An anonymous reader quotes a report from Ars Technica: It's no secret that students worldwide use AI chatbots to do their homework and avoid learning things. On the flip side, students can also use AI as a tool to beef up their knowledge and plan for the future with flashcards or study guides. Google hopes its latest Gemini feature will help with the latter. The company has announced that Gemini can now create free SAT practice tests and coach students to help them get higher scores. As a standardized test, the content of the SAT follows a predictable pattern. So there's no need to use a lengthy, personalized prompt to get Gemini going. Just say something like, "I want to take a practice SAT test," and the chatbot will generate one complete with clickable buttons, graphs, and score analysis.
Of course, generative AI can go off the rails and provide incorrect information, which is a problem when you're trying to learn things. However, Google says it has worked with education firms like The Princeton Review to ensure the AI-generated tests resemble what students will see in the real deal. The interface for Gemini's practice tests includes scoring and the ability to review previous answers. If you are unclear on why a particular answer is right or wrong, the questions have an "Explain answer" button right at the bottom. After you finish the practice exam, the custom interface (which looks a bit like Gemini's Canvas coding tool) can help you follow up on areas that need improvement. Google says support for the SAT is just the start, "with more tests coming in the future."
Cent soixante-cinq heures. C’est, en moyenne, le temps nécessaire au travail de correction pour un enseignant français. Cent soixante-cinq heures d’un travail éprouvant et répétitif. Cent soixante-cinq heures qui, multipliées par des centaines de milliers de profs, constituent une charge aussi indispensable que colossale pour le système éducatif. Et si l’IA pouvait changer la donne ?
Soixante-huit milliards d’euros de budget annuel. Près d’un million d’enseignants. Un ministère tentaculaire, aujourd’hui confronté à une crise profonde. Crise des vocations, d’abord, avec des difficultés croissantes de recrutement. Crise du niveau des élèves, ensuite, illustrée par le décrochage de la France dans les classements internationaux, notamment PISA. Dans ce contexte, difficile de ne pas s’interroger sur les leviers capables d’enrayer cette spirale.
L’intelligence artificielle, entrée brutalement dans notre quotidien avec ChatGPT il y a trois ans, pourrait bien faire partie des solutions. Non pas en remplaçant les enseignants, scénario qui relève toujours de la science-fiction, mais en s’attaquant à un angle mort du débat public : le travail invisible du professeur. Car près de la moitié de son activité se déroule hors de la classe, entre préparation des cours et correction des copies.
À ce titre, l’Éducation nationale dispose d’un gisement considérable de gains d’efficacité grâce aux outils numériques. Un potentiel que les instances ministérielles ont d’ailleurs commencé à explorer, en lançant une stratégie de développement d’agents conversationnels destinés à accompagner élèves et personnels, et en proposant des parcours de formation spécifiquement dédiés à l’IA.
Sujet tabou
C’est dans ce cadre qu’en octobre dernier, la région Hauts-de-France organisait un vaste webinaire à destination de l’ensemble des enseignants, afin de présenter les promesses, les enjeux, mais aussi les risques liés à l’usage de l’intelligence artificielle. De nombreuses pistes ont été évoquées pour alléger le travail de préparation pédagogique. Mais une question, pourtant centrale, brillait par son absence : celle de l’assistance à la correction.
Il aura fallu qu’un enseignant ose poser la question frontalement pour que le sujet soit enfin abordé… et aussitôt refermé. Selon les formateurs présents, l’usage de l’IA comme aide à la correction n’est aujourd’hui « pas recommandé », voire carrément proscrit. En cause, des outils jugés encore peu fiables, des interrogations sur la protection des données personnelles et, surtout, une conviction profondément ancrée : la correction manuelle ferait partie intégrante de l’acte pédagogique et de l’identité professionnelle de l’enseignant.
Certes, reconnaissent-ils, la correction est « coûteuse en temps ». Mais elle participerait directement à la « construction de l’acte pédagogique ». Comment, sans elle, identifier les obstacles rencontrés par les élèves et ajuster son enseignement en conséquence ?
Une position ferme, et qui semble frappée au coin du bon sens.
Un lourd tribut
Et pourtant, ce débat mérite mieux que des postures dogmatiques, même lorsqu’elles se parent d’une évidence de façade. Prenons donc le temps de regarder les chiffres, car la fameuse « construction de l’acte pédagogique » par la correction a un coût. Un coût élevé. Et pas uniquement pour la santé mentale des enseignants.
Rien que pour le baccalauréat, plus d’un million de copies doivent être corrigées chaque année, mobilisant environ un demi-million d’heures de travail cumulées. Une tâche souvent ingrate, faiblement rémunérée, qui représente pourtant à elle seule plus de 4 millions d’euros pour l’État. Mais se limiter aux examens serait réducteur. En moyenne, un enseignant consacre environ 4,6 heures par semaine à la correction, soit près de 11 % de son temps de travail total. Rapporté au coût global des enseignants en activité (environ 50 milliards d’euros par an), la seule correction des copies représente un ordre de grandeur de 5,5 milliards d’euros annuels. L’équivalent, peu ou prou, du budget du ministère de l’Agriculture.
