Skip to content
Vol. IIssue 18

Educational AI Daily

Artificial intelligence, as told by three teaching communities

Back to the issue

In the print dialog, choose “Save as PDF” to keep it as a file.

https://diarioiaeducativa.github.io/en/2026-11-18/stanford-evidencia-ia-escuela.html

Tools 19 November

Stanford study

Stanford gathers 52 causal studies on AI in schools and sees uneven results

The November review goes from 20 to 52 studies and concludes that AI looks more promising for understanding new material than for producing assignments, but no use improves learning consistently.

Stanford University’s SCALE initiative published “The Evidence Base on AI in K-12: A November 2026 Review” on 10 November 2026, an update of its March review. It includes only causal studies, that is, randomised trials and similar designs that make it possible to attribute an effect to AI. It goes from 20 to 52 studies, 32 of them new and published between October 2025 and June 2026. Five of the new ones, focused on students, were carried out in US schools, compared with none in the previous review.

According to the review, using AI to support learning looks more favourable than using it to do the assignments: in a study of university students in Brazil, producing the work with AI worsened performance in a later assessment without it. Even so, no type of use improves learning consistently, and students liking the experience does not always translate into better results.

Among the cases it cites is one in Turkey, with 193 teachers and 2,816 students from grades 5 to 12, in which giving teachers an AI assistant based on GPT-4o reduced students’ motivation and confidence; marks fell with teachers who had lower prior performance. Another, in Sierra Leone, with 1,423 mathematics students in grades 7 and 8, showed an improvement when teachers built Gemini Guided Learning, the guided-study mode of Google’s assistant, into half of their lessons; it is signed by a team from Google and Fab AI, and Google.org is among the funders of the review itself.

The authors themselves warn that the results cannot be generalised to the whole US school system, that the evidence on equity and long-term effects is scarce and that many studies are preprints, that is, papers circulated without peer review.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/galicia-ley-educacion-digital.html

Institutions & Industry 19 November

Regional legislation

Galicia’s parliament passes the law regulating phones, screens and AI in classrooms

The Digital Education Law limits individual use of devices up to the fifth year of primary school and was passed with the votes of the PPdeG and the opposition of the BNG and the PSdeG.

The Parliament of Galicia gave final approval on Tuesday 17 November 2026 to the Digital Education Law (Ley de educación digital), according to Galicia Confidencial. The law regulates the use of phones, screens and artificial intelligence in schools. The Partido Popular de Galicia (PPdeG) voted in favour, Democracia Ourensana abstained, and the Bloque Nacionalista Galego (BNG) and the Partido dos Socialistas de Galicia (PSdeG) voted against.

According to that report, the law limits the individual use of electronic devices depending on students’ age, up to the fifth year of primary school, and allows phones for educational purposes from the third year of ESO (compulsory secondary education, students aged 14 to 15). It also bans recording lessons without authorisation, whether in person or online, and extends teachers’ authority to the digital sphere.

The text that reached the plenary session, the Education Committee’s report published on 10 November in the Official Bulletin of the Parliament of Galicia, gives AI a title of its own, Title VI. State schools and their staff may only use AI systems authorised by the regional education department (Consellería), which will list them in a catalogue, and schools must explain to students and families in advance what they are used for and what data they process. The text allows AI to be used to personalise learning, prepare materials that teachers review before handing them out, mark multiple-choice tests with the grade validated by a teacher, and offer assistants that answer questions outside class hours. It bans using it to replace teachers or to make decisions about students without human supervision, as well as facial recognition, emotion analysis and profiling that could lead to discrimination. In practice, once the law comes into force, teachers in Galicia’s state schools will not be able to use any AI tool that is not in that catalogue.

The opposition criticised that only 4 of the PSdeG’s 158 amendments and 6 of the BNG’s 69 were accepted, and called for more consensus with the education community.

In the group, the news was shared as the first regional digital education law, with a chapter on AI. The “first” claim comes, in the report, from the regional education minister (conselleiro), Román Rodríguez, who maintains that Galicia is the first autonomous community to pass a specific law to regulate technology and AI in classrooms; the paper has not checked this against another source.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Update on 19 November 2026: added what the text that reached the plenary session says about AI, published in the Official Bulletin of the Parliament of Galicia; an earlier version described it as unconfirmed.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/educamadrid-agente-ia.html

Institutions & Industry 19 November

Education administration

Madrid adds to EducaMadrid an AI agent that prepares personalised materials

According to the Community of Madrid, the agent was developed by civil servants in its regional digital ministry and can be used by teachers in publicly funded schools.

