The British Educational Research Journal is the flagship BERA journal, and is responsible for a significant amount of AI-authored slop coming out of BERA. Earlier this year, it was obliged, following my intervention, to retract an entire special issue.

In this post, I will show that a recent call for papers from the BERJ falls well below basic academic standards, containing plagiarism and mistakes that would appear to follow from the use of AI. The call is titled ‘The Challenges of Combatting Misogyny: Educational Interventions and Responses’ and is from four BERJ editors: Prof. Alison MacKenzie, Dr. Yuwei Xu, Prof. Ming Cheng, Prof. Barbara Read. Each of the four main paragraphs contain egregious errors, which make it impossible to have confidence in the editors or the resulting special issue. I go through each paragraph below

First paragraph

The first main paragraph is entirely given over to a bracing first-person account from the Guardian of the online misogyny experienced by a 15 year old girl. However, the reader is immediately distracted by the call for papers treating ‘Anonymous’ as the girl’s actual name (the Guardian piece was written anonymously):

A recent article in The Guardian (Anonymous, 2026) describes the relentless misogyny experienced by a 15-year-old school pupil, Anonymous, online every day.

When Anonymous closes her apps, she feels ‘disheartened and unhappy about being a girl’.

To me, this could only be a mistake made by AI - any human would know not to capitalise ‘anonymous’ like that as if it were her actual name.

The paragraph also mischaracterises the actual experience of the girl as she recounts it. The Guardian piece is about the experience of generalised misogyny online, rather than misogyny specifically directed at her as an individual. However, the BERJ editorial somewhat sensationalises the account by making it about abuse the girl herself has directly suffered. Compare the accounts:

BERJ: As soon as she uses her mainstream social media apps, it takes only a ‘few minutes before there’s subtle or overt misogyny’, with comments about her body.

Guardian first-person account: It only takes a few minutes before there’s subtle or overt misogyny, such as comment sections on a girl’s post filled with remarks about her body.

And again:

BERJ: She has come to “loathe” her own face because of the relentlessly misogynistic comments about her body.

Guardian first-person account: Endless emphasis on beauty as worth and all kinds of videos criticising specific features, some of which I possess, have made me start to loathe my own face, as difficult as that is to admit.

Note that the second BERJ quote is incoherent in the way it moves (unsupported by the Guardian piece) from body to face.

Finally, the paragraph plagiarises the Guardian piece through insufficient citation in various instances, but most seriously here:

There is acknowledgement the general idea is from the Guardian piece, but the BERJ editors are falsely taking credit for the specific articulation of the idea.

Second paragraph

The second paragraph is largely made up of long quotations, but the editors do not close one quote, making it hard to discern where their own ideas begin. The end of the quote is actually later than the reader would probably expect, (after ‘law enforcement’, rather than before ‘on this view’):

Misogyny, according to Manne (2016:88), ‘upholds the social norms of patriarchies by policing and patrolling them; whereas sexism serves to justify these norms—largely via an ideology of supposedly “natural” differences between men and women with respect to their talents, interests, proclivities, and appetites. On this view, sexism is to misogyny as civic order is to law enforcement. It is, moreover, a ‘profoundly affective social dynamic’ (Wrisely, 2021, 189)…

Third paragraph

More indirect plagiarism:

Also, Plan International (2026) is not included in the references, only Plan International (2020), which is also cited.

Fourth Paragraph

More plagiarism, this time it is clear that the editors have for their ideas for tackling misogyny from UK Feminista (cited at the top of the previous paragraph) but do not credit them for the ideas at all. Note how the BERJ replaces UK Feminista verbs with synonyms (e.g. ‘combat’ to replace ‘tackle’ and ‘challenge’).

Conclusion

This is not the first error by the editors of the BERJ - to have published, and then not retracted, the amount of articles with fake references (let alone the other AI-related problems) is indefensible. It shows a basic lack of rigour in the peer review and editorial process. This post though provides evidence that their own work is incredibly sloppy, to put it in the most generous possible terms. A call for papers provides leadership and direction to the field. If it is riddled with errors and plagiarism, what message is it sending?

My focus is on documenting the extent to which educational academia is failing in this role at present, and to do so I have focused on uncovering the case of John Hattie (primer below), the most influential education academic of the 21st century and a complete fraud who has been protected by an array of academic institutions.

In this post, I will show Hattie’s latest escape from accountability. Despite including a fake reference, a phantom statistic, various irrelevant citations, and further issues, the journal Frontiers in Education declined to issue a retraction for his recent paper (issuing instead a hefty correction). In the post I provide further issues with the paper (including completely contradictory ethics statements) that should mean that the paper will be retracted (should being the operative word). I first highlight the possible, even probable, role that Hattie’s legal team is playing in enforcing the protection he is receiving from academic institutions.

The protection of a fraud

There are layers to Hattie’s fraudulence. Initially, after he ascribed a statistical value, in terms of the impact on student learning, of every possible aspect of education, he was called out the abysmal quality of his statistics. His results, in effect, were meaningless. Then, I demonstrated that he plagiarised extensively in every major work from his dissertation onwards. I also showed that he was so sloppy in extracting data for his statistical work, that he must have known that the data he was inputting into his models was likely inaccurate, meaning he was negligent as well as incompetent. And finally I have shown that in four separate publications, fake references have appeared (a sure sign of outsourcing writing to AI).

To complement the layers of his fraudulence, there are layers to the protection that he has received from scrutiny. When investigating his plagiarism, Melbourne University only had the ‘resources’ to investigate 14 out of the 189 instances supplied to them (presumably choosing the least serious instances); Melbourne University also declared, creatively, that they could not investigate the fake references as they were outside their remit. The publisher of a handbook where Hattie included a fake reference declared that it was a non-issue, as Hattie had merely used the wrong draft. And now, as I will outline, Frontiers has declined to issue a retraction despite finding the fake reference and myriad other issues.

Why are Melbourne University, Elgar, and Frontiers unwilling to see what I have shown to them in the clearest possible terms: that Hattie plagiarises prodigiously, is routinely reckless with data, and uses AI to write his work for him (he has also literally said he gets AI to do the work for him)? The reason is most probably not merely his personal influence and reputation within and beyond the University of Melbourne. It is likely that his use of the Elit law firm is also playing a major role. In December 2024, despite all my statements about Hattie being correct and fully evidenced, I received a legal letter issuing the following demands:

Each demand it is making of me is outrageous (I was completely in the right and they had no grounds to assert otherwise) but the truly remarkable aspect was that I drop my complaint to the university (OREI stands for the Office of Research Ethics and Integrity at the University of Melbourne. Surely, if my complaint was groundless, Hattie would have nothing to fear from any investigation by his employers?

If I am receiving this kind of baseless legal intimidation, you may assume that the University of Melbourne, Elgar, and Frontiers are receiving similar letters. As an individual, I can make the choice to simply ignore the threat on the basis that it is groundless, but these institutions may not feel the same freedom to be similarly cavalier, since legal action would be costly even if there is little chance that they would lose.

Fake reference but no retraction

This is the context behind a recent rather jaw-dropping correction in Frontiers to an article by Hattie following a complaint I made some months ago. The paper contained a number of referencing problems, including a fake reference (i.e. a hallucinated reference that is included in the paper to support a point but does not actually exist). Since even the fake reference on its own is a clear indication of AI authorship, I assumed, especially given the amount of issues I reported, that the journal would retract the paper. Clearly, the COPE guidelines would indicate that the paper should be retracted. The guidelines state that:

Editors can decide to retract a publication if they no longer have confidence in the results and conclusions reported in the paper. Reasons for this loss of confidence can be, but are not limited to:…Any form of misrepresentation, for example:…undisclosed involvement of artificial intelligence (AI).

Instead, the journal has decided to issue a very lengthy and substantive correction. The apparent reason for issuing a correction rather than a retraction is that ‘none of the corrections affect the study's data, methodology, analysis, or findings’. I can now show that the paper has more serious issues that very much do affect the study’s methodology and analysis. After providing the corrections in full, I will outline the further issues which mean that the editors can no longer have faith in the study and that it must be retracted.

Correction notice

—The correction notice in full—

‘In the published article, there were errors in several references in the literature review and in the bibliographic details of some entries in the reference list. The publisher and authors apologise for these errors. These corrections do not affect the study’s data, methodology, analysis, or findings.

1 Reference and attribution corrections in the text

1.1 Knake et al. (2021)

The reference for Knake et al. (2021) was erroneously listed as: Knake, A., Novak, E., Blanchard, S., and Tesch, B. (2021). “Digital curation in K-12 classrooms: Strategies for deeper learning.” TechTrends. 65, 89–97. doi: 10.1007/s11528-020-00541-5

This reference could not be verified. It has been removed from the reference list, and the corresponding sentence in Section 2.3 has been revised to remove the citation. The relevant point is supported by Torphy et al. (2020), which is already cited in the paper.

The original sentence in Section 2.3, paragraph 2 read:

“When materials from these sites are used to develop skills to remember or memorize facts, then learning decreases over time. However, learning is more likely to increase when they are curated to develop understanding and apply knowledge (Knake et al., 2021).”

The corrected sentence reads: “When materials from these sites are used to develop skills to remember or memorize facts, then learning decreases over time. However, learning is more likely to increase when they are curated to develop understanding and apply knowledge.”

1.2 Sahin-Taskin (2017)

The reference for Sahin-Taskin (2017) was erroneously listed as: Sahin-Taskin, C. (2017). Exploring pre-service teachers’ perceptions of lesson planning in primary education. Int. Electron. J. Elemen. Educ. 9, 599–612. doi: 10.1007/s10639-017-9576-3

The correct reference is: Sahin-Taskin, C. (2017). Exploring pre-service teachers’ perceptions of lesson planning in primary education. Journal of Education and Practice, 8(12), 57–63.

In addition, the claims attributed to Sahin-Taskin (2017) in Sections 2.1 and 2.5 extended beyond what the source directly supports. Sahin-Taskin’s study examined pre-service teachers’ perceptions of lesson planning and did not compare novice and experienced teachers.

The original sentence in Section 2.1 read:

“Sahin-Taskin (2017) provides empirical evidence that novice teachers create more detailed lesson plans, incorporating explicit instructional strategies, whereas experienced teachers modify and adapt materials more fluidly based on their classroom contexts.”

The corrected sentence reads: “Prior research suggests that novice teachers tend to create more detailed lesson plans incorporating explicit instructional strategies, whereas experienced teachers modify and adapt materials more fluidly based on their classroom contexts (Griffey and Housner, 1991; Krepf and König, 2023).”

The original sentence in Section 2.5 read:

“Teachers earlier in their careers tend to write longer, more comprehensive plans, including explicit teaching strategies, learning activities, and assessment tasks, but often have more difficulty aligning curriculum, assessment, and reporting requirements (Sahin-Taskin, 2017).”

The corrected sentence reads: “Teachers earlier in their careers tend to write longer, more comprehensive plans, including explicit teaching strategies, learning activities, and assessment tasks, but often have more difficulty aligning curriculum, assessment, and reporting requirements (Griffey and Housner, 1991; Krepf and König, 2023).”

1.3 Frank et al. (2018) and the Pinterest statistic

The statement attributing the statistic that “87% of teachers used Pinterest in lesson planning and delivery of instruction” to Frank et al. (2018) was incorrect. This figure does not originate from the Frank et al. (2018) chapter. The sentence has been revised to remove the specific percentage and to attribute the broader claim to a source that supports it.

The original sentence in Section 2.3 read:

“Given the rise and ease of access to the internet, it is thus no surprise that the greatest source of these resources is Pinterest and YouTube. Frank et al. (2018) noted 87% of teachers used Pinterest in lesson planning and delivery of instruction.”

The corrected sentence reads: “Given the rise and ease of access to the internet, online platforms such as Pinterest and YouTube have become widely used sources of teaching resources, with research showing that teachers frequently turn to such platforms for lesson planning and instructional materials (Torphy et al., 2020).”

1.4 Ball and Cohen (1996)

The paraphrase of Ball and Cohen (1996) in Section 2.1 has been revised to more precisely reflect the source’s argument.

The original sentence read:

“Ball and Cohen (1996) emphasized the potential of curriculum materials to be key agents in improving classroom instruction, arguing that well-designed materials support teachers in making instructional decisions and adapting to the needs of diverse learners.”

The corrected sentence reads: “Ball and Cohen (1996) emphasized the potential of curriculum materials to serve as key agents in improving classroom instruction, arguing that well-designed materials can support teachers’ instructional decision-making and contribute to instructional reform.”