Gardons toutefois les pieds sur terre : confier la correction à des IA ne ferait pas disparaître mécaniquement cette dépense. En revanche, cela pourrait libérer du temps enseignant pour des tâches moins pénibles et potentiellement plus utiles aux élèves. Moins de travail à la maison, c’est aussi la possibilité d’imaginer davantage de soutien individualisé, voire une évolution vers des classes à effectifs réduits. La perte de connaissance fine des élèves, invoquée pour justifier la correction manuelle, serait-elle alors réellement un problème… ou pourrait-elle être compensée autrement ?
D’autant qu’une correction automatisée pourrait fournir des retours détaillés sur les erreurs individuelles, identifier les difficultés récurrentes à l’échelle d’une classe, et n’exclurait nullement que l’enseignant lise un échantillon de copies à chaque devoir. Autrement dit, la correction par IA ne signifie pas l’abandon du regard pédagogique, mais sa réorganisation.
Et les bénéfices potentiels ne seraient pas uniquement comptables. Ils pourraient aussi être d’ordre pédagogique : davantage d’annotations pour les élèves, plus d’évaluations au cours de l’année, et peut-être même une plus grande équité, les résultats ne dépendant plus de la fatigue ou de la sévérité variable du correcteur.
À une condition essentielle, toutefois : que l’intelligence artificielle soit capable de faire ce travail avec un niveau de fiabilité suffisant.
L’humain face à la machine
Printemps 2024. En pleine « GPT mania », Thibaud Hayette, professeur de français dans l’académie de Lyon, décide de passer de la théorie à l’expérimentation. Avec l’accord de sa hiérarchie, il se procure sept copies de brevet et les soumet à ChatGPT. L’objectif est simple : tester, concrètement, la fiabilité d’un outil d’intelligence artificielle dans une tâche de correction.
Ce qui relevait encore, quelques années plus tôt, d’un « fantasme de prof » se révèle rapidement bien plus sérieux que prévu. La correction est quasi instantanée, richement commentée, et surtout, les notes attribuées par l’agent conversationnel s’avèrent remarquablement proches de celles des correcteurs humains. Une surprise, même pour l’enseignant.
Un point noir subsiste néanmoins : la reconnaissance de l’écriture manuscrite. L’OCR (la Reconnaissance Optique de Caractère), encore imparfaite, contraint Thibaud à retaper lui-même les copies. Un frein réel, mais depuis, de nouveaux modèles ont vu le jour, affichant des progrès constants.
Nous avons pu récupérer les copies utilisées par Thibaud afin de mener nos propres tests, et les résultats sont encourageants. Avec un taux d’erreur limité à 6%, Google AI Studio (utilisant le modèle Gemini 3 Pro) ne se trompe en moyenne que sur un mot sur dix-sept. Un ratio suffisamment faible pour ne pas entacher la compréhension globale d’une copie, mais qui reste problématique pour certains exercices, comme la dictée. Et sur des écritures particulièrement difficiles, l’OCR reste un obstacle. Du moins, pour l’instant.
Une fois cette étape franchie, les performances des modèles deviennent franchement impressionnantes. Nous avons testé quatre systèmes parmi les plus réputés du moment : Gemini 3 Pro (via Google AI Studio), Claude Sonnet 4.5, ChatGPT 5.2 et Mistral Large 3. À cela s’ajoute l’évaluation de la plateforme spécialisée Examino. Résultat : quel que soit le modèle, l’écart moyen sur la note globale par rapport à une correction humaine reste systématiquement inférieur à 10 %. Pour les modèles les plus performants, comme pour Examino, il tourne même autour de 6 %. Mieux encore, lorsque l’on répète plusieurs fois la correction d’une même copie, la variabilité aléatoire reste contenue, entre 3 et 4 %.
Mais un écart de 10 % avec un correcteur humain, est-ce déjà trop ? Pas nécessairement. Car il ne faut pas oublier que les correcteurs humains eux-mêmes ne sont pas infaillibles. S’il n’existe pas de statistiques officielles, les expérimentations disponibles évoquent généralement des écarts-types de 2 à 3 points sur 20, soit 10 à 15 %. Et dans certains cas, les différences entre correcteurs peuvent grimper à 10 points… voire davantage.
À l’aune de ces éléments, la variabilité observée avec les IA apparaît donc nettement inférieure à celle qui existe entre correcteurs humains. D’autant plus que la correction automatisée offre un avantage décisif : la possibilité de faire corriger plusieurs fois la même copie par la machine, puis d’en calculer une moyenne, afin de lisser les variations aléatoires. Une option tout simplement impossible à grande échelle avec une correction manuelle.
L’inertie du mammouth
Si la faisabilité technique de la correction par IA ne fait désormais plus guère de doute, les principaux freins se situent ailleurs. Pour qu’une adoption à grande échelle devienne possible, encore faut-il que l’institution s’empare réellement du sujet. Or, à en juger par le contenu des webinaires et formations estampillés Éducation nationale, l’intelligence artificielle reste d’abord perçue comme un risque potentiel, bien plus que comme une solution opérationnelle.
