Notebook or My Personal Learning Network

A journal that received 160 submissions in 2022 expects more than 1,000 this year. Its own editors say they will soon lack the resources to keep up. AI is to blame.

This comes from an editorial published on 25 September in the journal Accountability in Research.

The numbers around it are arguably even worse than the surge itself. A study of 7.3 million journal articles found that sentences showing GenAI influence rose from 12% in 2023 to 57% in 2025. One estimate puts 140,000 hallucinated citations in articles on preprint servers and scholarly indexes in 2025 alone. It is also likely that more than half of scientists now use GenAI to write peer review reports, despite most journals prohibiting this.

The editors are blunt about enforcement. They state that AI detection tools remain unreliable, and that even the watermarking Anthropic introduced earlier this year cannot always label AI-generated or AI-assisted text.

So, the editors changed the system.

They have capped each author at four submissions a year to make the journal a harder target for paper mills, made ORCID numbers a requirement for all authors, demanded underlying data for empirical papers, and banned AI in commentaries and letters to the editor.

Have they gone far enough?

Read the editorial in full here: https://lnkd.in/eU5GsVnv

#ResearchIntegrity #ScholarlyPublishing #PeerReview #AcademicPublishing #AIinResearch #CriticalAILiteracy #HigherEducation #OpenScience #AIEthics #Academia | 25 comments on LinkedIn

Le Crédit d'Impot Recherche est dépourvu de ciblage et de stratégie et a un effet de levier négatif sur l'investissement en R & D en France, quand la Belgique a un système d'aides focalisées, avec un effet de levier x4. Quatre euros privés investis pour un euro d argent public.

Cette niche fiscale désastreuse est l'un des plus gros facteurs de déclin économique et technologique. Les hauts fonctionnaires de Bercy, analphabètes scientifiquement, en font à nouveau la plus grosse niche fiscale du Projet de Loi de Finances.


Il est temps de réorienter le dispositif pour répondre aux besoins de la société, par une politique pensée, ciblée et évaluée.
https://lnkd.in/eg849fPx | 19 comments on LinkedIn

Academics love to moan about publishing, but for good reason. If you're not in academia, please read on and marvel at how we supposedly bright sparks have been sucked into one of the most bizarre economic systems of the modern world.

Take Elsevier, one of the biggest publishers with many journals across disciplines. Its parent company, RELX, reported £1.035 billion in adjusted operating profit in 2025, on £2.714bn revenue: an insane margin of about 38%. RELX’s largest disclosed shareholders are BlackRock (9.67%) and Invesco (c.5%); not companies we typically associate with the pursuit of knowledge and truth.

Those margins are only achievable because most of the underlying labour comes from academia. Researchers conduct the research, largely funded by universities, governments and charities. We write the papers. Other academics peer-review them, generally for free (peer review is vital for good science - unbiased, constructive feedback). Academics also provide much of the editorial expertise, often unpaid or modestly remunerated. After all that, publishing a paper still takes many months (often over a year). Publishers provide infrastructure, production, hosting and editorial systems.

THEN - here's the really good part - universities BUY access to that literature back, or pay article-processing charges so everyone else can read it. For Elsevier alone, most UK universities spend hundreds of thousands of pounds a year; large, research-intensive institutions often spend £1m+. This against a backdrop of 35.8% of English higher-education providers reporting a deficit and 15,000 job cuts in a year - University and College Union (UCU).

So academics simultaneously supply much of the product, much of the quality control, and are the customer. Cool.

In a fun plot twist, scholarly content and data are becoming valuable inputs for AI (of course). Major publishers are licensing content/data for AI applications and building new, paid AI products around scholarly literature. Good old AI is also being used to write articles (complete with AI hallucinations and outright plagiarism) - making even more work for those unpaid editors and reviewers.