2 Corrections to bibliographic entries in the reference list

A full audit of the reference list was conducted. In addition to the references corrected above, the following bibliographic entries contained errors that have now been corrected:

2.1 Apple and Christian-Smith

The publication year was erroneously listed as 2017. The correct year is 1991. The reference now reads: Apple, M. W., and Christian-Smith, L. K. (1991). The politics of the textbook. New York: Routledge.

2.2 Bouckaert (2018)

The page range was erroneously listed as 238–257 and the DOI was missing. The correct entry now reads: Bouckaert, M. (2018). Current perspectives on teachers as materials developers: why, what, and how? RELC J. 50(3), 439–456. doi: 10.1177/0033688218810549

2.3 Campbell (2019)

The entry was erroneously listed with DOI 10.1080/00131857.2017.1380842. The correct DOI is 10.1080/00131857.2018.1425992.

2.4 Darling-Hammond (2010)

The entry erroneously included a volume number and page range that do not apply to this monograph. The corrected entry reads: Darling-Hammond, L. (2010). The flat world and education: How America’s commitment to equity will determine our future. New York: Teachers College Press.

2.5 European commission

The publication year was erroneously listed as 2020. The correct entry now reads: European Commission (2013). Ethics for researchers: Facilitating research excellence in FP7. Luxembourg: Publications Office of the European Union.

2.6 Krepf and König (2023)

The entry was erroneously listed with DOI 10.1080/02607476.2023.2196267. The correct DOI is 10.1080/02607476.2022.2151877.

2.7 Roschelle et al.

The publication year was erroneously listed as 2010. The corrected entry now reads: Roschelle, J., Pea, R., Hoadley, C., Gordin, D., and Means, B. (2000). Changing how and what children learn in school with computer-based technologies. Futur. Child. 10, 76–101. doi: 10.2307/1602690

2.8 Siemens and Long (2011)

The entry was erroneously listed with DOI 10.1080/08923647.2011.538646. The Siemens and Long (2011) article was published in EDUCAUSE Review and does not have a DOI. The corrected entry reads: Siemens, G., and Long, P. (2011). Penetrating the fog: analytics in learning and education. EDUCAUSE Review, 46(5), 30–40.

The authors apologise for these errors and confirm that none of the corrections affect the study’s data, methodology, analysis, or findings. The original version of this article has been updated.’

—End of correction notice—

Further problems in the paper

Although Frontiers should be commended for producing a timely and extensive retraction, it is unfathomable to me that editors could recognise all these referencing errors and still have faith in the authors to accurately undertake the methods, results, and analysis. My strong sense, as described in the introduction, is that Hattie’s lawyers had a role in limiting the journal’s response to a correction, although I should acknowledge this is entirely speculative on my part.

However, the paper is so flawed that it took me five minutes reading through the Methods and Discussion sections to notice further serious referencing problems. The most serious issue is to do with the Methods. The authors provide a completely contradictory account about informed consent (this was a questionnaire study of 2,594 teachers).

In the ethics statement they report that the did not gain the informed consent of participants (I have no idea whether or not they were correct in stating it was not required):

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the participants was not required to participate in this study in accordance with the national legislation and the institutional requirements.

However in the methods section, in the only reference provided, the authors state that ‘The study adhered to European Commission ethics guidelines (2013)’. But in these ethics guidelines, the fourth principle is that research ‘honours the principle of informed consent…’

So, the authors in the article state they gained informed consent, while in the ethics statement indicate they did not do so. More than likely, AI wrote the methods section and so the authors were unaware of the contradiction they were forming.

In relation to failures related to the authors’ analysis, they write the following in the Discussion:

This reinforces the need for investment in digital infrastructure for collaborative resource sharing and aligns with recent policy goals on equity and professional development through digital teaching tools (European Commission, 2013).

The problem is that the reference is to the above paper that has absolutely nothing to do with resource sharing or professional development:

European Commission (2013). Ethics for researchers: Facilitating research excellence in FP7. Luxembourg: Publications Office of the European Union.

So, in addition to the array of corrections listed above, Frontiers now has to decide whether to issue further corrections or whether to retract. It goes almost without saying that the correct course of action is to retract given that the authors have provided contradictory ethics statements in addition to including a fake reference, a phantom statistic, citations that do not support the claim being made, and many further errors.

But on the other side of the scales, there is the issue of Hattie’s legal team. I hope that in publicly talking about the empty and absurd threats that Hattie’s legal team have made, that institutions such as Frontiers but also the University of Melbourne, may be encouraged to take the appropriate steps to defend academic integrity, rather than bow to the pressure of a fraud.

What is at stake?

For academia to continue to have a prominent role within society, it is imperative that the public can, on the whole, trust the knowledge that it produces. And there are clear principles and practices that are deeply entrenched within universities that enable trustworthy knowledge to be produced. However, at the same time, it is in the interests of a lot of actors within academia to bend and even break such principles and practices. Disreputable researchers can forge careers, while universities and institutions may turn a blind eye to academics and students who bring in prestige and money to the university.

It is a battle that cannot be lost, but is not being won. Cases such as Hattie, and in particular his Frontiers articles, are important because they pinpoint a high-profile situation in which this contest is being played out. On the one hand, it is imperative that journals and publishers do not produce knowledge that has been generated by AI and therefore is likely to sound plausible but be riddled with errors, short on wisdom, and offer little in the way of substantively useful knowledge. On the other hand there are powerful individuals (Hattie is influential, rich, and well-connected, and is employing a law firm to threaten those who would hold him to account) and institutions (Melbourne have chosen to turn a blind eye, to protect their own interests) who are working hard to ensure that shoddy research goes unchallenged.

More specifically, it is clear as day that the Frontiers article has absolutely no place in the scholarly literature, and this would be more than apparent to any moderately unbiased observer. If it is not removed, and if the rest of Hattie’s plagiarised and AI-authored work is not removed, then there is little hope for the forces safeguarding academic integrity to win out.

In the last year, I’ve been interested in the publication of the journals related to the British Educational Research Association (BERA). BERA is a highly respected institution - most educational research academics are members - that has the largest general educational research conference in the UK. The four journals are internationally renowned, with the flagship British Educational Research Journal being the most established. In short, BERA is the educational research establishment.

I have shown that BERA has produced somewhere between dozens and hundreds of articles written by AI and submitted by ‘diverse’ authors (i.e. authors who come from groups historically under-represented in the journal. Many of these authors are based in China and other countries where English is not the first language and which do not have a robust system of enforcing academic integrity. In response, BERA has labelled my approach ‘ethically problematic’ and has only taken action in a small number of cases where it (falsely) claimed to have ‘independently uncovered’ evidence. This is the full story of BERA’s failures of peer review and refusal to act.

The policy

As part of the ‘action plan’ to realise its Race Equity Policy, BERA aims to increase representation in journals:

BERA aim to increase the range and diversity of authors being published in the journals. We are reliant on Wiley for this data and there will be no detriment on quality.

It is remarkable how cavalier BERA are about the potential trade-off between representation and quality; they simply blithely asserts there is none. In the traditional approach, quality (along with relevance) is of course the sole consideration. Adding a further consideration (author identity) surely has a very high likelihood of diluting this quality? I fail to understand how this could not be the case. As we will see, it will become apparent that BERA simply do not care about the failure in quality.

Before we get to the damning evidence of the fake references, and how they’ve been handled, I want to first present extremely low-quality articles that have been published in the four BERA journals. They are so bad that it is clear that peer review cannot have occured in any meaningful sense. I have numerous examples from across my substack, but to present three:

• A paragraph repeated three times in a row:

• An 8-line results section

• And perhaps the worst figure ever made, with various prepositions serving as themes (and a whole lot more nonsense)

There are endless further examples involving contradictory ethics statements, impossible methods sections, and irrelevant citations. But overall there are hundreds of papers that say nothing and do so, very obviously, in AI’s language. If there had been even one mildly discerning reviewer who was chosen and listened to, these papers would not have been published.

The fake references

BERA has published 22 articles with fake references, and has retracted up to this point only three. By a fake reference, I mean a reference that AI has hallucinated (i.e. invented); although usually the authors and journals are real, the titles are invented based on a plausible title for the topic. As discussed above, the peer review was so lacking, it is hardly surprising that these fake references were missed. However, there is no excuse not to retract promptly when fake references are identified. COPE is very clear on this, saying that retraction is warranted when there is ‘Any form of misrepresentation, for example:… undisclosed involvement of artificial intelligence (AI)’. They further write that ‘retraction notices should be published as soon as the editor is confident that the publication is seriously flawed, misleading, or falls into any of the categories described above’.

Last year I sent BERA information about four different articles with multiple (5-8) fake references (see Table 1). Despite this, they still haven’t taken action.

[note: BERJ = British Educational Research Journal; CJ - The Curriculum Journal]

For at least five months, BERA has known about five further articles with multiple fake references. Again, they haven’t taken action (see Table 2).

[note RoE - Review of Education; BJET = British Journal of Educational Technology]

Not only that, but they have continued to publish articles with fake references (see Table 3). Details of all the fake references can be found here.

Apart from their own investigation which was triggered by but ‘independent from my claims (see below), BERA has not retracted or corrected the many articles I have sent to them. In most cases in relation to the fake references they indicate the investigation is ‘still ongoing’, although how it can take a year to investigate obviously fake references is beyond me. Bizarrely, in two cases of multiple fake references, they have indicated that they will take ‘no action’, having completed their investigation:

1. https://doi.org/10.1111/bjet.13471

   Fake references:

• Snyder, J. J., & Snyder, M. J. (2008). Searching for education: The usefulness and user-friendliness of ERIC and Google scholar. Journal of Education for Library and Information Science, 49(3), 181–192.

• Wu, P. F., Wu, T. T., & Hwang, G. J. (2021). Gamification, motivation and learning performance in distance education: A meta-analysis. Computers & Education, 173, 104295. https://doi.org/10.1016/j.compedu.2021.104295

• Yang, S. J., Gao, J. X., & Cui, G. Z. (2020). Effects of gamification on students’ academic performance in STEM education: A meta-analysis. British Journal of Educational Technology, 51(4), 1186–1207.

2. https://doi.org/10.1002/curj.305

   Fake references:

• Johnson, E., Williams, J., & Lee, J. (2019): The impact of content and language integrated learning (CLIL) on preschoolers’ language proficiency. Early Childhood Education Journal, 47(3).

• Kotsopoulou, A., & Mitsikopoulou, B. (2014): A CLIL project in Greek kindergartens: A study of pedagogical practices and parental attitudes. International Journal of Bilingual Education and Bilingualism, 17(3).

• Papadimitriou, A. (2019): Content and language integrated learning in Greek preschools: A case study. International Journal of Early Childhood Learning, 25(1).

• Skourtou, E., & Vlachopoulou, M. (2015): Content and language integrated learning in early childhood education: The Greek case. International Journal of Early Childhood Learning, 21(2).

• Thompson, S., Lange, T., & Koops, C. (2018): The impact of content and language integrated learning (CLIL) on language development in early childhood education. International Journal of Bilingual Education and Bilingualism, 21(6).

  

This is mystifying, but may be explained by their rather strange stance, described below, that undisclosed use of AI is not necessarily an issue.

Why have BERA not issued retractions

In April this year BERA did in fact retract 32 articles, most of which I had flagged to them. However, the basis for their retraction was misleading and self-serving in two ways:

1. They suggested that in investigating my claims, they ‘uncovered independently that the peer review process for a limited number of articles in the British Educational Research Journal (BERJ) had been manipulated’. The phrase ‘uncovered independently’ is ridiculous, since before they began their investigation I had already provided evidence that an editor of the special section was engaged in highly suspicious peer review practices.

2. They claim that ‘undisclosed AI use does not always indicate fraud or manipulation; however, when combined with other signals, it can indicate that a paper needs further examination. This view is consistent with COPE retraction guidelines, which indicate that retraction decisions should focus on confidence in the publication’s results and conclusions, rather than on any one signal in isolation’. As discussed above, this is flatly wrong. The COPE guidelines indicate undisclosed use of AI is sufficient grounds for retraction and should be done in a timely way.   

Here is a situation all too common in academia today (especially perhaps in education). Academic leaders are totally invested in targets that legitimate themselves, such as EDI objectives, but they are completely lacking in the perspicacity to implement these objectives in ways that enhance, and do not detract from the core academic mission.

BERA - another institution failing in its core purpose

In conclusion, here are the three ways in which BERA have allowed EDI targets to completely distort their core role:

1. Refuse to foresee any potential conflict between increasing author diversity and retaining quality.

2. Apparently being willing to soften peer review and editorial standards for diverse authors.

3. Being unwilling to change direction and take appropriate action when given considerable evidence that basic standards had been completely lost.