Le « Cadre d’usage de l’IA en éducation » donne le ton sans ambiguïté, en rappelant que « l’utilisation de l’IA pose des questions éthiques, déontologiques et écologiques ». L’intention est louable : poser des garde-fous clairs pour prévenir les dérives. Mais le mieux est parfois l’ennemi du bien, et une focalisation quasi exclusive sur les risques supposés, au détriment des gains possibles, pourrait freiner durablement le déploiement de l’IA au sein des services.
Et quand bien même une volonté politique affirmée émergerait en faveur de l’usage de l’IA pour la correction des copies, sa traduction concrète dans les pratiques prendrait du temps. Avec ses effectifs pléthoriques et ses processus lourds, le ministère de l’Éducation nationale n’est pas réputé pour sa souplesse, et l’inertie institutionnelle y est souvent considérable.
À cela s’ajoute la réticence d’une partie du corps enseignant lui-même. Dans un climat général marqué à gauche par une technophobie ambiante, l’intelligence artificielle reste souvent perçue comme un outil superficiel et dispensable… voire comme une menace, en raison notamment des usages détournés qu’en font certains élèves.
Si, pour l’Éducation nationale, la perspective d’économies constitue sans doute le levier le plus efficace pour faire bouger les lignes, il ne faudrait pas perdre de vue l’essentiel. La correction automatisée pourrait avant tout apporter des bénéfices très concrets aux personnels, et plus encore à la jeunesse. Du temps et du confort retrouvés pour les enseignants, avec à la clé des heures précieuses libérées. Et, pour les élèves, des retours plus riches sur leurs productions, ainsi qu’une évaluation plus cohérente, plus lisible et mieux harmonisée.
S’il est probable que l’intelligence artificielle s’impose, dans un horizon relativement proche, comme un outil d’aide à la correction des copies d’examen, pour les usages plus quotidiens, en revanche, il faudra sans doute attendre que la profession ouvre les yeux sur cet outil, dont les promesses apparaissent aujourd’hui aussi enthousiasmantes que vertigineuses.
U.S. college graduates "have historically found jobs more quickly than people with only a high school degree," writes Bloomberg.
"But that advantage is becoming a thing of the past, according to new research from the Federal Reserve Bank of Cleveland."
"Recently, the job-finding rate for young college-educated workers has declined to be roughly in line with the rate for young high-school-educated workers, indicating that a long period of relatively easier job-finding prospects for college grads has ended," Cleveland Fed researchers Alexander Cline and BarıÅY Kaymak said in a blog post published Monday. The study follows the latest monthly employment data released on Nov. 20, which showed the unemployment rate for college-educated workers continued to rise in September amid an ongoing slowdown in white-collar hiring... The unemployment rate for people between the ages of 20 to 24 was 9.2% in September, up 2.2 percentage points from a year prior.
There is a caveat. "Young college graduates maintain advantages in job stability and compensation once hired..." the researchers write. "The convergence we document concerns the initial step of securing employment rather than overall labor market outcomes."
Their research includes a graph showing how the "unemployment gap" first increased dramatically after 2010 between college-educated and high school-educated workers, which the researchers attribute to "the prolonged jobless recovery after 2008". But that gap has been closing ever since, with that gap now smaller than at any time since the 1970s.
"Young high school workers are riding the wave of the historically tight postpandemic labor market with well-below-average unemployment compared to that of past high school graduates, while young college workers are experiencing unemployment rates rarely observed among past college cohorts barring during recessions."
The labor market advantages conferred by a college degree have historically justified individual investment in higher education and expanding support for college access. If the job-finding rate of college graduates continues to decline relative to the rate for high school graduates, we may see a reversal of these trends. The convergence we document concerns the initial step of securing employment rather than overall labor market outcomes. These details suggest a nuanced shift in employment dynamics, one in which college graduates face greater difficulty finding jobs than previously but maintain advantages compared with high school graduates in job stability and compensation once hired.
Two key quotes:
"Declining job prospects among young college graduates may reflect the continued growth in college attainment, adding ever larger cohorts of college graduates to the ranks of job seekers, even though technology no longer favors college-educated workers."
"Developments related to AI, which may be affecting job-finding prospects in some cases, cannot explain the decades-long decline in the college job-finding rate."
U.S. college graduates "have historically found jobs more quickly than people with only a high school degree," writes Bloomberg.
"But that advantage is becoming a thing of the past, according to new research from the Federal Reserve Bank of Cleveland."
"Recently, the job-finding rate for young college-educated workers has declined to be roughly in line with the rate for young high-school-educated workers, indicating that a long period of relatively easier job-finding prospects for college grads has ended," Cleveland Fed researchers Alexander Cline and BarıÅY Kaymak said in a blog post published Monday. The study follows the latest monthly employment data released on Nov. 20, which showed the unemployment rate for college-educated workers continued to rise in September amid an ongoing slowdown in white-collar hiring... The unemployment rate for people between the ages of 20 to 24 was 9.2% in September, up 2.2 percentage points from a year prior.