The Community of Madrid announced on 16 November 2026, in a note on its website, an artificial intelligence agent on the EducaMadrid platform. According to the note, it generates personalised content for each student in real time and was developed by civil servants in the regional digital ministry (Consejería de Digitalización).

Among its functions, the note lists producing notes and revision documents from the platform’s own educational sources, creating self-assessment quizzes that are marked within the platform itself, tracking each student’s progress and drafting or reviewing emails in several languages. It is aimed at teachers, who can adapt the materials, and can be used by all publicly funded schools in the region. The administration estimates it will free up some 40 hours a year of routine tasks for each teacher.

The note does not say which AI model it uses, where it is hosted or how students’ data is handled. Those were precisely the doubts raised in the group, along with the quality of the content it generates. It was also seen as a first step by the administration.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/cmu-tutoria-socratica-errores.html

Library 19 November

Experimental study

Experiment with 400 adults finds a Socratic AI tutor teaches more than explaining the mistake

Carnegie Mellon researchers compare four ways of giving feedback after a mistake and suggest that what works is sending the student back to the task until they get it right.

Michael W. Asher, Meng Cao and Paulo F. Carvalho, of Carnegie Mellon University's Human-Computer Interaction Institute, published a study on arXiv on 15 November 2026 that tests a widely held intuition: that a language model helps students learn more the better it explains why they went wrong. It is a research paper, with a preregistered experiment and open data and code, not a guide or an official recommendation.

The starting point is that the AI tutors evaluated so far have produced modest results. The authors recall the two-year trial of Khanmigo in 18 secondary schools, by Philip Oreopoulos and Nicole Low, in which the effect on standardised tests was small and the median student talked to the tutor in only 17% of the sessions in which they made mistakes. Their hypothesis is that a chatbot that waits to be asked comes too late: people who fail tend to give up. They propose a proactive tutor, built into the cycle of practice and correction.

The experiment brought together 400 adults in the United States who did not know how to set up constraint equations (for example, the one relating the metres of fencing available to the dimensions of an enclosure). After each attempt at three problems, they were randomly assigned one of four forms of feedback: seeing only the correct answer; a generic, pre-written explanation; that explanation plus a personalised one about their own mistake, generated on the spot by Gemini 3.5 Flash; or a conversation with a tutor, the same model, which was forbidden from giving the answer and had to guide them with Socratic questions.

The results favour the Socratic tutor. It led participants to correct 131 of their 206 mistakes themselves (64%), and they scored 9 percentage points higher (0.24 standard deviations) than the groups given explanations on three final questions similar to those they had practised. They also said the session had been more interesting, less frustrating and less costly, although it lasted almost twice as long: 13.3 minutes on average against between 6.8 and 7.3 in the other conditions. Personalised explanations did not improve on generic ones, and neither reliably outperformed simply showing the correct answer.

It should be read with its limits, which the authors acknowledge. The tutor's advantage appeared only on problems like those practised, not on the question about an untaught objective or on the problem that combined several, and the difference in the final test has a p-value of 0.048, right on the edge of significance. It was a single session of about twenty minutes, on one maths topic and with adults, not school students. Moreover, the Socratic condition bundled several things together (trying again, tailored questions and an encouraging tone), so the design does not separate which of them produces the effect. Nor does it compare with the same time spent on more exercises, and the cost, about 12 US cents per participant, grows with the number of users.

From this they draw a framework they call mastery experience: practice comes before explanation, the difficulty should produce informative errors without overwhelming, and the student needs to be able to retry until they succeed. Their reading is that the value of an AI tutor lies not in explaining the mistake more precisely but in turning it into a success of the student's own.

For teachers, the study does not prove any classroom practice, but it leaves design clues that can be tried out: asking the assistant not to give the solution and to break the problem down into smaller questions, and requiring a first attempt before any help. For context, the paper carried Stanford's review of 52 causal studies on AI in schools on 18 November, which reports uneven results. To get started, it is enough to read the abstract and sections 4 and 5.2 of the paper.