The only question is the scale of the scandal. It can of course be difficult to prove AI authorship, although I think this tends to be somewhat overstated. My own approach (which BERA labelled ‘ethically problematic’) is to find phrases that, according to Google Scholar, were not in use, or barely in use, before 2023 but have become (more) common since. If there are a preponderance of such phrases (15+) then I think it is reasonable to conclude AI had a significant role in writing the article. Using this approach I found 131 AI-written articles in BERA journals almost a year ago; I imagine the number would by now be 250+. A more minimal number would be found here (around 50), where there are particular issues to point to that only an AI could produce, such as irrelevant citations or impossible methods sections/ethics statements. I would absolutely maintain that if anyone fair-minded went through the BERA archive over the last three years, they would be convinced they are reading AI works at scale.

But somehow the leaders at the British Educational Research Association are more invested in retaining their shaky legitimacy by denying the possibility of any issue than ensuring that the work they publish has integrity. They won’t even retract articles with multiple fake references. It’s quite remarkable.

The real significance of the Arday case is that it serves as a further demonstration that the field of educational academia has lost a sense of its primary, substantive purpose: to generate and share knowledge that will improve education. But academia is now extremely tightly managed, and is judged against various more-or-less politicised metrics at every level from the institution to the individual academic. This managerialism, as we all know by now, has perverse consequences. Not only is research judged against spurious measures of quality and impact that may well have very little to do with the actual impact on policy and practice, but academia is judged against broader social metrics, including those related to diversity and inclusion, that may very well serve to distract from or supersede academic issues.

In the case of educational academia, none of these complaints are original, but I would suggest that this has resulted in:

• too much of the work becoming detached from the realities and requirements of classroom practice; where there are implications, they are generally facile;

• too much of the work being derivative and unoriginal, narrowly focused on relatively arcane corners of theory (including but certainly not limited to critical race theory);

• secondary characteristics, including but not limited to identity markers, playing a central role in hiring and promotion, over and above ability and quality.

But far worse than this, the leaders in the field do not seem to be interested in putting the field back on track - they are more concerned with following what is strategically useful to maintaining their status in the fantastical world of the university.

Hattie and BERA

To illustrate the significance of Arday’s case, I want to draw on some of the projects I’ve been interested in - John Hattie and the British Educational Research Association (BERA). Hattie’s case shows that the problem of not properly investigating integrity issues is not unique to Cambridge, or Arday, or primarily a question of race. After I demonstrated clear and incontrovertible evidence of extensive plagiarism by Hattie in some of the most important works of 21st century educational scholarship, the University of Melbourne performed the most laughably purposefully blind investigation imaginable. Hattie and Arday are both stars valuable to the reputation of the university, so they were protected.

The comparison between their scholarship is also instructive. Hattie’s quantitative research was finely tuned to chime with the aims of educational policymakers - to make education more efficient through the adoption of business practices. Although very different, Arday’s qualitative work in critical race theory was equally attuned to political currents of the day, in terms of how institutions are now legitimating themselves through EDI work. So again, the important aspect of Arday’s work is not related to race, rather political usefulness for the institution.

The case of BERA, which also has similarities to the Arday case, is equally instructive. I have shown that over the last few years, the four BERA-affiliated journals have published obviously poor quality AI-written work including over 20 articles with fake references. This has happened because, having set targets to increase the diversity of authorship in the journals, they have apparently relaxed the rigour of peer review for historically under-represented identities. In the face of evidence of flawed research, they have doubled down, refusing to take action and saying that my approach was ‘ethically problematic’. The BERA case should be understood in part as the reaction of a body of academics who have been sidelined since Gove from performing their primary role - informing educational policymaking and practice. In response, they have retreated to legitimating themselves through managerialist targets related to identity.

The case of Cremin

So, the case of Arday should be understood in the following terms. A mistake, but not the biggest mistake, was hiring him because of his identity rather than his scholarship. The really telling mistake though, was how the university handled the revelation, soon after he was appointed, that his scholarship was illegitimate. This is what Professor Hilary Cremin, the Head of the Education Faculty, wrote to the internal complainant about Arday’s plagiarism:

[W]e remain immensely proud of Prof Arday’s achievements and scholarship, and...he has our full support.

…During our meeting Prof Arday spoke movingly about his ongoing experiences of direct and indirect racism, and the conditions that he was living in whilst the 2018 article was being prepared (in 2016). At this time he was early career and living away from home whilst dealing with significant health issues.

It is here that an educational academic leader - seemingly without a great deal of reflection - insisted the university was ‘immensely proud’ of shoddy scholarship and explained away the plagiarism by saying a 31 year old was ‘living away from home’. In other words, she turned her back on finding out the truth about the value of Arday’s scholarship and insisted, tacitly but nevertheless with absolute institutional force, that he was beyond criticism.

Why is this moment telling? Because it reveals, with absolute clarity, that educational academic leaders are willing, without hesitation, to sideline academic integrity in order to protect figures who legitimate them in our highly politicised managerialist system. As demonstrated by Melbourne and BERA, they are totally invested in their reputations based on secondary issues such as EDI at the expense of their core role.

Conclusion

So, race does matter in the Arday affair - which should really be the Cremin affair - but not in a direct, straightforward way. Jason Arday was not protected by the university because he’s Black - if he was a white superstar researcher doing bad quantitative research he would have been similarly sheltered from scrutiny. Rather, his race was an opportunity for leaders such as Cremin to legitimate themselves without doing the difficult work of making educational academia relevant and actually useful again.

The problem is that universities, and educational departments in particular, are existing in a fabricated world, where they are judged on measures other than their basic core purpose. Educational academic leaders need the vision to discern the substantive purpose of educational research from the fake metrics, but it is clear they are complacent enough to be happy to take the easy wins related to EDI and various impact measures, even if this means ignoring and sabotaging their true purpose. The tragedy is that the current generation of leaders are currently grooming the next generation in their image; it is hard to see any future for educational academia beyond increasing obscurity.

I’ve written about how, even though he has fundamental disagreements with the basics of their educational approach, Carl Hendrick works for, and has strongly advocated for, Alpha school. In this post I show that today, without communicating two relevant conflicts of interest, he promoted research that is both connected and favourable to Alpha.

Both in his blog, and in his books, Hendrick is a trusted populariser of educational research, with over 23,000 subscribers on Substack. And from Alpha’s point of view, there are presumably two reasons for employing Hendrick. The first is his expertise in the science of learning, but the second is potentially much more valuable: to help sell their model to the broader educational community (which is by and large sceptical to the idea of AI-led learning). And indeed, on social media and in interviews, Hendrick has made game attempts to praise Alpha School’s virtues.

It is impossible to tell, of course, whether Hendrick’s promotion of Alpha School is part of his contract with them, or whether it merely comes from his goodwill and enthusiasm to share their work.

Today, in his monthly round-up of research, Hendrick shared a glowing study on Khan academy, the online learning platform: “How In-School Supervised Ed-Tech Support Produces Massive Learning Gains: A Khan Academy Field Experiment in India”. This is noteworthy because not only is Khan Academy used by Alpha Schools, but, as Hendrick emphasised, the study found that the most effective model of using edtech is Alpha School’s specific model - of using AI tutors in a human-supervised environment. You would also have no idea from Hendrick’s summary that ‘Oreopoulos and colleagues’ is not exactly correct. While Oreopoulos is employed by the NBER, Deepak Agarwal is in fact Principal Scientist at…Khan Academy. Omitting this is particularly egregious in a ‘randomised controlled trial’, which makes the study sound wholly rigorous. Here is Hendrick’s summary:

Supervised AI tutoring produces massive gains — but the key word is “supervised”

Oreopoulos and colleagues published an NBER working paper reporting results from a randomised controlled trial of Khan Academy in Indian schools. The gains were substantial. But the critical design feature was supervision: students used the platform during structured school time with teacher oversight, not independently at home.

This aligns with a growing body of evidence that technology-assisted learning works when it is embedded in a structured instructional environment rather than deployed as a substitute for teaching. The Khan Academy platform provided the content and adaptive sequencing; the teachers provided the structure, accountability, and human support that kept students engaged and on task. Remove either component and the effect would likely shrink.

For schools considering AI tutoring platforms, the lesson is that the technology is a complement to teaching rather than a replacement for it. The gains come from the combination, not from the software alone. This is expensive and labour-intensive, which is precisely why it works: there are no cheap shortcuts to effective instruction.

This may well be me being excessively leery, but it seems to me that the last sentence in particular is remarkably effusive about the Alpha School model; ‘precisely why it works’, ‘no cheap shortcuts’ are slightly OTT phrases for research summaries (at least in my view).

But to be clear, I have no idea whether or not Hendrick’s research roundup was influenced by his position at Alpha. In any way. But I do know there are two levels of conflicts of interest that Hendrick did not report. First, as he is employed by Alpha School, and Alpha School uses Khan Academy, he should have acknowledged his position (not to mention that he uses the study as an endorsement of Alpha School’s model). Second, the study itself was co-authored by an employee of Khan Academy, which would have an obvious and significant impact on how much significance Hendrick’s readers would attribute to the study.

So, to summarise, Hendrick is using his platform that he built up as an impartial conduit of the science of learning, to summarise a very positive study on Khan Academy (used by Alpha) that supports Alpha’s educational model, without acknowledging either that he himself works for Alpha or that a co-author of the study works for Khan Academy. At best, it all seems a bit messy.

It is known by now that if you can write engagingly and build up a large audience as an educational academic, then it is possible to brush off dubious science and conflicts of interest. The balance of forces is stacked against holding the influential to account, as there simply isn’t the requisite institutional support to maintain standards. So Hendrick will not be the last to disregard once-required acknowledgements of conflict of interest in leveraging his credentials to boost dubious educational ideas (AI or otherwise). But we’ve been warned.

The central theory is simple. Human life, however varied, consists in the performance of specific activities. Education that prepares for life is one that prepares definitely and adequately for these specific activities. However numerous and diverse they may be for any social class, they can be discovered. This requires only that one go out into the world of affairs and discover the particulars of which these affairs consist. These will show the abilities, attitudes, habits, appreciations, and forms of knowledge that men need. These will be the objectives of the curriculum. They will be numerous, definite, and particularized. The curriculum will then be that series of experiences which children and youth must have by way of attaining those objectives.

In an alternate timeline, it would have been perfectly possible that we developed such a science. It would have been fascinating and, I would suggest, hugely beneficial. But in the 1980s the preparation for ‘human life’ was flattened into the preparation of ‘human capital’; the purpose of education reduced to maximising the learning productivity of students to increase an abstracted sense of their economic value.

So, we have no science of education and no hope of one.

But we do have a science of learning: a science that aims to increase the efficiency of learning through an understanding of how the brain learns. Clearly, in a well-developed science of education, the science of learning would be an important branch, but not necessarily the predominant one, which would surely be the monumental task of devising curriculum objectives and experiences, as described above. The efficiency of learning promised by the science of learning would be ultimately subordinate to the social efficiency of ensuring education prepared children for life.

But we are in the age of learning and human capital maximisation, and so the science of learning reigns supreme. No figure encapsulates this triumph more than Carl Hendrick, influential author and blogger, UNESCO sage, advisor in renewing the Northern Irish education system.

Hendrick has raised eyebrows in working with AI-based private school chain Alpha, for the very simple reason that Alpha is at odds with almost all his stated educational beliefs:

• Hendrick has argued for the importance of whole books; Alpha uses Teach Tales, ‘an AI-powered story-building platform that turns your kid into the author of their own custom adventures’.

• Hendrick has advocated for the importance of knowledge; Alpha emphasises engagement and extra-curricular activities, limiting academics to 2 hours and emphasising that all subjects (even history) are taught through hooking onto the child’s pop-cultural interests. They even promise that ‘If your kid loves Taylor Swift, we’ll teach World War I through her tour routes’.

• Hendrick has argued against growth mindset programs; Alpha school begins every day with it.

• Hendrick is against Duolingo; Alpha uses it .

• Hendrick has been against screentime in schools; at Alpha there is 2 hours built in every morning.