There is a caveat. "Young college graduates maintain advantages in job stability and compensation once hired..." the researchers write. "The convergence we document concerns the initial step of securing employment rather than overall labor market outcomes."
Their research includes a graph showing how the "unemployment gap" first increased dramatically after 2010 between college-educated and high school-educated workers, which the researchers attribute to "the prolonged jobless recovery after 2008". But that gap has been closing ever since, with that gap now smaller than at any time since the 1970s. "Young high school workers are riding the wave of the historically tight postpandemic labor market with well-below-average unemployment compared to that of past high school graduates, while young college workers are experiencing unemployment rates rarely observed among past college cohorts barring during recessions."
The labor market advantages conferred by a college degree have historically justified individual investment in higher education and expanding support for college access. If the job-finding rate of college graduates continues to decline relative to the rate for high school graduates, we may see a reversal of these trends. The convergence we document concerns the initial step of securing employment rather than overall labor market outcomes. These details suggest a nuanced shift in employment dynamics, one in which college graduates face greater difficulty finding jobs than previously but maintain advantages compared with high school graduates in job stability and compensation once hired.
Two key quotes:
"Declining job prospects among young college graduates may reflect the continued growth in college attainment, adding ever larger cohorts of college graduates to the ranks of job seekers, even though technology no longer favors college-educated workers.
"Developments related to AI, which may be affecting job-finding prospects in some cases, cannot explain the decades-long decline in the college job-finding rate.
Jeunes parents, vous n’avez pas pu y échapper. Il y a une quinzaine d’années, l’éducation positive a déferlé sur la France, portée par un cocktail de grands principes, de références aux neurosciences… et d’injonctions parfois culpabilisantes. Mais ce modèle éducatif tient-il réellement ses promesses ? Et au fond, est-il vraiment si « positif » ?
Lorsque Isabelle Filliozat publie J’ai tout essayé en 2011, rien ne laissait présager un tel raz-de-marée éditorial. Jusqu’alors, l’auteure prêchait plutôt dans le désert. Son précédent ouvrage consacré à l’éducation, Au cœur des émotions de l’enfant, paru en 1999, n’avait rencontré qu’un écho modeste. Mais cette fois, le succès est immédiat et massif, et sera le point de départ d’un véritable phénomène de société : En quelques années à peine, l’éducation « positive » qu’elle promeut s’impose progressivement comme une évidence, presque comme une nouvelle norme. Et son premier livre, rétrospectivement, est élevé au rang de texte fondateur du mouvement.
Dans la foulée, en 2014, la pédiatre Catherine Gueguen publie Pour une enfance heureuse. Là encore, le succès est spectaculaire. Forte de sa formation médicale, elle vient appuyer les intuitions de Filliozat (psychologue de formation) par des arguments issus des neurosciences. Les principes de l’éducation bienveillante gagnent alors un vernis scientifique qui achève de la légitimer aux yeux du grand public.
Mais de quelles thèses parle-t-on exactement ?
Les axiomes de l’éducation bienveillante
Tous les parents le savent : la parentalité ressemble souvent à un parcours du combattant. Dans ce contexte, la soif de réponses est immense, et c’est précisément là que ces ouvrages rencontrent leur public. Ils promettent des clés simples, directement applicables, presque des modes d’emploi pour parents démunis. Le tout enveloppé de valeurs auxquelles personne ne peut décemment s’opposer : le bien-être de l’enfant, la sérénité familiale, l’amour, l’épanouissement. Rien d’étonnant, dès lors, à ce que ces livres deviennent des best-sellers.
D’autant que le message est clair : tout se joue pendant l’enfance. Faites les mauvais choix, commettez des erreurs éducatives (sous-entendu : ne respectez pas nos conseils), et vos enfants en paieront le prix toute leur vie.
Et parmi ces erreurs, la plus grave serait de stresser l’enfant. Catherine Gueguen alerte ainsi sur les effets délétères du cortisol, l’hormone du stress, qui provoquerait des dégâts irréversibles sur le cerveau en développement. À l’inverse, l’ocytocine — hormone de l’attachement et du bien-être — favoriserait un développement cérébral optimal. Conclusion : éviter à tout prix colères, frustrations et sanctions, au profit des câlins, et autres témoignages d’affection. Sous peine, là encore, de conséquences durables.
Dans cette logique, toute « violence éducative ordinaire » est proscrite. Bien sûr, les violences physiques sont bannies. Mais la notion s’étend bien au-delà : crier, punir, contraindre un enfant à faire quelque chose qu’il ne souhaite pas deviennent, eux aussi, des formes de violence psychologique.
Et si, malgré toute cette vigilance, une colère incontrôlée survient, il ne s’agit surtout pas d’un caprice. Les caprices, nous dit-on, n’existent pas. Ces débordements seraient des appels à l’aide, auxquels il faudrait répondre par de l’affection, de l’écoute, du réconfort, et surtout du dialogue. D’autant que, chez le jeune enfant, le cortex préfrontal, siège du raisonnement et du contrôle de soi, est immature. L’enfant serait donc gouverné par ses pulsions, incapable de planifier ou de se réguler. Dans ces conditions, toute sanction serait non seulement injuste, mais contre-productive. Alors qu’un sain dialogue permettrait de « reconnecter » le néocortex, qui reprendrait le dessus sur le cerveau limbique et le fameux « cerveau reptilien », responsables des émotions et des réactions instinctives.