Surely there is a way for universities to use their existing infrastructure and keep more of the money that currently leaves the sector inside research itself?! Other models include university and learned-society publishing, diamond open access where neither reader nor author pays, consortial funding, Subscribe-to-Open, institutional repositories, and open science platforms. It takes a brave researcher, especially in the early years, to ditch the prestige that comes with a big name 'high impact' journal...and a lot of research to find these ethical routes.

Hit me up with viable solutions, or publishers you trust.

Also, are there industries out there with economic models more bonkers than this?! | 41 comments on LinkedIn

Advancing medical knowledge from research to practice

L'engagement est le vrai angle mort des démarches de veille : on investit dans l'outil, les sources, le paramétrage… et on oublie la question la plus décisive : comment faire vivre la démarche dans le temps ?

Résultat : beaucoup de lecteurs passifs, quelques contributeurs, et une dynamique qui repose sur deux ou trois personnes. Ce n'est pas un problème d'outil mais un enjeu d'engagement et ça, ça se travaille !

RDV mercredi 7 octobre 2026, pour un webinaire de 45 minutes, à l'occasion de la sortie de notre livre blanc « Engager au sein des démarches de veille ». Pas de recettes magiques mais des grilles de lecture issues du terrain et des sciences sociales, et des retours d'expérience de démarches qu'on a accompagnées 😉

Inscription (gratuite) : https://lnkd.in/eqRFHJVv

#Veille #IntelligenceÉconomique #Engagement #VeilleCollaborative

"The Death of the Academic Author? Implications of Artificial Intelligence for Academic Writing and Publishing."

That was the title of a symposium I had the chance to attend at the AMEE - The International Association for Health Professions Education conference in August. Jennifer Cleland moderated, and the panel was Lorelei Lingard, Erik Driessen, Yu-Che Chang, Ayelet Kuper, and Ken Masters.

A title like that plus this line-up of speakers? Sign me up!

You would anticipate excellent debate, and it lived up to the billing. I had a lot of thoughts immediately after the session, but I finally had a chance to sit down and write/think/dictate my reflections.

Wait! Write/think/dictate?!? What does that mean? You'll just have to read it to find out! 😉
The link is in the first comment.

Then let me know your thoughts on writing as thinking versus writing as writing in the comments. | 14 comments on LinkedIn

Les premiers résultats de l'enquête par questionnaire pilotée par le Gis Marsouin sur les usages de l'IAGen par les étudiants, obtenus sur un échantillon de 2 276 étudiants, viennent nourrir nos réflexions en matière de recherche mais également d'enseignement.

- Un usage largement répondu
85% des étudiant·es interrogé·es l’utilisent dans le cadre de leurs études.

- Pour réviser et réaliser des "devoirs maison"
75% l’utilisent pour réviser des examens, plus de 60% pour préparer un TD ou un TP ou réviser un cours magistral, et plus d’un tiers lors d’un examen réalisé à domicile.

- Peu discuté avec les enseignants
Malgré un fort taux d'usage, plus de la moitié des étudiant·es déclarent que l’usage de l’IAG n’a pas été discuté avec leurs enseignant·es et seul·es 18% rapportent une autorisation explicite.

- Confiance et fiabilité limitées mais peu de vérification
D'un côté, les étudiant·es interrogé·es portent un regard globalement prudent sur les productions de l’IAG : seul·es 16% les jugent fiables et 11% considèrent les informations fournies comme exactes. De l'autre, cette faible confiance ne se traduit pas systématiquement par un contrôle des réponses : 36% vérifient systématiquement les sources.

- Un compagnon virtuel ?
En dehors des usages académique, l'enquête explore également une pratique spécifique : le recours à l’IAG comme « compagnon virtuel », pour discuter, obtenir des conseils ou rechercher un soutien émotionnel. Encore minoritaire, elle concerne tout de même 17% des étudiant·es et est davantage associée à la solitude et à l’anxiété déclarées.