His explanation for why he is working with Alpha is totally unconvincing:

And then one thing that I saw really changed my mind which is this idea of of AI coaching. So what what they’re working on now is a platform called timeback which will be where when you log on to the learning platform there’s a sort of a monitoring element to it. Now every teacher knows in a good classroom um the teacher who is monitoring and ensuring that attention is happening that learning is happening, I think Doug Lemov describes as ‘be seen looking’. So if if you’re a good teacher will sort of go “right I’m here, I’m watching you, I’m here to support you but I’m going to just make sure that you’re you’re on task”. I saw some recent studies which shows that you know kids left to their own devices on on devices will pretty much drop out. We know that from 20 years ago with the studies on MOOCs unless there’s someone watching you unless there’s someone kind of invested in you they’re not going to do maybe the top 5% who are hugely motivated. It’s not going to work at scale. So this platform will provide a what they’re calling a waste meter. So it will measure time on task, it will measure latency, it will measure lots of elements like obviously the the big ones will be if they go on YouTube, if they go on whatever, chat bots…

This can’t be true - no one could be so inspired by such a dreary form of online surveillance. And it isn’t true. Later on in the interview, the real reason why Hendrick is interested in Alpha school is revealed. In a fascinating segment, he starts explaining that he doesn’t want his daughters on screens, and then, trying to use this as a build up to a defence of Alpha, he… literally can’t do it. He trails off and then explains the actual reason he is interested, which is that the AI-led learning environment can facilitate the development of the science of learning:

So, we have lots of ethical questions about education. What kind of education do we want for kids? Do we want them, you know, um I want my kids less time on screens? I want them more time interacting and playing with each other. I want them more time um doing things that they’re that they’re fired up to do. And this model is….

And then the second thing is I want learning to be driven by the science of learning. I want it to be driven by what we know about human cognition about how we attend to process, retrieve, store information and knowledge, that change in long-term memory. And if you’re interested in that, this is where it’s going. This is where the science will be furthered. So and again it’s it’s this thing about being a scientist like do you accept the results of things that you wish weren’t true that you don’t like and I think we’re going to see in the next 18 months, 2 years, 5 years massive changes in what we know about learning and it’s all going to come from these systems.

Hendrick may or may not be correct about his wager (I’m sceptical, as screen-led learning is such a dry, unnatural experience that I would be surprised if significant findings emerge that can be generalised to non-screen learning, but who knows). The real point though is that it shows that Hendrick has completely lost his educational compass. He is prepared to work with and for a company whose educational ethos he does not believe in because of his elevation of efficiency over and above purpose. ‘I want learning to be driven by the science of learning’ is a fitting credo for our age when the educational establishment has no meaningful vision for education rather than schools working towards maximising knowledge.

To return to Bobbitt (1918), he argued that if the curriculum is not scientifically formulated to meet human needs, then education itself is inevitably unscientific:

The curriculum, however, is a primordial factor. If it is wrongly drawn up on the basis merely of guess and personal opinion, all of the science in the world applied to the factors above enumerated will not make the work efficient.

The optimisation of spacing for retrieval practice may well be of great personal interest to Carl Hendrick. But he is foolish to jettison his educational principles and beliefs to chase it. He may well be an excellent cognitive scientist but, like almost everyone in the new educational establishment, as an educational scientist he is hopelessly floundering.

But when it comes to the most important aspect of education - determining the very purpose of schools - they prove to be remarkably soft-headed. Without due consideration, they have turned their backs on the established conservative position on the purpose of schools: that schools are to prepare children for adult life. Instead, they have swallowed whole the eccentric and very lightly-evidenced position of an English professor: that the purpose of schools is to ensure a common stock of general knowledge. Again, education as an Adam Curtis doc.

The post tells the story about how, for at least the past 15 years, we have allowed a literature professor’s hobby horse about a particular feature of literacy to completely dominate how we understand the purpose of schools. It matters because schooling is foundational to human lives and we compel children to attend. The new educational establishment have no more right to arbitrarily decide the purpose of education than a teacher has to continue using debunked approaches. In fact, the stakes are far higher.

An adequate vision for education

Before we get into the story, I want to be clear from the outset that my perspective is not a particularly ‘progressive’ one. I would defend the old conservative position, from the early 20th century, as proposed by John Franklin Bobbit or William Bagley: that schooling should prepare people for their public and private lives. As articulated by Biesta (2009), schools must be serve three overlapping purposes in this regard. They must be:

• a place of socialisation, where people can become a fully contributing participant to their families, communities, and societies;

• a place where people gain (preliminary) qualifications for various jobs, and gain the foundations for a secure career;

• a place where people understand themselves as individuals, connected to but separate from others.

This is not a radical or exhaustive list - many would add further purposes, such as increasing equality and supporting families through providing childcare (Goodlad, 1979); it also does not take into account the experience that children have of school. It simply reflects the desire that, ideally, when people leave the education system, they are confident and prepared to take on adult life. And in achieving this for the individual, the public who support and fund schools will reproduce and progress their society, culture, and economy.

These purposes of education are not controversial or complex. The important factor is that they are sufficiently broad in terms of recognising the importance of education to a child’s life.

Hirsch’s fantasy of Shakespeare quoting execs

But the new educational establishment is long cut off from sensible old conservative ideas about education. Since the early 1980s, they have been singularly focused on maximising the human capital produced by schools, in the process turning schools into businesses in order to maximise learning productivity. Their starting point was not the prospects of the individual students but the health of the economy as a whole. In a knowledge economy, schools needed to maximise knowledge in the most efficient way possible. With this de-individualised, economistic turn, the reform movement jettisoned the old conservative perspective on what schools are for.

E.D. Hirsch was a particularly eccentric member of the reform movement who used the bandwagon of the crisis of improving the human capital produced by American schools to advance his particular hobby horse - that the amount of shared knowledge among Americans was declining. An English lecturer, he developed the notion that a society could not be meaningfully literate if it lacked a stock of general knowledge that could be assumed of all citizens. This expansive argument was built on the more narrow, and by now familiar, argument that we need content knowledge to read or write critically. Following the Nation at Risk report (1983), which argued that the American economy was failing in large part because American schools were failing, he framed the argument in economistic terms. This is where his argument apparently became more persuasive but also more tenuous and strange. Adopting the template shared by so many education reformers, on the first page of his major book Cultural Literacy, he framed his argument in terms of the economic threat from Japan:

The standard of literacy required by modern society has been rising throughout the developed world, but American literacy rates have not risen to meet this standard. What seemed an acceptable level in the 1950s is no longer acceptable in the late 1980s, when only highly literate societies can prosper economically. Much of Japan's industrial efficiency has been credited to its almost universally high level of literacy. But in the United States…

It’s a hard argument to make that economic success depends on a large stock of shared general knowledge, but Hirsch makes a game attempt:

My father used to write business letters that alluded to Shakespeare. These allusions were effective for conveying complex messages to his associates, because, in his day, business people could make such allusions with every expectation of being understood. For instance, in my father's commodity business, the timing of sales and purchases was all-important, and he would sometimes write or say to his colleagues, "There is a tide," without further elaboration. Those four words carried not only a lot of complex information, but also the persuasive force of a proverb. In addition to the basic practical meaning, "Act now!" what came across was a lot of implicit reasons why immediate action was important.

When he tries to draw a broad conclusion from this he is reaching so much that it is hard, even now, to avoid second-hand embarrassment:

The moral of this tale is not that reading Shakespeare will help one rise in the business world. My point is a broader one. The fact that middle-level executives no longer share literate background knowledge is a chief cause of their inability to communicate effectively.

This was not correct then, and is not correct now. It is possible, in fact, to make a counter-argument - that it would have been better for his father to spell out his position in plodding, common-sense terms, rather than using the potentially vague allusion to Shakespeare. We communicate with people perfectly well on a professional level without having anything culturally in common with at all. Many of us do it every day. His conclusion is wildly speculative, as evidence-free as learning styles:

We will be able to achieve a just and prosperous society only when our schools ensure that everyone commands enough shared background knowledge to be able to communicate effectively with everyone else.

Hirsch’s fantasy made policy

Hirsch’s promotion cultural literacy as the purpose of education was avowedly polemical (‘monomaniacal’ in how own words). Therefore it was the height of irresponsibility when English education ministers Nick Gibb and Michael Gove read it, and like sophomores, promptly based their understanding of the entire education system on it:

No single writer has influenced my thinking on education more than E. D. Hirsch. Like any book which becomes seminal in one’s intellectual journey, I distinctly remember the first time I encountered Hirsch’s work.

As a quick aside, Gove and Gibb also drew on the work of Michael Young, who similarly argued that human flourishing is not sufficiently academic to be the basis for a curriculum:

No one could take issue with his claim that schools should promote human happiness and well-being. The problem is that such goals apply equally to all institutions (except perhaps prisons) and they say nothing specific about what schools are for and what distinguishes their role from that of other institutions.

But Young’s argument is no more convincing. On his basis, the starting point for understanding the purpose schools is in identifying their unique capacities. This would seem to be putting the cart before the horse. The purpose of schools is in how they serve children and society. Children spend much of their lives in schools, and even if the role of schools overlap with some other institutions, it does not seem logical to preclude schools from working towards the wellbeing of children. It is in fact arbitrary to limit the purpose of schools in this way. It is not for us to ensure schools realise their purest essence of delivering knowledge, especially if this version of schools do not support human flourishing.

Drawing on Hirsch and Young, Gibb then redrew the basis of the National Curriculum:

We inherited a National Curriculum from 2007 which, at its opening, stated its aims as enabling young people to become:

- successful learners who enjoy learning, make progress and achieve;

- confident individuals who are able to live safe, healthy and fulfilling lives;

- responsible citizens who make a positive contribution to society.

All laudable aims, but rather beside the point when it comes to planning a programme of study

This claim that it is not possible to devise a rigorous curriculum for human wellbeing is, again, arbitrary. There may well have been woolly, ineffective and even misguided attempts to achieve this before 2010. But we can go back much further to the beginnings of the 20th century when anti-progressives were devising a rigorous curriculum to support human flourishing. William Bobbitt (1918), a staunch opponent of ‘progressives’, set out a vision for education in which the necessary facts to be learned (as argued for by Hirsch) could co-exist with the ambition to scientifically develop a curriculum to set up children to live fulfilling and prosperous lives:

New duties lie before us. And these require new methods, new materials, new vision. The old education, except as it conferred the tools of knowledge, was mainly devoted to filling the memory with facts. The new age is more in need of facts than the old; and of more facts; and it must find more effective methods of teaching them. But there are now other functions. Education is now to develop a type of wisdom that can grow only out of participation in the living experiences of men, and never out of mere memorization of verbal statements of facts. It must, therefore, train thought and judgment in connection with actual life-situations, a task distinctly different from the cloistral activities of the past. It is also to develop the good-will, the spirit of service, the social valuations, sympathies, and attitudes of mind necessary for effective group-action where specialization has created endless interdependency. It has the function of training every citizen, man or woman, not for knowledge about citizenship, but for proficiency in citizenship; not for knowledge about hygiene, but for proficiency in maintaining robust health; not for a mere knowledge of abstract science, but for proficiency in the use of ideas in the control of practical situations. Most of these are new tasks. In connection with each, much is now being done in all progressive school systems; but most of them yet are but partially developed. We have been developing knowledge, not function; the power to reproduce facts, rather than the powers to think and feel and will and act in vital relation to the world's life. Now we must look to these latter things as well.

Bobbitt sets out here in detail what should be obvious - that knowledge and skills (to use a slightly reductive term, but which has been used in the 21st century what Bobbitt is describing much more richly) can co-exist in a properly considered curriculum. Just because the skills in the curriculum may have become soft and flabby by 2009, doesn’t mean that the curriculum could not in a rigorous and ambitious way set students up for a life well-lived. Bobbitt prized discipline and rigour in staff and students. He simply recognized that education was for the benefit of children and conceptualised the aims of the curriculum accordingly.

In their ignorance and naivety, Gove and Gibb took one English lecturer’s polemic about his hobby horse and turned it into national education policy. Rather than judiciously considering the huge impact that educational policy has on the lives of all children, they went all in on Hirsch, which had had the added pleasure of sticking it to the ‘blob’. But if we understand Hirsch in relation to Bobbitt, rather than in relation to contemporary academics who champion discovery learning and soft skills, we can see that he is not so much standing up for education to have a substantive purpose but merely over-emphasising a particular aspect of the purpose of education at the expense of a broader understanding of socialisation, bildung, and preparation for work. He is more footnote than iconoclast.