Isabelle Filliozat pousse le raisonnement encore plus loin. Un enfant agité, qui court, grimpe, déborde d’énergie ? C’est parfaitement normal. « La vie, c’est le mouvement », affirme-t-elle. Chercher à canaliser ces comportements reviendrait à aller contre sa nature biologique. À l’inverse, un enfant trop obéissant devient suspect : il serait en réalité en état de stress ou de sidération. Face à un parent autoritaire, son cerveau reptilien déclencherait une réponse de paralysie.
Et chez l’adolescent ? La crise est présentée comme inévitable. Le cortex préfrontal, encore immature, empêche toute autorégulation efficace, tandis que le cerveau est inondé de dopamine, rendant le circuit de la récompense hyperactif. Punir un ado pour ses excès serait donc profondément injuste. Cela reviendrait à reprocher à un tout-petit de ne pas savoir marcher.
Pire encore, le cerveau adolescent étant en pleine reconfiguration, il doit expérimenter pour développer ses propres facultés. Si le parent décide à sa place, contrôle trop, cadre trop, le cerveau de l’adolescent ne « câble » pas l’autonomie attendue.
La conclusion logique est que, pour être juste et bienveillant, il faudrait laisser faire. Prendre du recul. Accepter les débordements comme des passages nécessaires. Insultes, violences, conduites à risque, consommation de substances ? Rien d’anormal : ce serait, là encore, la simple expression de sa nature.
Sous le vernis de la science
Pris isolément, ces axiomes peuvent sembler relever du bon sens. Mieux encore, ils paraissent solidement adossés à des résultats scientifiques. Mais dès que l’on gratte un peu le vernis, la fragilité de l’argumentation apparaît rapidement.
A new article from Bloomberg says dozens of America's colleges "succumbed to a fundamental problem killing colleges across the US: not enough students. The schools will award their final degrees this spring, stranding students not yet ready to graduate and forcing faculty and staff to hunt for new jobs."
The country's tumbling birth rate is pushing schools toward a "demographic cliff," where a steadily dropping population of people in their late teens and early 20s will leave desks and classrooms empty. Many smaller, lesser-known schools like Cazenovia have already hit the precipice. They're firing professors, paring back liberal arts courses in favor of STEM — or closing altogether. Others will likely reach the cliff in the next few years... [T]the US birth rate ticked upward slightly before the 2008 financial crisis, and that brief demographic boost has kept enrollment at larger schools afloat. But the nationwide pool of college-aged Americans is expected to shrink after 2025. Schools face the risk that each incoming class could be smaller than the last. The financial pressure will be relentless...
Since 2020, more than 40 schools have announced plans to close, displacing students and faculty and leaving host towns without a key economic engine... Close to 400 schools could vanish in the coming decade, according to Huron Consulting Group. The projected closures and mergers will impact around 600,000 students and redistribute about $18 billion in endowment funds, Huron estimates... Pennsylvania State University, citing falling enrollment at many of its regional branches, plans to shutter seven of its 20 branch campuses after the spring 2027 semester... [C]ampuses in far-flung places, without brand recognition, are falling out of favor with students already questioning the value of a college degree. For example, while Penn State's flagship University Park campus saw enrollment grow 5% from 2014 to 2024, 12 other Penn State campuses recorded a 35% drop, according to a report tasked with determining whether closures were necessary.
The article notes that "Less than half of students whose schools shut down before they graduate re-enroll in another college or university, according to a 2022 study."
But even at colleges that remain, "The shrinking supply of students has already sparked a frenzied competition for high school seniors..."
Some public institutions are letting seniors bypass traditional requirements like essays and letters of recommendation to gain entry automatically... Direct-admission programs, which allow students to skip traditional applications, are one potential response. Some 15 states have them, according to Taylor Odle, assistant professor of educational policy studies at the University of Wisconsin-Madison. He found in a 2022 paper that direct admissions increased first-year undergrad enrollment by 4% to 8%... And they don't require nearly as many paid staff to run, since there are no essays or letters of recommendation to read.
A researchers' propensity for risky projects is passed down to their doctoral students -- and stays with trainees after they leave the laboratory, according to an analysis of thousands of current and former PhD students and their mentors. From a report: Science involves taking risks, and some of the most impactful discoveries require taking big bets. However, scientists and policymakers have raised concerns that the current academic system's emphasis on short-term outcomes encourages researchers to play it safe. Studies have shown, for example, that risky research is less likely to be funded. Anders Brostrom, an economist studying science policy at the University of Gothenburg in Sweden, and his colleagues decided to examine the role of doctoral education in shaping risk-related behaviour -- an area that Brostrom says has been largely overlooked.
