. Psychol Sci 23, 524–532 (2012).
19. Muñoz, J. & Young, C. We Ran 9 Billion Regressions: Eliminating False Positives through Computational Model Robustness. Sociological Methodology 48, 1–33 (2018).
20. Borjas, G. J. & Breznau, N. Ideological bias in the production of research findings. Science Advances 12, eadz7173 (2026).
21. Rainero, V., Stolz, J. & Luijkx, R. The Faith Factor. How Scholars’ Religiosity Biases Research Findings on Secularization. Sociological Science 13, 154–177 (2026).
22. Aronson, J. K. When I use a word . . . “Publish or perish”: adverse effects. https://doi.org/10.1136/bmj.r1577 (2025) doi:10.1136/bmj.r1577.
23. Paruzel-Czachura, M., Baran, L. & Spendel, Z. Publish or be ethical? Publishing pressure and scientific misconduct in research. Research Ethics 17, 375–397 (2021).
24. Carey, J. Scientific Communication Before and After Networked Science. Information & Culture 48, 344–367 (2013).
25. Larivière, V., Haustein, S. & Mongeon, P. The Oligopoly of Academic Publishers in the Digital Era. PLOS ONE 10, e0127502 (2015).
26. Rossello, G. & Martinelli, A. The effect of lobbies’ narratives on academics’ perceptions of scientific publishing: A survey experiment. Information Economics and Policy 71, 101148 (2025).
27. Butler, L.-A., Matthias, L., Simard, M.-A., Mongeon, P. & Haustein, S. The oligopoly’s shift to open access: How the big five academic publishers profit from article processing charges. Quantitative Science Studies 4, 778–799 (2023).
28. Else, H. Dutch publishing giant cuts off researchers in Germany and Sweden. Nature 559, 454–455 (2018).
29. Lancet, T. & Board, T. L. I. A. Reed Elsevier and the arms trade. The Lancet 366, 868 (2005).
30. Jureidini, J. & Clothier, R. Elsevier should divest itself of either its medical publishing or pharmaceutical services division. The Lancet 374, 375 (2009).
31. Scholastica. New fully-OA publishing toolkit and stakeholder reflections on 20 years of the BOAI. https://blog.scholasticahq.com/post/oa-publishing-toolkit-and-stakeholder-reflections-BOAI20/ (2022).
32. Normand, S. Is Diamond Open Access the Future of Open Access? The iJournal: Student Journal of the Faculty of Information 3, (2018).
33. Kumari, M. & A, S. Perceptions of open access publishing: A comparative study of gold and diamond models among global researchers. Alexandria 35, 55–73 (2025).
34. Plan S. Working collectively towards an equitable, community-driven and academic-led scholarly publishing model. Action Plan for Diamond Open Access https://www.coalition-s.org/action-plan-for-diamond-open-access/ (2022).
35. UNESCO. Diamond Open Access. Public Service Press Release vol. Online Report (2026).
36. UNESCO. UNESCO Recommendation on Open Science. https://unesdoc.unesco.org/ark:/48223/pf0000379949.locale=en (2021).
37. Simard, M.-A., Basson, I., Hare, M., Larivière, V. & Mongeon, P. The Value of a Diamond: Understanding Global Coverage of Diamond Open Access Journals in Web of Science, Scopus, and OpenAlex to Support an Open Future. Proceedings of the Annual Conference of CAIS / Actes du congrès annuel de l’ACSI https://doi.org/10.29173/cais1845 (2023) doi:10.29173/cais1845.
38. Trueblood, J. S. et al. The misalignment of incentives in academic publishing and implications for journal reform. Proc. Natl. Acad. Sci. U.S.A. 122, e2401231121 (2025).
39. Why I’m leaving the American Sociological Association. Family Inequality https://familyinequality.wordpress.com/2021/11/06/why-im-leaving-the-american-sociological-association/ (2021).
40. Philip Cohen’s ASA Publications Committee platform. Family Inequality https://familyinequality.wordpress.com/2018/01/14/philip-cohens-asa-publications-committee-platform/ (2018).
41. Singh Chawla, D. Open-access row prompts editorial board of Elsevier journal to resign. Nature https://doi.org/10.1038/d41586-019-00135-8 (2019) doi:10.1038/d41586-019-00135-8.
42. Szkalej, K. The Paradox of Lawful Text and Data Mining? Some Experiences from the Research Sector and Where We (Should) Go from Here. GRUR Int 74, 307–319 (2025).
43. Breznau, N. & Nguyen, H. H. V. An Introduction to Generative Artificial Intelligence for Academics. F1000 Research Preprint Status, (2025).
44. Kwon, D. Open-access publishing fees deter researchers in the global south. Nature https://doi.org/10.1038/d41586-022-00342-w (2022) doi:10.1038/d41586-022-00342-w.
Ethical Statement
I have no conflict of interest to report. I used Google search which includes Gemini by default, ChatGPT, NotebookLM and Nano Banana to support my literature review, check arguments, suggest words or phrases and to extract facts. No writing was copied from Gen AI, it is all my own. Any mistakes or opinions are also my own.