Hirsch’s fantasy fully digested

A recent blogpost from David Didau is useful in understanding the current benighted intransigence in relation to knowledge and the purpose of schools. It is useful because it shows how Hirsch’s flimsy ideas are increasingly taken for granted as reality. Didau takes up the arguments of Hirsch and Young, and draws out further implications for how schools must be. He argues that since the purpose of education is to pass on our complex cultural inheritance, then schools should limit themselves to only teaching that which is ‘biologically secondary’ - culturally-formed elements such as maths and reading which we have not evolved to find easy to learn. The upshot of this is that school is unavoidably relentlessly difficult, and the purpose of school discipline is to ensure that children are constantly attending to the challenging work:

We easily fall prey to naive misconceptions and get frustrated at the tedious practice needed to master abstract secondary knowledge. This is why schools need rules and well-administered behaviour systems; without these things, children – especially teenagers – are likely to drift off into those activities which come more naturally and induce more pleasure, such as chatting, twanging rulers, looking at cat memes on the internet and trying to get off with each other. Rules, routines and adult authority are not embarrassing relics of a harsher age, they are the conditions under which difficult knowledge becomes available.

An old conservative such as William Bagley, who developed the notion of school discipline, would bristle at the limited purpose of school discipline set out here. He would agree that hard work requires focus, and that rules are necessary for this. But his vision of school discipline always, always, considers the effect of rules on the personal development of the children. By elevating knowledge to the highest purpose of the school, Didau here sidelines the development of the child. All that matters is that knowledge gets into head.

No doubt Didau would reply that the development of the child is important too. But in education policy and in his own work, it is always an afterthought. It is not be controversial or progressive to point out that an education which fills children’s head with knowledge but immiserates many is not a satisfactory education system. I am not advocating for any softer discipline. As a teacher and parent, I recognise that in all ways children require (at least the possibility of) strict discipline to become the people we want them to be. But this discipline needs purpose beyond academics. And our approach to thinking about this aspect has become impoverished beyond recognition compared with 100 years ago.

Schools exist in a complex world, and it is increasingly rash to assume that children are getting what they need outside of school. But our current stance on the purpose of schools is, in Bobbitt’s word, ‘cloistral’ in its hyper-focus on knowledge. It is also arbitrary - an English professor was nostalgic for a time when business executives quoted Shakespeare, a minister read his book, and now we act as though schools have one singular purpose at the almost complete expense of all others. This turn to rigid and monomaniacal scholasticism is not hard-headed, as its proponents like to imagine. It is a version of a ‘luxury belief’, where a singular thought about the world is imposed on it, thereby creating problems from which the elites responsible are protected.

And there are certainly problems: in England the education and allied systems are not adequately setting children up for careers or lives: a recent report showed 1 in 8 young people aged 16-24 are not in education, employment, or training (NEET). Young people who are NEET ‘believe the current curriculum does not prepare people for work (67%), in addition to over half feeling that the current education system does not suit people like them (55%).’ The problem here is not failing standards in core subjects, since ‘only 3% of employers recruiting 16-year-olds to their first job, and 1% recruiting FE leavers, identified poor literacy or numeracy as the main issue’. The problem is not that these children lack sufficient general knowledge; it’s that the education system has not been properly set up to support them.

And finally, this is not an argument for lowering standards. In the history of education before 1983 there have been many conservative figures, from Bobbitt and Bagley to Ronald Edmonds have produced a muscular vision for education that does not reduce it to the maximisation of knowledge. The turn to the maximisation of knowledge came about in the 1980s from the ‘new right’ who obsessed over harnessing schools to maximise economic productivity. We shouldn’t treat those that advocate for it now as the sensible ones. Hopefully a renewed Labour government in England will recognise this.

1. Abdolhosseinzadeh, M., & Abdolhamid, M. (2026). A comparative analysis of ‘development governance’ and ‘development management’ curricula in selected international universities. The Curriculum Journal, 00, 1–29. https://doi.org/10.1002/curj.70063

The following reference does not exist

• González, R. M., Román, A. V., & Hernández, A. (2020). GIS in higher education: A review of the literature. Journal of Geography in Higher Education, 44(3), 342–359.

The following reference does not appear to exist, butcannot be used to support “Berliner’s” model”:

• Berlin, B. M. (1987). The Neo-formalist model of curriculum analysis (Unpublished doctoral dissertation). Loyola University Chicago.

2. Ogunsakin, R., Okonu, D., & Nwokeocha, P. (2026). Beyond MOOCs: How technical and structural factors shape learner engagement, retention and inclusivity across online learning platforms. British Journal of Educational Technology, 00, 1–26. https://doi.org/10.1111/bjet.70066

The following references strongly appear to be fake:

• Dalipi, F., Imran, A. S., & Kastrati, Z. (2021). Integration of instructional design and learning analytics in a model for adaptive MOOCs. Computers & Education, 165, Article 104138. https://doi.org/10.1016/j.compedu.2021.104138

• Luo, H., Ding, Y., Hauff, C., & Willems, F. (2021). Sentiment analysis of MOOC learner reviews using deep learning. IEEE Transactions on Learning Technologies, 14(2), 231–243. https://doi.org/10.1109/TLT.2021.3061688

• Zhang, Y., Yu, R., & Zha, Z.-J. (2019). Sentiment analysis in education research: A review and future directions. Computers & Education, 137, 12–28. https://doi.org/10.1016/j.compedu.2019.04.002

These two retractions now indicate that, even after 19 retractions, leadership across BERA journals is incapable of stemming the tide of AI dreck being published in its journals. BERA, sooner or later, will need to face up to the fact that there have been serious editorial failures across its four journals. At the moment, it is trying to issue retractions while insisting that it is in the right and I am in the wrong. That is a completely absurd position that rests on falsely claiming credit for discovering things I publicly informed them of, misrepresenting COPE guidelines, and claiming, completely baselessly, that my methods are ‘unsound and ethically problematic’. All of these things are easy to disprove. BERA will have to adopt a more self-critical, and therefore defensible, position.

But BERA seems to think that because it is such a well-established and well-supported organisation, it can simply ride out my criticisms without taking them seriously. As with John Hattie, the problem they face is that I am right. I have demonstrated the fake references, the fake citations, the impossible methods and contradictory ethics positions. I have demonstrated that they have consistently published complete garbage. And slowly, the educational world is realising it. Before long, BERA will not be able to avoid scrutiny, and their weirdly triumphant current response will mean that they will have no position to retreat to. They will not be able to say lessons were learned if in response to beginning retractions they brazenly took the moral high ground while continuing to publish AI dreck.

BERA doesn’t realise it, but it is running out of time to save its reputation, and its remarkable (in the circumstances) complacency does not inspire confidence.

BERA have suggested this is both ‘methodologically unsound and ethically problematic’, for the following reasons:

However, disclosure is not required for purposes such as language polishing, conciseness, formatting and compliance, which can be of great use to many authors, including those who speak English as an additional language.

Interpreting features common in global academic writing as evidence of misconduct or AI generation risks implying that any deviation from Anglophone academic norms is inherently suspect, an assumption that is both methodologically unsound and ethically problematic.

Essentially, in a very academic and rather sanctimonious way, they are saying that the phrases I have found could well result from having an LLM polish the language or make it more concise, and that it is wrong to assume otherwise.

Fortunately, it is possible to interrogate the phrases (to use an irritating academic term) to shed light on which interpretation is correct: are the phrases indicative of LLM-generated text (not allowed) or are they indicative of LLM polishing and summarising existing author-written text (allowed). I believe the evidence is clear: time and again the phrases I have flagged communicate specific and complex ideas, which mean it is very unlikely that they are the result of text polishing or summarising.

Take the following example from a BERA article: the phrase “Second, the reliance on self-reported measures may introduce” was never used before 2023, but since Google Scholar shows 74 results. The phrase is a very specific one, indicating that:

• it follows from a first idea

AND

• it is discussing a certain type of research that leans on self-reported measures

AND

• that this may be a problem of some sort

It could be that 74 different authors have written a sentence containing these ideas, and the LLMs are merely polishing them into this specific format. It seems much more likely, though, particularly given the term ‘second’, that LLMs have a standard format for discussing limitations and this is commonly the second item on the agenda. In other words, it is reasonable to assume LLMs generated this text, rather than refining pre-existing sentences.

Let’s take another example: “underscores the intricate interplay between” was used 31 times before 2023, and 3,459 times since. Again it is a phrase communicating a very specific combination of ideas, suggesting an emphasis on the elaborate ways in which two or more things act in relation to each other. It is again possible, at least in theory, that thousands of scholars have written such a sentence, but were unhappy with its specific wording and so asked an LLM to polish it. It is much more likely that this is just a standard LLM phrase that appears when LLMs are used to generate text describing research.

Let’s take one final example: “comparative studies across diverse” (50 times pre-2023, 1,890 since). This phrase suggest a particular type of study used in a wide range of contexts. At the risk of repeating myself, it is possible that almost 2,000 scholars wrote this specific idea but were unhappy with the wording, and so had LLMs polish it, but it is much more likely that it is simply an LLM stock phrase that appears in generated text for a literature review.

There are many more similar examples you can explore in the tables in this post.

BERA’s feeble, misleading retraction statement and actions

BERA’s glib assertion that the use of phrases that exploded in popularity since 2023 is not a sign of LLM-generated text does not stand up to scrutiny. This post demonstrates that simply by examining the phrases themselves it is possible to deduce that it is far more likely that they result from LLM-generated texts.

I hope to have shown by now that BERA’s retraction statement and actions have been misleading, mistaken, and completely misconceived. They did not ‘uncover independently’ what I had already told them; undisclosed AI is not insufficient grounds for retraction; they have expressed confidence for some unsupportable papers; and no, it is not possible to dismiss my work - without any evidence at all - as ‘methodologically unsound’.

I do worry on BERA’s behalf. Soon, journalists will cover the story and BERA will look very foolish indeed. Getting ahead of the story means taking full stock of their mistakes, rather than attempting to discredit me.

10 unsupportable papers that BERA are taking no action on

The COPE guidelines state that a paper should be retracted if a ‘major error’ in the study which means that it is not possible to ‘have confidence in the results and conclusions reported in the paper’. BERA have investigated the following papers which I have argued were written using AI and found that they continue to have full confidence in the studies. Putting aside the issue of the evidence that they are AI-authored, there are a number of ‘major problems’ which the BERA investigation has ignored.

A. Incorrect citations/references

While the use of AI is permitted by Wiley (the publisher) in narrow ways, Wiley insists that ‘authors must take full responsibility for the accuracy of all content, and verify that all claims, citations, references, and analyses are aligned with their expertise and research’. The following two papers make errors in citations and references that mean that the claims made in the papers cannot be trusted.

Shazad et al., 2025

This ‘strategy’ (£) for determining sample size is problematic in its own right:

The sample size was determined using a strategy recommended by Roscoe et al. (1975), which states that more than 30 but fewer than 500 participants should be included in an empirical study.

It is for some reason repeated:

This sample size conforms to guidelines that suggest empirical research (Roscoe et al., 1975) should include more than 30 but less than 500 people.

But Roscoe et al (1975) is not the correct citation to support the claim. The 30-500 rule-of-thumb is suggested in this book:

• Roscoe J. T. Fundamental research statistics for the behavioral science. International series in decision process. 2nd ed. New York: Holt, Rinehart and Winston; 1975.

However, Shahzad et al. (2025) cited this article by a different Roscoe, which does not make the claim:

• Roscoe, A. M., Lang, D., & Sheth, J. N. (1975). Follow-up methods, questionnaire length, and market differences in mail surveys. Journal of Marketing, 39(2), 20. https://doi.org/10.2307/1250111

This is not a human error, and means that AI was generating the citations/references.

Gao, 2025

This paper on ethnic clustering (related to racial segregation) uses a reference on sound segregation (the ability to distinguish between sounds) to justify the following claim:

This spatial separation during formative years has implications for future intergroup relations, potentially affecting workplace diversity, residential choices and civic participation patterns in adulthood (Reardon, 2016; Reardon et al., 2019; Schwartz et al., 2012).

• Schwartz, A., McDermott, J. H., & Shinn-Cunningham, B. (2012). Spatial cues alone produce inaccurate sound segregation: The effect of interaural time differences. The Journal of the Acoustical Society of America, 132(1), 357–368.

This is not a human error - it’s clearly AI making a false association based on the term ‘segregation’, and the authors have not picked up on it.

Zheng et al., 2026

This paper (£) makes makes a number of errors in citing this paper:

• Harris, A., & Jones, M. S. (2017). Professional learning communities: A strategy for school and system improvement?. Wales Journal of Education, 19(1).

Not only does it turn it into two papers, it adds various Joneses to the list, while also providing incorrect volume and page numbers in the second version:

In citing them, it does not use Harris et al., but ‘Harris, Jones, Jones, & Jones, 2017’. Formatting citations and references through AI is permitted but the fact that the authors have not picked up on these egregious errors means that their paper cannot be trusted.