"We often focus on thinking about how we can change the funding systems to make it more likely for people to take risks, but that's not the only lever we have," says Chiara Franzoni, an economist at the Polytechnic University of Milan in Italy. This study is "refreshing" because "we've discussed policy interventions a lot, but we haven't discussed training," she adds. [...] The team found that students' risk-taking dispositions matched those of their supervisors. This link was stronger when students and their supervisors communicated frequently, and weaker when students were also mentored by scientists outside their lab.
DesScorp writes: CNBC reports that new data from the National Student Clearinghouse indicates that enrollment growth in four year degree programs is slowing down, while growth in two year and certification programs is accelerating: Enrollments in undergraduate certificate and associate degree programs both grew by about 2% in fall 2025, while enrollment in bachelor's degree programs rose by less than 1%, the report found. Community colleges now enroll 752,000 students in undergraduate certificate programs -- a 28% jump from just four years ago.
Overall, undergraduate enrollment growth was fueled by more students choosing to attend community college, the report found. "Community colleges led this year with a 3% increase, driven by continued rising interest in those shorter job-aligned certificate programs," said Matthew Holsapple, the National Student Clearinghouse Research Center's senior director of research.
For one thing, community college is significantly less expensive. At two-year public schools, tuition and fees averaged $4,150 for the 2025-2026 academic year, according to the College Board. Alternatively, at four-year public colleges, in-state tuition and fees averaged $11,950, and those costs at four-year private schools averaged $45,000. A further factor driving this new growth is that Pell Grants are now available for job-training courses like certifications.
The abstract of a paper on NBER: School boards have statutory authority over most elementary and secondary education policies, but receive little attention compared to other actors in education systems. A fundamental challenge to understanding the importance of boards is the absence of data on the policy goals of board members -- i.e., their ideologies -- forcing researchers to conduct tests based on demographic and professional characteristics -- i.e., identities -- with which ideology is presumed to correlate.
This paper uses new data on the viewpoints and policy actions of school board members, coupled with a regression discontinuity design that generates quasi-random variation in board composition, to establish two results. The first is that the priorities of board members have large causal effects across many domains. For example, the effect of electing an equity-focused board member on test scores for low-income students is roughly equivalent to assigning every such student a teacher who is 0.3 to 0.4 SDs higher in the distribution of teacher value-added. The second is that observing policy priorities is crucial. Identity turns out to be a poor proxy for ideology, with limited governance effects that are fully explained by differences in policy priorities. Our findings challenge the belief that school boards are unimportant, showing that who serves on the board and what they prioritize can have far-reaching consequences for students.
The "talent is everywhere" approach that U.S. employers adopted during the white-hot pandemic job market is quietly giving way to something much older and more familiar: recruiting almost exclusively from a small set of elite and nearby universities. A 2025 survey of more than 150 companies by Veris Insights found that 26% were exclusively recruiting from a shortlist of schools, up from 17% in 2022.
Diversity as a priority for school recruiting selection dropped to 31% of employers surveyed in 2025, down from nearly 60% in 2022. GE Appliances once sent recruiters on one or two passes through 45 to 50 schools each year; now the company attends four or five events per semester at just 15 universities, including Purdue and Auburn. McKinsey, the consulting firm that expanded recruitment well beyond the Ivy League after George Floyd's murder, recently removed language from its career page that said "We hire people, not degrees." The firm now hosts in-person events at a shortlist of about 20 core schools, including Vanderbilt and Notre Dame.
Most companies now recruit at up to 30 American colleges out of about 4,000, said William Chichester III, who has directed entry-level recruiting at Target and Peloton. For students outside elite schools or those located near company headquarters? "God help you," he said.
A survey of 2,000 teachers, students and parents conducted by the New York Times found that many high schools have stopped assigning full novels to students, opting instead for excerpts that are often read on school-issued laptops rather than in print. The shift stems from multiple factors: a belief that students have shorter attention spans, pressure to prepare students for standardized tests, and the influence of Common Core standards adopted by many U.S. states more than a decade ago.
Schools increasingly rely on curriculum products like StudySync, which takes an anthology approach to literature rather than requiring complete books. Teachers acknowledge that teens now read far fewer full novels than previous generations, though some educators push back against the trend. "Many teachers are secret revolutionaries and still assign whole books," said Heather McGuire, a New Mexico English teacher who responded to the survey.
Public opinion surveys paint a picture of Americans souring dramatically on higher education, as Pew found that the share of adults calling college "very important" dropped from 70% in 2013 to just 35% today, and NBC polling shows that 63% now believe a degree is "not worth the cost," up from 40% over the same period. Yet enrollment data tells a different story.
Four-year institutions awarded 2 million bachelor's degrees in 2023, up from 1.6 million in 2010, and the fraction of 25-year-olds holding a bachelor's degree has steadily increased for the past 15 years. The economic case remains strong. The average bachelor's degree holder earns about 70% more than a high-school graduate of similar work experience, and after factoring in financial aid, the cost of attending a public four-year college has fallen by more than 20% since 2015.