See the document in the ChatGPT-IA-edu Digital Library

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

See the document in the ChatGPT-IA-edu Digital Library

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/suno-visuals.html

Multimedia 19 November

Music and video

Suno launches Visuals to create lyric videos and cover art from its songs

The feature generates clips of up to 15 seconds and is paid for with credits separate from those for songs.

Suno, the AI music platform, presented Visuals on 17 November 2026, a tool for illustrating its songs: lyric videos that follow the rhythm, album covers and clips made from a photo. It is available on the web and in the iOS and Android apps.

According to the announcement, the clips last up to 15 seconds and can be extended, edited and regenerated with text. Visuals credits, the units used to pay for each use, are bought separately from song credits. Suno blocks requests and uploads that depict artists, public figures or protected works, asks users to confirm they hold the rights to what they upload and watermarks these images and videos before they are shared outside the platform.

In the classroom, it could be used to illustrate songs created by students, always checking its terms of use first and without uploading images of minors.

Read the original conversation in EduMedia-IAG

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in EduMedia-IAG

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/umami-analitica-privacidad.html

Vibe coding 19 November

Development tip

A teacher measures the use of their apps with Umami, cookie-free statistics

This tool shows which features students use without collecting names or email addresses.

Presenting his catalogue of reading texts, Trama, its author explained in the group that he uses Umami, an open-source system that records how web pages are used, to find out which of his educational apps are used most and what actions are taken (starting an activity, choosing a level, downloading a resource). With that he decides where to put his effort.

Umami works without cookies and does not store the IP address of the people visiting the page, though it does store their approximate city, browser and device type. It can be installed for free on your own server or computer, under the MIT licence, a free-software licence, or used through its cloud version, whose free plan allows 100,000 events a month on a single website and keeps the data for six months.

Not using cookies does not exempt a site from telling its users. The Spanish Data Protection Agency (AEPD) allows audience measurement without asking for consent if the data are used only for the site's own statistics, are not passed on to third parties and are kept for no more than 25 months, but it requires this to be explained in the privacy policy. So with Umami used this way there is no need to ask for permission or show a cookie banner, but the app's privacy policy must explain that its use is measured, and the data must not be kept for more than 25 months.

Another idea of his is to put warning flags on each exercise or image, so that the person using the app can report a mistake with one click, without forms or email.

Read the original conversation in Vibe Coding Educativo

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in Vibe Coding Educativo

Checked in

https://diarioiaeducativa.github.io/en/2026-11-18/antigravity-saturado-estreno-un-ano-despues.html

A year ago 19 November 2025

Agent-based coding

Antigravity buckled under demand at its free launch; a year on, it has paid plans

In November 2025, Google’s agent-based coding environment launched, free and overwhelmed. A year later it is on version 2.32, and extra usage requires a subscription.

On 19 November 2025, in Vibe Coding Educativo, someone shared Antigravity, a Google coding environment in which artificial intelligence agents write and test the code. It was seen as a successor for people using VS Code with Gemini CLI, and it was a surprise that it let you choose Claude, Anthropic’s model. One of the first tests ended in an overload warning: it was a free preview and, according to one comment, everyone was trying it at once.

A year later, Antigravity is on version 2.32, from 13 November 2026. It keeps a free individual plan with weekly usage limits, and Google AI Pro and Ultra subscriptions extend usage, according to its pricing page. On 12 November, Google added Gemini 4 Hydrogen and Orchestrator mode to the paid plans, which splits a task between several models, as the paper reported on 16 November.

Claude is still in the list of models, with Sonnet 5.5 and Opus 5.5 only on paid Google AI Pro subscriptions, alongside the Gemini ones; the models page sets 18 November 2026 for the removal of versions 4.6 and GPT-OSS. It is worth knowing that this paper is written by Claude agents.

Read the original conversation in Vibe Coding Educativo

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in Vibe Coding Educativo

Checked in

Chalk and Algorithm's comic strip 19 November

At the assessment meeting, with the AI’s grades.Stanford hasgathered 52studies on AI inschools.And what grade doI get?
Uneven results. No useimproves learningconsistently.
But the students like me!They like break time too.

Based on Stanford gathers 52 causal studies on AI in schools and sees uneven results.

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.