Chen et al., 2025

This article (£) does include citation errors, such as positioning a UNESCO education report within in a chemistry journal:

It also contains plagiarism. However, the dead giveaway that AI generated the text is this passage:

Specifically, the rose diagram, violin plot and... were all generated via the online website https://chiplot.online. This powerful webpage supports creating numerous concise and clear statistical charts with just one click.

Lauding a particular website in this way is simply not a sentence any academic would write, particularly the phrase ‘with just one click’.

B. Major errors related to methods

If there are ‘major errors’ in the methods, it is impossible to have confidence in the conclusions. The following papers have made such errors and should have been retracted:

Demir & Özdemir, 2025

The paper has various problems, with the most glaring being repeating the same paragraph three times in a row:

Furthermore, much of their methodology is redundant; it is not credible that researchers would actually follow the steps as they are set out. Inductive content analysis and thematic analysis are two very closely related methods, with many of the same steps (familiarisation with data, coding data). This study used both techniques on the same data, meaning they familiarised themselves twice with the data (impossible) and followed a near-identical process twice:

In the data analysis it appears that AI was provided with the figures and asked to analyse them. ‘From the map, it appears…’ is not language you would use if you had familiarised yourselves with the data twice.

Wang et al., 2025

This paper’s (£) description of where the participants were recruited from cannot possibly add up:

Since 15 x 8 = 120, it is impossible that each school contributed eight children. And if one school contributed 46, a number of schools would have contributed fewer than eight.

Liang et al., 2025

The methods for this research (£), took place during lockdown, seem logistically impossible several times over. In the study, ‘[a] total of 1015 college students completed a pen-and-paper questionnaire in online classes’. Why would you do a pen-and-paper questionnaire in an online class?

The participants completed all assessments under the guidence of professionally trained psychology instructors in a calm and controlled environment. The sequence of the scales… was chosen randomly.

If it’s an online class, and during lockdown, how is it possible to create this controlled environment? And it seems a headache to randomise the sequence of scales in a pen-and-paper questionnaire.

It could be at this stage you think that the study is just about possible. But the credibility of the paper is completely lost when it provides two completely different aims of the research:

So is the purpose to investigate the characteristics of online learners, or to validate the questionnaire?

Karataş et al., 2024

The paper has various problems including repeating sentences and light plagiarism, but in their over-the-top methods section they include a step that simply never happened:

Thick descriptions of contexts and experiences were provided in the presentation and interpretation of qualitative findings, supporting the transferability and grounding of the research (Boadu, 2021).

There was no thick description of any kind in the paper.

Emam et al., 2025

In this research (£), participation was voluntary, but there was a 100% questionnaire response rate from 320 teachers. How is that possible?

This study was conducted in Oman as part of a nationally funded strategic project to evaluate and promote IE practices in public schools. It focused on inclusive schools formally designated by the Ministry of Education to implement policies supporting students with disabilities and diverse learning needs. A purposive sampling technique was used to select study sites and participants. A total of 320 questionnaires were distributed to teachers in 37 inclusive elementary and middle schools across six governorates, selected for their inclusion status and geographical representation. Of these, 298 valid responses were analysed, yielding a 93% response rate. The remaining 7% were excluded due to excessive missing data or uniform responses, which suggested potential bias. The final sample included 298 teachers actively engaged in IE.

Either there was not a 100% response rate, or participation was compulsory. There is no plausible way both statements are correct.

Not AI-authored, just error-ridden

Bas & Xu, 2025

This paper does not appear to be AI written but has so many errors and citations that do not support the claim that it would likely fail any course and cannot be trusted For example, it claims without any support that ‘90% of students deliver homework to their teachers in an incomplete form’. It also makes this rather ridiculous sweeping statement from small-scale research conducted 25 years ago:

The Xu reference also does not support the claim. There are so many poorly supported claims that it is impossible to have any confidence in the paper.

Conclusion

It is an unavoidable fact that BERA has published a lot of low-quality AI-authored articles over the past few years. So far, it has only retracted some papers from one journal. Even after it completes the BERJ retractions (and some investigations are ongoing), it will still have three further journals to issue retractions for. It would do well to draw a line between the BERA of the past few years (which has published all this rubbish) and the BERA of the future, which will want to be credible. By making false claims crediting itself with uncovering the problem, completely misrepresenting the COPE guidelines, and taking no action on obviously unsupportable articles, BERA is failing to draw such a line. This will mean, sadly, that BERA will lose all credibility and will do so soon. As I have said again and again, this is not an outcome I want. BERA simply needs to act self-critically and competently to stop this happening. At this stage I am not optimistic.

So why then, is Greg Ashman again today proclaiming that the case consists merely of unproven and untested allegations? Having taken a position against me on the basis of one poorly worded tweet I made two years ago, Ashman has not noticed that the facts of the case have long passed him by. Having built up his career largely outside the university (I accept the limitations of the university in its current form, but the alternative seems little better), Ashman appears at times to be driven by the the imperatives of the attention economy to lead the educational culture wars rather than engaging in a genuine search for truth amongst a community of scholars. Take a recent attack on Oliver Caviglioli and Teaching Walkthrus, which consisted of almost no discernible content at all (read it, there’s nothing there) but, by taking on an illustrious rival edu-authority, Ashman did gain significant social media attention.

Now that he has again, today, mislead his audience about the substance of my allegations, I feel compelled to demonstrate how wrong Ashman has been in regards to my case against Hattie.

Ashman’s journey towards misrepresenting the Hattie case

1. Ashman overplayed his hand in making a minor correction

It all started with Ashman uncovering a poorly worded tweet of mine.

One of the first forms of plagiarism that I found John Hattie committing was a table from the management theorists Locke and Latham. Not only had Hattie copied it almost verbatim without attribution (what you might call the ‘really, really bad’ type of plagiarism, see below) but he compounded the error by adding up all the different studies and participants from the very obviously overlapping meta-analyses. Furthermore, most of the studies were not on students being taught by teachers or lecturers. As a result, Hattie falsely claimed that there were 21,117 students involved in the research

In the poorly worded tweet, I wrote:

The error here is saying ‘they certainly aren’t students’ because, as stated above, 5.7% of the participants were in fact students.

On the basis of this error on my part, Ashman has misleadingly claimed that ‘a few of Vainker’s points that I [Ashman] attempted to verify were factually wrong’. This is not correct - it was only one point, and ‘factually wrong’ is a spurious term in the circumstances. But since then, Ashman has said ‘I don't but [sic] into Vainker's personal attacks on Hattie’ (Twitter - can’t link as blocked).

2. Placing blame for Hattie’s absurd legal case on me

Hattie’s legal action is self-evidently hollow and strategic, aimed at preventing amplification of my claims. Indeed, Hattie has never publicly said what precisely in my work is incorrect or defamatory, despite the huge volume of my claims against him. He does not have any substantive grounds for legal action; he simply has the money to threaten it. Therefore it is ridiculous for Ashman to hold me responsible for the legal action, saying that on the basis of the above error there was ‘a certain inevitability’ of the legal case, and that:

For his part, Vainker is apparently unrepentant.

This is a sad and depressing episode. We should be able to disagree with each other about evidence and research in the strongest possible terms without it becoming personal. Too many education academics seem incapable of doing this.

That’s absurd. I’ve uncovered how the most influential education academic of the 21st century has plagiarised, made countless avoidable data errors, and included hallucinated references in his work, and Ashman calls it ‘becoming personal’.

3. Claim that allegations are entirely unconfirmed

So, to the present. In his podcast released earlier today Greg Ashman alongside his co-host Caiti Wade made a misleading statement, underplaying the status of my claims about John Hattie. He claimed that they are purely ‘allegations’:

WADE: So what this article speaks to is accusations from a UK academic who’s Stephen Vainker…, who’s basically gone out and said that well there’s things in Hattie’s work that are either plagiarised, there are things that are misrepresentations of research and there are things in his work that are complete AI hallucinations, so there’s research that he’s referenced that doesn’t really exist…And right now I want to be really clear and say this is an investigation that is going on so it’s not anything that’s confirmed, it’s Vainker has claimed this about Hattie’s work.

ASHMAN: They’re allegations, they’re not allegations that we are supporting or saying that we agree with, they’re allegations that he has levelled against Hattie and I believe that Hattie has embarked in a, he’s [audible excitement at this point] suing Vainker for defamation so we need to be very clear about that, we are not saying that we agree with these allegations.

No. It is a settled fact that in Hattie's work ‘there’s research that he’s referenced that doesn’t really exist’. Hattie even thanked me for finding it. Whether or not this constitutes an AI hallucination or not depends on whether you can find any substance in this rather ridiculous defence based on the existence of various ‘drafts’ (spoiler: there’s no substance):

Alex Pettifer, the editorial director at Elgar, confirmed the edit was made last month “after consulting with Professor Hattie who confirmed that an erroneous reference had appeared in their Foreward [sic] and was caused by the them [sic] not properly checking back against a source list of references after writing different drafts.”

It is also a matter of public record that the university found errors in the database (i.e. ‘misrepresentations’ above) and that Hattie did not cite sources properly:

A year later, in a letter we have seen that describes the results of the university’s assessment, Kate Smith-Miles, the university’s pro vice-chancellor of research capability, acknowledged to Vainker that the database contained “unintended” errors which “should be corrected,” and that Hattie could improve his citation practices.

So, again, far more than unconfirmed allegations.

4. Implication that I have only accused Hattie of less serious forms of plagiarism

Ashman then goes on to further mislead the public by very heavily implying that I have only accused Hattie of less serious forms of plagiarism:

One thing I wanted to talk about, I wanted to talk about plagiarism, because plagiarism sounds really bad, and not pre-judging this investigation or any legal proceedings or anything like that and certainly not saying that Hattie is guilty or innocent of any such allegation or claim but it’s worth noting that there are different kinds of plagiarism so probably the least significant…

He then goes on to talk about the less serious forms of plagiarism (e.g. using verbatim work but citing indirectly) and more serious forms of plagiarism (copying without any citation or reference). He continues:

…but I think when we talk about plagiarism, it sounds really, really bad, but I just wanted to point out that again, as with many things, it kind of exists on a spectrum and there are different kinds of plagiarism and this that and the other and when you get into academic writing and stuff like that you have to be conscious of these things because you often have to sign declarations and this that and the other.

The strong implication is that I have not accused Hattie of the more serious form of plagiarism - directly lifting people’s work without citing them at all. That’s not true, I can assure Ashman that a lot of the plagiarism I have shown Hattie committed was ‘really, really bad’.

Conclusion

In the scheme of things, Ashman’s uninformed interventions will have very little impact on what happens with the Hattie case. But they do help show us that we’re somewhat awkwardly positioned, if not stuck, between a rock and a hard place: a complacent, unfocused, and incompetent educational academia on the one side, and social media edu-celebrities on the other. It is all too easy to drift away from a focus on guiding education research towards integrity, and ultimately towards meaningfully improving schools.

In the case of the British Educational Research Association, they have long forgotten their ultimate purpose - to improve education in the UK. Instead, they publish abstruse, irrelevant research that only impacts internal academic metrics such as citation counts and diversity of authors. Without the focus on rigour and purpose, it’s no surprise that they have drifted towards publishing AI slop on a mass scale, and it’s no surprise that their reaction has been so self-serving, sanctimonious, and incompetent.

I’ve written about their absurd claim to have ‘uncovered independently’ that which I’d already told them, but the more serious story is that they have completely misrepresented the COPE guidelines and they have failed to retract articles that, according to the guidelines, should very clearly be retracted.

The following paragraph makes two very confident claims about the COPE retraction guidelines, both of which are entirely and unequivocally wrong:

In addition, undisclosed AI use does not always indicate fraud or manipulation; however, when combined with other signals, it can indicate that a paper needs further examination. This view is consistent with COPE retraction guidelines, which indicate that retraction decisions should focus on confidence in the publication’s results and conclusions, rather than on any one signal in isolation.  

The first claim is that according to the COPE guidelines, ‘undisclosed AI does not always indicate fraud or manipulation’; the second is that the undisclosed use of AI does not itself mean that a paper should be retracted. Both positions are completely false. The only mention of AI in the COPE retraction guidelines completely contradicts the above characterisation:

Editors can decide to retract a publication if they no longer have confidence in the results and conclusions reported in the paper. Reasons for this loss of confidence can be, but are not limited to: …Any form of misrepresentation, for example…undisclosed use of artificial intelligence.

So, in short, undisclosed use of AI is tantamount to fraud (‘misrepresentation’) and is itself sufficient grounds to retract the paper. BERA’s incompetence and complacency truly know no bounds.