Even after accounting for student-debt payments, college graduates net about $8,000 more annually than those without degrees. Part of the disconnect may stem from misunderstanding how college pricing works. Nearly half of U.S. adults believe everyone pays the same tuition, though fewer than 20% of families actually pay the published sticker price.
Starting in July 2026, every public school district in California will be required to have policies on the books that restrict or prohibit students from using smartphones during the school day, thanks to Assembly Bill 3216 that Governor Gavin Newsom signed into law back in 2024.
The legislation also mandates that districts update these policies every five years. Newsom had previously signed related legislation in 2019, though that earlier law merely affirmed that school districts have the authority to regulate smartphone use rather than requiring them to do so.
A group of researchers is calling on universities to treat consulting work as a strategic priority, arguing that bureaucratic obstacles and inconsistent policies have left a massive revenue stream largely untapped even as higher education institutions face mounting financial pressures. (Consulting work refers to academics offering their advice and expertise to outside organizations -- industry, government, civil society -- for a fee. It's one of the most direct and scalable ways academics can shape the world beyond campus, and the projects are typically shorter in duration and easier to set up than alternatives like spin-out companies.)
Writing in Nature, the authors found that fewer than 10% of academic staff at nine UK universities engaged in consulting work, and the number of academic consulting contracts across the country fell 38% over the past decade -- from around 99,000 in 2014-15 to fewer than 62,000 in 2023-24.
Academic consulting in the UK is currently worth roughly $675-810 million annually, a figure that represents just 0.6% of the country's $124 billion management consulting market. The authors examined policies at 30 universities and surveyed 76 fellows from a UK Research and Innovation programme. Two-thirds of the surveyed institutions had publicly available consulting policies, and two outright prohibit private consulting. Permitted consulting time ranged from unlimited to 30 days or fewer per year, institutional charges varied from 10-40% of fees, and contract approval timelines stretched from 24 hours to several months.
Private consultancy firms are moving into this space, capturing opportunities that universities neglect. Small-scale projects under $6,750 are commonly sidelined by university contract offices because they represent too small an income for strained institutional resources. The authors propose standardized policies across institutions, shared consulting income with departments, and faster approval processes -- reforms similar to those already implemented for university spin-out companies.
An anonymous reader quotes a report from Wired: Two years ago, Lizmary Fernandez took a detour from studying to be an immigration attorney to join a free Apple course for making iPhone apps. The Apple Developer Academy in Detroit launched as part of the company's $200 million response to the Black Lives Matter protests and aims to expand opportunities for people of color in the country's poorest big city. But Fernandez found the program's cost-of-living stipend lacking -- "A lot of us got on food stamps," she says -- and the coursework insufficient for landing a coding job. "I didn't have the experience or portfolio," says the 25-year-old, who is now a flight attendant and preparing to apply to law school. "Coding is not something I got back to."
Since 2021, the academy has welcomed over 1,700 students, a racially diverse mix with varying levels of tech literacy and financial flexibility. About 600 students, including Fernandez, have completed its 10-month course of half-days at Michigan State University, which cosponsors the Apple-branded and Apple-focused program. WIRED reviewed contracts and budgets and spoke with officials and graduates for the first in-depth examination of the nearly $30 million invested in the academy over the past four years -- almost 30 percent of which came from Michigan taxpayers and the university's regular students. As tech giants begin pouring billions of dollars into AI-related job training courses across the country, the Apple academy offers lessons on the challenges of uplifting diverse communities.
[...] The program gives out iPhones and MacBooks and spends an estimated $20,000 per student, nearly twice as much as state and local governments budget for community colleges. [...] About 70 percent of students graduate, which [Sarah Gretter, the academy leader for Michigan State] describes as higher than typical for adult education. She says the goal is for them to take "a next step," whether a job or more courses. Roughly a third of participants are under 25, and virtually all of them pursue further schooling. [...] About 71 percent of graduates from the last two years went onto full-time jobs across a variety of industries, according to academy officials. Amy J. Ko, a University of Washington computer scientist who researches computing education, calls under 80 percent typical for the coding schools she has studied but notes that one of her department's own undergraduate programs has a 95 percent job placement rate.
Goldman Sachs, in a note this week, via India Dispatch: There are various reasons that explains this: (i) A large part of the global education spend goes towards formal education (schools, colleges and universities), which are typically either run by governments or are not-for-profit institutions;
(ii) It is difficult to replicate education quality at scale in our view, since most teachers would have a different pedagogy, and thus standardization is harder to achieve vs that in other internet categories;
(iii) Education is fragmented - it includes various fields (schools, undergrad courses, medicine, engg, management, etc.), each with their own curriculum, and the same being vastly different across countries globally; this makes scalability difficult beyond a few certain specializations and regions.
Additionally, we believe the ability for online education to capture a sizable value share of supplemental education is limited since the perceived value of offline, including that from community, in-person engagement and doubt solving, rigour, etc., is typically higher.
However, we note that before China's double reduction policy in 2021, TAL and EDU had market caps of up to US$50 bn; these companies were mostly domestic focused and on the K-12 tutoring segment, which has large volumes. Similarly in India, Byju's reached a peak valuation of US$20 bn+ (link; again, focused on K-12), before issues around governance etc. impacted the business.