Le questionnaire a été construit par Agnès Grimault-Leprince, Bruno Daucé, Marianne Lumeau, Laurent Mell et notre regrettée collègue Séverine Erhel.
Les données ont été traitées par @Pierre Le Bras, qui a rédigé la note complète à retrouver sur le site de Marsouin : https://lnkd.in/e6hiPR7S

Reading 100 research papers is not the same as understanding 100 research papers.
The real challenge in a literature review is turning a large collection of papers into clear themes, relationships, contradictions and research gaps.

This framework gives you 20 practical prompts to help analyse 100 papers and move from scattered information to structured research themes.

You can use these prompts to:
• Extract key information from papers
• Identify major topics and recurring themes
• Compare findings and research methods
• Find contradictions and research gaps
• Map relationships between themes
• Identify emerging trends
• Build a thematic table
• Develop a literature review structure
• Connect themes to your research topic
• Refine and validate the final themes

AI can help you process and organise information, but the interpretation and final research judgment should remain yours.

#ResearchMethodology #AcademicResearch #ResearchTools #ResearchGap #SystematicReview #Researchers | 10 comments on LinkedIn

London School of Economics researchers say Gen Alpha will talk, not type at work—they predict that voice notes will make keyboard practically obsolete by 2030.

Is up to 80% of the content of the research literature already available in an LLM?

In 1950 Norbert Wiener worried less about machines doing the work of people than about people becoming components in processes they had neither designed nor could control. My new paper, The machine use of human beings, turns his title round to ask how AI systems come to know research that sits behind a paywall.

A machine that may lawfully read only open-access papers still learns a good deal about subscription work, because the researchers who read it go on to cite and restate it in open articles. We call this human mining. Across 124 million articles in Dimensions and Unpaywall (https://unpaywall.org/), three quarters of the subscription articles that have ever been cited have been cited from the open literature. The median wait for that first open citation has fallen from thirteen years for work published in 1990 to one year for work from 2020. That change comes from the growth of open publishing, not from any change in how we cite.

What crosses over is selective, though. It carries single results rather than the method, the caveat or the approach that failed. It favours the recent and the much-cited, it can pick up errors along the way, and the researchers acting as the machines' sensors get no credit for it.
The chart splits each year's output into open, subscription-but-cited-openly, cited only behind the paywall, and uncited. The black line is roughly the most an open-only machine reader can see.

https://lnkd.in/eBpC595H

L'intégrité de la recherche est fondamentale pour la confiance dans la production scientifique, mais les systèmes qui la protègent sont soumis à une pression croissante. Des défis de longue date, tels que la fraude, la manipulation et les pratiques de recherche problématiques, se conjuguent désormais avec l'augmentation du volume des soumissions, l'évolution des incitations et les progrès rapides de l'intelligence artificielle. Ces pressions affectent différemment les différents maillons de l'écosystème de la communication scientifique, soulevant des questions importantes quant à l'origine des risques, à l'efficacité de leur détection et à la capacité des politiques, pratiques et ressources actuelles à suivre le rythme.

L'intéressant sondage #PulseCheck de la Society for Scholarly Publishing examine comment les organisations de l’écosystème de l’édition scientifique réagissent aux défis actuels en matière d’intégrité de la recherche. Les résultats offrent un aperçu opportun des priorités actuelles, de la confiance de la communauté et des causes sous-jacentes.

👉 https://lnkd.in/eexVmJ6g

Quelques points saillants :