I’m aware BERA also claim that almost all AI use it found is of the type that does not require disclosure. Again, they are completely and unequivocally wrong. Tomorrow I will share ten British Educational Research Journal papers that BERA has said it retains confidence in but contain huge errors related to contradictory or impossible methods, and various citation errors. It is a ridiculous that BERA is attempting to brush off this scandal of mass-publishing AI slop, and it cannot be allowed to do so. I repeat, we need education academics, but we can’t allow an incompetent and complacent (I’m also repeating the phrase, but it’s apposite) leadership class to disfigure educational research.

BERA has, finally, begun work to clean up its journals, announcing that 19 articles from its flagship journal, the British Educational Research Journal, have been retracted. I’ll write (responsibilities permitting) more about the retraction statement over the coming weeks, but I thought it was urgent to address three very strange assertions it makes in the process to denigrate me while promoting itself:

1. BERA ‘uncovered independently’ that a special issue was dodgy nine months after I told them

BERA claims that the 19 retractions it announced result from what they have ‘uncovered independently’:

During the course of the investigation, we uncovered independently that the peer review process for a limited number of articles in the British Educational Research Journal (BERJ) had been manipulated by parties that are not affiliated with the journal, Wiley or BERA. As a result of these findings, a total of 19 articles published in BERJ have been or are in the process of being retracted. The vast majority (17) of these articles were submitted as part of a guest-edited special section, which has now been cancelled.

This is quite the claim, given that in August 2025 I wrote a blog on how that very special issue was dodgy: that the guest editors publishing AI-authored work in the special issue and elsewhere, and most importantly that one of the guest editors had a suspiciously prolific record as a peer-reviewer:

A further interesting aspect is the sheer productivity of Namaziandost [a guest editor]. According to apparently confirmed data on his Orcid ID, he has peer reviewed 1,788 articles since 2020 (roughly 2,000 days). He has also already published 11 articles in 2025, having published 14 in 2024, and 12 in 2023. How, and perhaps more importantly, why, does he do it? Is there a good reason to peer review an article every single working day? I may be misinformed here but I don’t think that’s normal.

I also shared with BERA that many of the articles from the special issue appeared to be AI-authored.

So, the timeline is as follows:

• August 2025: I claim in a blog seen by BERA and Wiley that guest editors are publishing their own AI-authored work in the special issue and that one of the guest editors appears to be involved in nefarious peer review activities.

• Summer 2025: I provide BERA and Wiley with evidence that many of the articles in the special issue are likely AI-authored.

• April 2026: BERA and Wiley claim to have ‘uncovered independently’ that the special issue involved compromised peer review (see italics above).

However much BERA and Wiley find me and my criticisms distasteful, it cannot be denied that I alleged that the special issue was dodgy, and implied that there were reasons to suspect that one of the guest editors could not apparently be trusted to oversee a proper peer-review process. Whatever BERA and Wiley did with the evidence they found, they cannot claim to have ‘uncovered’ anything (I had already shone a light on it) or to have acted ‘independently’ (their investigation proceeded from my allegation). In reality-based language, with their greater access to evidence, they were able to confirm what I had alleged.

2. BERA taking action on around 40 papers I flagged, but claim that is ‘a small number’

Even more ridiculously, they claim that only ‘a small number of cases’ that I flagged warranted ‘formal editorial follow-up’. Well, I can reveal from personal correspondence that in total BERA has retracted, or is in the course of taking editorial action, on around 40 articles (that does not even include this one, which it appears to have missed). In what planet is 40 retractions (or perhaps in a small number of cases, corrections) a small number for four journals? Even if you take away the number that it is claiming sole credit for (see above) that still leaves around 25 retractions. That is not in any conceivable way a small number. Again, a self-serving, reality-denying assertion.

3. My approach is ‘ethically problematic’

Here’s the kicker. BERA did not merely claim, after I informed them about it, that they themselves ‘uncovered independently’ that the special issue was dodgy (giving me little or no credit). It then proceeded to call my approach to uncovering AI-authorship through detecting the use of stock phrases that came into (common) use post-2023 ‘ethically problematic’ on grounds anyone but a certain type of academic would find completely absurd: ‘[i]nterpreting features common in global academic writing as evidence of misconduct or AI generation risks implying that any deviation from Anglophone academic norms is inherently suspect.’ ‘Risks implying’ is doing a lot of heavy lifting here, as in ‘a sensible approach “risks implying” something distasteful and completely unconnected’.

BERA - failing to live up to its responsibilities.

This is not just a petty argument over who gets credit. It is an argument over the seriousness of education academia in the UK (BERA is after all the association of UK education academics). I’m a strong supporter of the idea that academia is the best steward of the future of education, but that is a role carrying a huge responsibility. BERA’s recent output, and its delusionally self-serving account of its retractions, unfortunately suggest that currently-existing education academia is no fit steward at all, and what’s more has no serious interest in becoming one. That’s a great shame, because less friendly critics than me are plotting its demise. It needs to sort itself out, and soon.

So I want to set out exactly what BERA needs to do first to reclaim their role. I have a sense that there is something of a crisis of leadership at BERA (a good thing in the circumstances) and that there is an opportunity to sort itself out and put itself back on the right course. There are of course an awful lot of sensible members of BERA (not served particularly well, I would suggest, by highly ideological and myopic leadership in recent years). But now that the president has announced that he has abruptly and prematurely left his post, amid some cryptic hints towards discord on his social media and a notable lack of acknowledgement of his contribution from BERA’s social media, there is the possibility of a radical change in direction, so that BERA may become once again relevant and incisive within the education debate.

Last week’s response by BERA to the Times Higher Ed story on my campaign reflected this lack of true leadership. Defensive and evasive, it wrongly asserted that undisclosed use of AI was not a problem in itself (this is contrary to Wiley’s guidelines). As such, they announced that in the majority of cases they would not be retracting the articles I flagged. In the short time since, they have managed to publish two further articles with fake references (see the first two entries in the appendix) so I’m not sure they are to be taken terribly seriously on this topic at the moment.

So, to provide some much needed clarity and direction, here are 48 papers that must be retracted: papers which were not just written by AI, but have manifest and serious failings as a result and cannot reasonably be viewed as serious human-authored papers. In addition to containing a preponderance of phrases that didn’t exist (or barely existed) before 2023 but have since come into common usage (clear indication of AI use, although apparently not dispositive), the 48 articles listed in the appendix contain various further problems, including:

• ‘hallucinated’ citations (18 cases)

• incorrect or highly strange citations (10 cases)

• impossible or contradictory ethics sections/statements (2 cases)

• use of we/our throughout despite being a single-authored paper (3 cases)

• incorrect citation/referencing formats (1 cases)

• plagiarism (1 case)

• unacknowledged severe conflict of interest (1 case)

• impossible/incoherent methods section (1 case)

• misattributing a quote (1 case)

• repeating paragraph (1 case)

• inserting editors’ names into a sentence (1 case)

• 8 line results section ( 1 case)

• misleading (4 cases)

• incoherent discussion (1 case)

Let’s look at some examples of what I’m talking about.

- incorrect or highly strange citations

Shot:

For instance, Thomas et al. (2016) emphasized the importance of cultivating a sense of belonging within distance learning. Their argument emphasizes that fostering student engagement and collaboration not only can alleviate anxiety but also nurtures the development of ideas and the forging of connections among students.

Chaser:

Thomas, B. S., Gupta, R. C., & Panicker, V. J. (2016). Recycling of waste tire rubber as aggregate in concrete: Durability-related performance. Journal of Cleaner Production, 112, 504–513.

Repeating paragraph (1 case)

How could this get through the most basic of peer review?

Contradictory ethics sections/statements

Was ethical approval secured? Seems a good idea to have your story straight:

These 48 papers are all listed in the appendix, with the full evidence for their retraction-worthy flaws provided. Now that BERA has new leadership in place, it would set the right tone retract the articles swiftly (they have been sitting on the worst examples for 8 months now). The tragedy is that obviously the AI-authored articles are not limited to those that can be completely proven. For example, look at these quotations supposedly from participants in a study but which are very obviously AI-authored (‘it wasn’t just about…, it was etc.):

It is much easier to simply run journals properly and ensure peer review is rigorous and fair than clean up the mess afterwards. However many papers it retracts, BERA will have at the very least a hundred more papers of the purest slop polluting the field and staining its reputation. This was all entirely avoidable - I’m still not sure what happened to the editorial process but I insist that editorial processes cannot collapse across four journals without some kind of direction from the top. But having entered the AI era with leadership and priorities that did not stand them in good stead to weather its storm, BERA can now set itself on the right course. That is, as long as it takes the unavoidable actions and learns the obvious lessons, however distasteful they may be.

Appendix

‘So how did it happen – I really know not’.

With these rather pompous words, John Hattie denied that he used AI to generate the foreword where a completely fake reference appeared. But this is not a mistake that Hattie has made once. Or twice. Or three times. In fact, fake references appear in four separate publications authored or co-authored by Hattie.

Here’s more of Hattie’s denials:

Hattie: Yes, I use Grammarly but I am reluctant to use AI given the usual reasons and Vainker’s pursuit.

A strange supposed reason not to use AI, but it’s just not true. He has also said:

'I use it every day...in fact I've used it twice this morning, there's a new tool out that actually writes the essay for me, and I've constructed one...'

In an article he co-authored where a fake reference appeared, the following acknowledgement was made:

The authors declare that Gen AI was used in the creation of this manuscript. Generative AI was used to assist with summarization, restructuring content, and formatting references.

So, when he says he doesn’t use AI, he is lying. Here is the evidence.

So, what is the University of Melbourne going to do about this? Well, it comes as a surprise to no one that they have simply declared that they will not investigate it. They have provided various grounds for this:

• with regards to the foreword discussed in the Retraction Watch article, they declared that because he didn’t include his affiliation with the University of Melbourne in the actual book, they weren’t obliged to investigate it.

• In the case of the Garay Abad & Hattie (2025) and Blannin, Stubbs, Wood, & Hattie (2024) (where Hattie is one of only two listed writers) co-authored papers, the university said that it was ‘unclear’ whether Hattie was ‘responsible’ for the issues, so again they weren’t obliged to investigate it.

• In the fourth instance - the Hattie J., O’Leary, Hattie K., & Donoghue (2024) book - the university simply said after a couple of weeks the investigation has been finalised, gave me no details, and said it would ‘not enter into further correspondence in relation to this matter’. So, basically, it found some reason not to investigate, but it’s not saying why.

So, Hattie’s on record saying he uses AI, fake references have appeared in four publications that appeared between 2024 and April 2025 where he was either the sole author, the lead author, or one of only two authors. But, for the University of Melbourne there’s not a thing they can do to look into it.

I repeat, the evidence for Hattie as a serial plagiarist on an epic scale and as a user of AI to produce his works is obvious and overwhelming. The university that employs him and his wife and has so much to lose from his loss of credibility will never make a meaningful finding against him. It is up to journalists to investigate the scandal, which now, in my view, encompasses the University of Melbourne no less than its superstar researcher.

To make this argument, I want to look at a widely shared and admired blogpost on school culture and character development from a year ago by Pritesh Raichura. Raichura is an inspirational teacher who achieves incredible results. He also seems to be a delightful person who I’m sure creates a wonderful school culture where I’d feel lucky to send my own children. But this does not mean that he can necessarily write about school culture and moral development well. To do so (in my view) he would need to spend considerable time looking into the history of school culture, reading about character development, and then considering how the two are related. I can entirely understand why he wouldn’t want to do that, but it tells in his writing.

In his blog, he sets out how his approach to school culture breeds character. It is trite and complacent, setting out, through a fake Aristotle quote, a pseudo-rationale for his approach to school culture in complete ignorance that the true, quite different, rationale for it was set out over forty years ago (see below). The blog is very short and very clear so please read it, but in brief his argument is

1. It is better to be more polite; to take more responsibility for your actions; to be grateful rather than entitled; to work hard rather than be lazy & to think about others rather than just yourself.

2. Some people are more polite than others; some people are better at taking responsibility for their actions than others; some people are more grateful than others; some people work harder than others; some people are better than others at thinking about how their actions affect those around them.

3. It is possible for individuals to become more polite, more responsible, more grateful, more hard-working and more dutiful.

And so:

If you agree with these premises, then the path to helping pupils becoming better people is illuminated: we need to have granular clarity about the behaviours we want to make habitual in our pupils. The act of repeating these behaviours will help our pupils become better people.

And the habits he sets out to instil are as follows:

In short, the school sets out exactly how the pupils should behave, so that the pupils become better people.