Long-time Slashdot reader theodp writes: Last September, tech-backed nonprofit Code.org pledged to engage 25 million K-12 schoolchildren in an "Hour of AI" this school year. Preliminary numbers released this week by the Code.org Advocacy Coalition showed that [halfway through the five-day event Computer Science Education Week] 13.1 million users had participated in the inaugural Hour of AI, attaining 52.4% of its goal of 25 million participants.
In a pivot from coding to AI literacy, the Hour of AI replaced Code.org's hugely-popular Hour of Code this December as the flagship event of Computer Science Education Week (December 8-14). According to Code.org's 2024-25 Impact Report, "in 2024–25 alone, students logged over 100 million Hours of Code, including more than 43 million in the four months leading up to and including CS Education Week."
Minecraft participated with their own Hour of AI lessons. ("Program an AI Agent to craft tools and build shelter before dusk falls in this iconic challenge!") And Google contributed AI Quests, "a gamified, in-class learning experience" allowing students to "step into the shoes of Google researchers using AI to solve real-world challenges." Other participating organizations included the Scratch Foundation, Lego Education, Adobe, and Roblox.
And Microsoft contributed two — including one with their block-based programming environment Microsoft MakeCode Arcade, with students urged to "code and train your own super-smart bug using AI algorithms and challenge other AI bugs in an epic Tower battle for ultimate Bug Arena glory!"
See all the educational festivities here...
An anonymous reader quotes a report from the Guardian: The Massachusetts Institute of Technology (MIT) community is grieving after the "shocking" shooting death of the director of its plasma science and fusion center, according to officials. Nuno FG Loureiro, 47, had been shot multiple times at his home in the affluent Boston suburb of Brookline on Monday night when police said they received a call to investigate. Emergency responders brought Loureiro to a hospital, and the award-winning scientist was pronounced dead there Tuesday morning, the Norfolk county district attorney's office said in a statement.
The Boston Globe reported speaking with a neighbor of Loureiro who heard gunshots, found the academic lying on his back in the foyer of their building and then called for help alongside the victim's wife. The statement from the Norfolk district attorney's office said an investigation into Loureiro's slaying remained ongoing later Tuesday. But the agency did not immediately release any details about a possible suspect or motive in the killing, which gained widespread attention across academic circles, the US and in Loureiro's native Portugal.
Portugal's minster of foreign affairs announced Loureiro's death in a public hearing Tuesday, as CNN reported. Separately, MIT president Sally Kornbluth issued a university-wide letter expressing "great sadness" over the death of Loureiro, whose survivors include his wife. "This shocking loss for our community comes in a period of disturbing violence in many other places," said Kornbluth's letter, released after a weekend marred by deadly mass shootings at Brown University in Rhode Island -- about 50 miles away from MIT -- as well as on Australia's Bondi Beach. The letter concluded by providing a list of mental health resources, saying: "It's entirely natural to feel the need for comfort and support."
The conventional wisdom that English prose has gotten easier to read because sentences have gotten shorter is wrong, according to a new analysis published in Works in Progress by writer and Mercatus Center research fellow Henry Oliver. The real transformation happened centuries ago in the 1500s and 1600s when Bible translators like William Tyndale and Thomas Cranmer developed a "plain style" built on logical syntax rather than the older rhythmic, periodic structures inherited from medieval prose.
Oliver argues that much of what modern datasets measure as declining sentence length is actually just changing punctuation habits. Writers now use periods where earlier generations used colons and semicolons. One dataset shows semicolon usage dropped from one every 90 words in 1781 to one every 390 words today. The cognitive complexity of a paragraph often remains the same regardless of how it's punctuated. Even wildly popular modern books don't follow the "short sentences equal readable" formula. Oliver points to Onyx Storm, the 2025 fantasy novel that has sold tens of millions of copies, which opens with sentences of 24 and 30 words. The 30-word sentence has a subordinate clause twice as long as its main clause. The book reads easily not because sentences are short but because the language is plain and the syntax is logical.
The traditional signals that employers used to evaluate entry-level job candidates -- college GPAs, cover letters, and interview performance -- have lost much of their value as grade inflation and widespread AI use render these metrics nearly meaningless, writes The Atlantic.
The recent-graduate unemployment rate now sits slightly higher than the overall workforce's, a reversal from historical norms where new college graduates were more likely to be employed than the average worker. Job postings on Handshake, a career-services platform for students and recent graduates, have fallen by more than 16 percent in the past year. At Harvard, 60% of undergraduate grades are now A's, up from fewer than a quarter two decades ago. Seven years ago, 70% of new graduates' resumes were screened by GPA; that figure has dropped to 40%.
Two working papers examining Freelancer.com found that cover-letter quality once strongly predicted who would get hired and how well they would perform -- until ChatGPT became available. "We basically find the collapse of this entire signaling mechanism," researcher Jesse Silbert said. The average number of applications per open job has increased by 26% in the past year. Students at UC Berkeley are now applying to 150 internships just to land one or two interviews.
Connexion