🎯 les manuscrits générés ou manipulés par l’IA constituent le principal défi, et les inquiétudes l’emportent sur la confiance dans la détection ;
🎯 bien que 94 % des organisations aient mis en place au moins une mesure d’intégrité, 63 % d’entre elles ont peu confiance dans la capacité de l’écosystème à protéger efficacement le patrimoine scientifique ;
🎯 l’IA est à la fois un défi et un outil. Plus de la moitié ont recours à un filtrage assisté par l’IA, mais la communauté reste divisée quant à savoir si l’IA améliore la détection et la prévention des atteintes à l’intégrité ;
🎯 la place qu’occupent les organisations au sein de l’écosystème détermine leur perception. Les éditeurs mettent l’accent sur les menaces pesant sur le processus de publication, tandis que les bailleurs de fonds, les institutions et les bibliothèques accordent une importance relativement plus grande aux pratiques de recherche et aux défis systémiques.
🎯 La pression liée au « publish or perish » est considérée comme la principale cause profonde, les personnes interrogées soulignant comment cette pression, l’IA, le volume des soumissions, la fraude et les contraintes de ressources se renforcent mutuellement.

Que pèsent vraiment les think tanks sur la fabrique de la décision publique ?

Le 10 mars 2025, les députés adoptent un amendement sur les soins relationnels dans la profession infirmière. Ses auteurs citent les recommandations de l'Institut Sapiens et d'organisations professionnelles. Quelques mois plus tard, la notion figure dans la loi du 27 juin. Qui a fait avancer cette idée ? L'institut qui a formulé des recommandations, les organisations infirmières, les élus ? Le texte permet de suivre une contribution partagée, sans désigner un inventeur unique.

Ce cas éclaire la question que pose « Ce que le Parlement retient des think tanks », l'étude que je publie aujourd'hui pour Suasoria et qui inaugure l'Observatoire de l'influence intellectuelle : que deviennent les travaux des think tanks une fois publiés ? Un rapport peut inspirer une réforme, fournir un chiffre à ses adversaires ou rester une référence longtemps après sa parution. Pour comprendre le rôle de ces organismes, il faut regarder ce que les responsables publics font de leurs travaux.

J'ai suivi, de janvier 2017 à septembre 2026, les traces laissées par 36 think tanks et organismes assimilés, et par quatre organismes de recherche pris comme points de comparaison, dans les travaux de l'Assemblée nationale, du Sénat et dans les rapports officiels. L'étude s'en tient à ce que les textes permettent d'établir ; elle ne reconstitue pas les conversations qui précèdent une décision.

Premier constat : il n'y a pas un classement, mais plusieurs. Depuis les législatives de 2024, l'Institut Montaigne reste le think tank le plus cité au Parlement (110 citations), devant la Fondation Jean-Jaurès -Jaurès et I4CE - Institut de l'économie pour le climat / Institute for Climate Economics , à égalité (96). Mais I4CE - Institut de l'économie pour le climat / Institute for Climate Economics , l'Institut de l'économie pour le climat, passe en tête dans les rapports officiels, et Jean-Jaurès dans les propositions de loi. Quant à l'OFCE et à l'Institut des Politiques Publiques , ils devancent tous les think tanks dans les rapports officiels.

Quel est le think tank le plus influent de France ? La réponse change selon le texte que l'on ouvre.

La synthèse et l'atlas des 40 organismes sont en ligne : https://lnkd.in/e66J-m6f | 24 comments on LinkedIn

🔬 AI Tools for Research & Review Article Writing

AI can significantly accelerate research, but the researcher must remain responsible for scientific judgment, originality, accuracy, and integrity.

A practical workflow is

🔹 Research Question & Planning: ChatGPT, Claude, Gemini, Perplexity
🔹 Literature Discovery: Scopus, Web of Science, Google Scholar, Semantic Scholar, OpenAlex
🔹 Related Paper Discovery: ResearchRabbit, Connected Papers, Litmaps, Lens
🔹 Literature Screening & Synthesis: Elicit, Consensus, Scite, SciSpace
🔹 Reference Management: Zotero, Mendeley, EndNote
🔹 Bibliometric Analysis: VOSviewer, Bibliometrix/Biblioshiny, R, Python
🔹 Article Drafting & Refinement: ChatGPT, Claude, Gemini, Copilot
🔹 Academic Language Editing: Grammarly, Paperpal, Writefull, LanguageTool
🔹 Figures & Visualization: BioRender, Canva, PowerPoint, Python, R
🔹 Citation Verification: Crossref, PubMed, Google Scholar, Scite
🔹 Originality & Quality Checks: Turnitin, iThenticate, journal guidelines
🔹 Final Submission: Journal submission system and AI disclosure requirements