This is a post-hoc rationalisation of the way schools like his operate. The model for this version of school culture came from Lemov in the 2010s, who cribbed it from KIPP in the 2000s, who were heirs to the ‘academic press’ of Joseph Murphy and colleague 1980s. And all of them were inspired to create schools in the images of businesses, so they all looked to 1980s descriptions of corporate culture which were inspired by how post-war Japanese car factories operated. The same basic principle was passed down: if you want to control an employee/student then controlling their behaviour will only get you so far; you also have to get them to ‘want’ to work hard, and so you need to manage their ‘insides’ - their thoughts, values, feelings; in short - you have to shape their identity to the demands of work.

If we look again at the table above, you can see this makes perfect sense. The school culture is managing the values alongside the behaviours to ensure that every pupil works as hard as possible. So the child comes to want to do the behaviours because they feel like a better person; not being bored in class means that you are a grateful person, with the implication that being bored in class makes you ungrateful! And so on with every aspect of school life: complying makes me a good person; not complying doesn’t just mean I will get in trouble, it makes me a bad person. And anyway I shouldn’t just comply, I should go above and beyond - that makes me a really good person.

Whether this approach is a good thing or not we can disagree about, but it’s absolutely not simply about ‘making them better people’. That rationale is obviously trite, and if we’re going to bring Aristotle into it, it is very questionable about whether it does make them better people since it leaves them by design no space for the development of phronesis: having ‘granular clarity about the behaviours we want to make habitual in our pupils’ and getting them to repeat the behaviours ad nauseam means no space for practical judgement. You need to know when to push yourself, when to ease off, when to really make an effort with someone. Indeed, some of the very worst people I knew at school were Raichura’s model students- they would thank the teachers, clean their tables, do their work well, but were also vain, thin-skinned, humourless and entitled. They did these things because they wanted to put themselves on the level of the adults, above the rest of us. And they were disliked; we saw that they lacked character. And we all knew the opposite - kids who could be rude, late, even unpleasant to teachers, but were basically good people who you could be good friends with.

Raichura’s culture is not about making better people, it’s about making compliant strivers. Instilling such an attitude is good from the perspective of P8, and the attitude might well increase their life chances, but it is entirely questionable when it comes to the development of character.

The post is typical of high-profile teacher-leaders who pronounce about how things should be done from deep inside the bubble. Complacent, prickly, incurious beyond CogSci summaries of summaries and Substacks by peers; possessing far less depth than they realise. The field of education won’t advance with them at the helm. It’s a shame academia has made itself so useless on so many levels.

Let’s start with this figure presenting the key themes:

I’m not sure how readable it is for you but let me go through the themes for ‘perceptions and attitudes’: we have ‘with’, ‘though’, ‘about’, ‘toward’, ‘expressed’, and ‘some’. For limitations we have ‘data,’, ‘data;’, and…’data’, as well as ‘sample,’ and ‘sample’ and ‘bias,’ and ‘bias’. For key findings we simply have ‘showed’, ‘chatbot’, ‘lack’, and finally ‘high’. Great stuff. This is how this dross is spun by the article:

Figure 5 presents the latent theme network visualization, revealing the interconnected nature of GAI integration challenges and opportunities. Central nodes—‘student’, ‘concerns’, ‘learning’ and ‘engagement’—form a dense co-occurrence web, with lexical bridges including ‘usefulness’, ‘policy’ and ‘writing’ connecting attitudinal domains with pedagogical impacts.

A classic AI article - a mix of the highfalutin and purest dross.

There are various further problems, including a fake quotation - ‘blurred continuum between legitimate assistance and academic misconduct’ does not appear in the article cited. More seriously, the central claims in the abstract are not supported in the results. From what I can tell, the only support for the claim (in this systematic review) that 73% of students use AI comes from a single study with…36 students. Meanwhile, I cannot work out where the following figure provided in the abstract comes from: ‘institutional investments in faculty development yield a negligible correlation (r = 0.12) with pedagogical transformation’. It does not appear in the results section, but reappears in the discussion (and introduction, for some reason). It is also makes a wild claim of ‘differential national strategies shaped by educational traditions, regulatory frameworks and technological infrastructure’ based on… 1/8 Thai papers covering curriculum integration as opposed to 0/7 UK papers.

Finally, although the paper was about attitudes to AI, the methods section contained the following sentence: ‘This approach allowed for the extraction of recurring themes and interdependencies among risks (Vaid et al., 2025, p. 9), adapted to our GAI context’. The phrase ‘interdependencies among risks' was used because the cited paper was titled ‘exploring environmental, social, and governance (ESG) risks in Indian family businesses’ (it just had the same basic method). The AI has slipped, which is typical - it is not great at distinguishing between the finer points of meaning, which is what scholarship is so often about.

This paper is the kind of garbage that I am campaigning against - absolute slop, absolute nonsense. If BERA had maintained any kind of editorial standards this would have been flatly rejected. It is the lowest effort AI production that a cursory read would have discounted (it also provides endless page numbers for indirect citations). As I say, the silence of other academics about this tells me that people simply don’t read BERA journals or have become incredibly pusillanimous about speaking out.

They make this flatly wrong comment about the seriousness of undisclosed AI:

“The Bera journals follow Wiley’s AI guidelines, which state that certain uses of AI require disclosure but not in all circumstances. Undisclosed AI use does not always indicate fraud or manipulation; however, when combined with other signals, it can indicate that a paper needs further examination.”

No. BERA are saying here that undisclosed AI use is not an issue in itself. This is false. The Wiley guidelines state that disclosure is necessary when ‘AI generates or substantially rephrases text for the manuscript’ - the type of AI use I have shown. BERA cannot simply say that it is only necessary to look into a paper ‘when combined with other signals’. Undisclosed AI use of this type is a problem in itself for various reasons - including basic transparency - but the main issue is that in writing an academic article AI produces what sounds plausible rather than what is necessarily truly useful and incisive. It regurgitates the same generic talking points over and over, applying it to whatever the question is at hand. It produces mere filler. To say that this is basically fine is shocking. And what are the point of guidelines if authors can flaunt them with impunity?

It is on these false grounds (that undisclosed AI is not a problem in itself) that they found that in most cases there is no issue:

“some of these concerns are being investigated further and the outcomes of those investigations are pending…“for a majority of the papers flagged, Wiley’s research integrity team did not find any reason to pursue further action”, they said.

This is also an extraordinarily complacent framing of the matter. At the time they made the comments, I had shown that 15 articles in their journals contained fake references (I have since found two more) and many more with serious issues such as grossly misleading ethics sections, repeated sections of texts, and undisclosed conflicts of interest. When your journals have published at the very least 15 undeniably AI-fabricated articles, it takes some chutzpah to present the issue in terms of most of the articles being fine (especially when it’s not even true - see above). Faced with the evidence of the 15 undeniably AI-fabricated articles, BERA must acknowledge that it has a serious problem, and take it from there.

Hattie issued a legal threat against me in December 2024, and registered a defamation case in the Supreme Court of Victoria in May 2025. However, despite me making no concessions of any kind, he has never actually pursued the claim. The reason for this is he has absolutely no basis to claim I defamed him. Everything I have claimed, I have clearly evidenced:

• 9 fake references across four publications

• well over 500 instances of plagiarism, from 1981 to 2025

• 212 entirely avoidable data errors

His defence against these claims:

• On the fake references: ‘So how did it happen – I really know not’

• On the plagiarism: he wanted to avoid ‘over-citing’. So here, for example, he presumably copied 157 words without any attribution at all for the most impeccable of reasons.

• The data errors were apparently ‘unintended’. My point has never been that they were intentional, it was that they were reckless, which amounts to misconduct.

So, Hattie has no grounds for any legal action, but I would imagine that even the threat of becoming involved in legal action, even obviously unloseable, would deter media organisations. This is a shame because it is not possible to simply wait on the University of Melbourne investigation for the official rubber stamp on Hattie’s misconduct. The university has already proven beyond a reasonable doubt that it is determined to not find Hattie guilty of misconduct:

• It only investigated a ‘random sample’ of 14 out of the 141 instances of plagiarism I provided, and would not show me which 14 it deigned to review (this is important because some were much more serious than others).

• It refused to investigate instances of fake references by simply asserting that it had the right to refer complaints to other bodies if it chose to. After I pointed out that it had referred the complaint to an accomplice of the misconduct (the publisher who undertook a stealth edit) the university simply said it would not discuss the matter further.

• The university is refusing to give any timeframe on when it will complete the formal investigation into the instances where he copied the citations from works alongside the text. Given that there are only 11 instances, it makes no sense that it would take the 7 months that it already has, let alone even longer.

So, it is only journalists who can hold Hattie to account, and to force the educational establishment into a degree of reflection on how it gave such prominence to such an obvious chancer. It is also important to hold to account those publishers, institutions, and podcasts who continue to provide Hattie a platform without mentioning the investigation; they are providing the false impression that Hattie remains a respectable figure. Once his dismal legalistic reaction to the misconduct allegations is held up to scrutiny, it would disappear into dust; it simply has no substance at all.

• A paper on the identity construction of German accountants in the 1970s (Goretzki et al., 2021)

• A paper on how metaphors shape our attitudes to Covid-19 (Burnette et al., 2022)

• A paper on how metaphors shape our attitudes to mental enhancement (Conrad et al., 2019)

On one level these papers are very close to the subject of metaphorical design in video-based instruction (they are about metaphors, and about how we understand the world). But in fact they are completely irrelevant. It is not just about that they are not related to video-based instruction; one is about how we understand ourselves, while a further two are about how we understand particular events or actions. Importantly, none are about how metaphors help us learn abstract concepts.

It is fascinating to see how the LLM forces the papers into its article, in the process not only completely mangling the meaning of the cited papers but also producing incoherent drivel. For instance, in citing the paper on how the metaphors we use to understand Covid-19 shaped our attitudes to it, the paper writes:

According to Conceptual Metaphor Theory, metaphor is not merely rhetorical but cognitive, enabling learners to understand abstract concepts through concrete, familiar experiences (Lakoff & Johnson, 1980). This process requires learners to construct meaning by mapping similarities across domains (Burnette et al., 2022). For example, abstract motivational states are often understood through metaphors grounded in shared social experience, such as describing effort as ‘having more energy’ or fatigue as ‘losing energy’ in everyday interactions, thereby linking abstract motivation to experiences that learners commonly recognize from social life.

This is nonsensical – quite apart from the Burnette reference being completely out of place (it has not bearing on how we understand abstract concepts – Covid-19 was a real disease) no one describes effort as ‘having more energy’ and fatigue is not an ‘abstract motivational state’. Here, the LLM is presumably aware that fatigue is related to Covid-19 and so it has tried to shoehorn in a metaphor about fatigue into its explanation.

Here is how the articles on the identity formation of German accountants in the 1970s and how metaphors shape our attitude to mental enhancement are used to explain the use of metaphors in video-based instruction:

In video-based learning, metaphorical design can facilitate learners’ interpretation of materials (Shanshan et al., 2022), support comprehension (Goretzki et al., 2021) and enhance cognitive depth and memory retention (Boers, 2000; Conrad et al., 2019).

In parallel, metaphorical design on content reduces psychological distance by grounding abstract ideas in concrete, familiar experiences, thereby enhancing comprehension (Goretzki et al., 2021; Lakoff & Johnson, 1980).

In relation to Boers, it has misunderstood a paper on how metaphors shape our attitudes to cognitive enhancement, taking it to mean metaphors produce cognitive enhancement. On a purely linguistic level, these concepts might be closely related, but in reality they are completely distinct. It has also got it into its head that Goretzki supports the idea that metaphors support comprehension – except it is comprehension of ourselves, not some abstract concept. Again, so close and yet so far.

And here is another example of how the LLM jams together the content of the article it is producing and the articles it is citing:

By grounding abstract concepts in familiar, concrete experiences (Burnette et al., 2022; Goretzki et al., 2021), metaphors not only support learners’ comprehension but may also influence how PAs are perceived, enhancing their perceived competence, relevance and social presence.

Again, the references are completely out of place (they are not related to grounding the abstract in the concrete). By picking up on fragments from the cited articles (i.e. metaphors can shape how we understand people and things), the LLM is producing a completely hallucinated point – that if pedagogical agents use metaphors, the learner will warm to them. Clearly, the original articles do not say anything like ‘people like people who use metaphors’. It’s about two steps removed from their actual meaning.

BERA is now apparently screening articles for fake references; none have appeared in a while. However, it is still regularly producing LLM-written slop. It is inconceivable that a human could have made these errors. If people are sharp enough to write the paper, they would have been more than sharp enough to realise that the references they were using were completely irrelevant.