⚠️ Ethical AI Principles

✅ Use AI as an assistant, not a substitute for the researcher.
✅ Verify every reference, DOI, claim, statistic, and interpretation.
❌ Never fabricate references, data, or results.
❌ Never submit AI-generated text without critical human review
✅ Protect confidential and unpublished research data.
✅ Follow the target journal's AI use policy.
AI can accelerate the research process.
Scientific contribution, critical thinking, and responsibility remain with the researcher.

Dr. Raffi Mohammed
Professor | Department of Mechanical Engineering
Ramachandra College of Engineering, Eluru, Andhra Pradesh, India


#ArtificialIntelligence #AIforResearch #ResearchTools #ReviewArticle #ResearchMethodology #AcademicWriting #BibliometricAnalysis #VOSviewer #Bibliometrix #SystematicReview #ResearchIntegrity #AcademicResearch #ScholarlyPublishing #ResponsibleAI #Researcher #MechanicalEngineering

What is the most appropriate use of generative AI (e.g., ChatGPT) in research?
This was the question I heard most often this summer when interacting with junior scholars (PhD students and assistant professors).
My initial instinct was to say NONE. Zero use of AI in management research.
After all, many of us received our doctoral training when AI existed mainly in science fiction. Remember Star Trek?
We learned to write, rewrite, and revise every sentence, agonize over every word and punctuation mark, and second-guess every phrase.
For those of us who learned English as a second, third, or fourth language, meeting the writing standards of leading journals was especially challenging. But we persisted and learned.
When I advocated zero tolerance for generative AI in academic writing and reviewing, however, my views did not seem to resonate, particularly with younger scholars. For them, AI feels far less unfamiliar than it does to some of us from an earlier generation.
This fall, the The University of Alabama offered UA AI Experience, a three-hour, self-paced online course on responsible generative AI use. I took it and came away with a clearer perspective.
My key takeaway: Use generative AI as an editor, not as a substitute for your own thinking and writing.
Consider this post. I developed the idea, drafted the text in Microsoft Word, and worked to ensure it reflected what I wanted to say, a process that is not always easy when you do not think in English.
Only after I was satisfied with the substance did I ask ChatGPT to polish the writing.
That is fundamentally different from asking ChatGPT to write the post from a few prompts.
At least, that is how I interpreted the lessons from UA AI Experience.
Why does this distinction matter?
Using AI to edit and refine writing can help level the playing field. Many scholars, including senior academics and native English speakers, have relied on professional copyeditors to sharpen their prose.
But these services can be expensive and inaccessible to scholars outside well-resourced institutions.
Generative AI puts a capable writing editor within reach of almost everyone, often at little or no cost.
Asking AI to generate the substance of scholarly writing raises different questions about intellectual ownership, integrity, and authorship.
AI cannot take responsibility for research as a human author can, and many journals restrict or prohibit listing AI tools as authors.
Thanks to UA AI Experience, I now have a more thoughtful answer when asked about appropriate AI use in research.
But I may still be wrong (as I often am).
Where do you draw the line between using AI to improve your writing and using it to do your writing?
I welcome your perspectives in the comments or via direct message!
Jef Naidoo Pratyush Nidhi Sharma Kay Palan Dr.Jaspreet Kaur Banu Goktan Bilhan Sandra Mortal Robert Hayes Peter Mohler Athina Skiadopoulou, Ph.D. Jeffrey Martin Craig Armstrong

“I will miss the creativity of teaching.”

On #WorldTeachersDay, check out this Working Life from a retired professor emeritus on how she challenged students to think beyond facts—and how she learned to teach like a scientist. https://scim.ag/4e3ygGr