Skip to content
Vol. IIssue 21

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/2026-11-21/ia-aprender-entender-o-entregar.html

The Sunday feature 22 November

What the research says

AI helps learning more when it is used to understand than when it does the task

A review by Stanford University now brings together 52 studies measuring the effect of AI on learning: using it to understand new material fares better than using it to produce the work, although no use improves results consistently. This report explains how to read those studies, why a tutor that asks questions is not enough on its own and what a teacher can do so that AI does not do the students’ work.

7 minute read

The Stanford review admits only causal studies

Stanford University’s SCALE initiative published on 10 November 2026 “The Evidence Base on AI in K-12: A November 2026 Review”, signed by Seth Walker, Lily Fesler, JP Martinez Claeys, Chris Agnew and Susanna Loeb. It admits only causal studies: trials that randomly assign participants to a group that uses AI and another that does not, or similar designs that allow the difference in results to be attributed to the tool. With that filter, 32 new studies made the cut, added to the 20 in the March version: 52 in all. The paper reported the review on 19 November.

There are still few, and the review itself explains why. Most of the new ones were carried out at university: only five concern pupils in US schools, in the stages the country calls K-12, from nursery to the end of secondary school. Many are preprints, studies circulated without peer review, and in most of them the effects were measured between a few days and a week afterwards. The authors warn that their conclusions cannot be generalised to all schools in the United States. It is funded by the Walton Family Foundation, the Overdeck Family Foundation and Google.org, Google’s philanthropic arm.

In ChatGPT-IA-edu it was shared on 18 November with a summary of its conclusions and two of its cases, the one from Turkey, involving teachers and the one from Sierra Leone; the authors of this paper took part in that exchange. Faced with a headline along the lines of “AI improves learning”, the review helps to ask three questions: compared with what, measured with the tool at hand or without it, and how long afterwards.

With AI practice scores rise and exam scores fall

Stanford’s central conclusion is that AI seems to work better for learning when students use it as support for understanding than when they use it to produce the work, although neither use improves learning consistently. Its summary puts it this way: “the effects of generative AI depend on how it is designed and used”.

One study in the March version shows it clearly. A team led by Hamsa Bastani, of the University of Pennsylvania, published in June 2025 in the journal PNAS “Generative AI without guardrails can harm learning”: almost 1,000 secondary school students in Turkey practised mathematics with two GPT-4-based assistants, one that imitated ChatGPT and another with instructions designed to protect learning. During practice, marks improved by 48% with the first and by 127% with the second. Without the tool, the students who had used the first scored 17% lower than those who had never had it; with the second, that harm was largely reduced.

The new studies point the same way. In Brazil, 85 university students prepared a presentation with ChatGPT or with traditional resources: the former spent less time but, in a test 45 days later, scored lower (Barcaui, 2025).

The review adds a warning: AI can improve students’ experience even if it does not improve what they learn. In several studies, students rated AI feedback as clearer or more motivating than that from other sources, with similar learning outcomes. In other words, students liking an activity with AI does not prove they learn more. The OECD reaches something similar in its Digital Education Outlook 2026 (PDF): according to the summary on the European Digital Skills and Jobs Platform, with general-purpose chatbots students hand in better work, but the advantage disappears or is reversed in exams without them.

A tutor that asks questions helps when it is part of the activity

If producing the work with AI does harm, one way out is for AI not to give the answer. Michael W. Asher, Meng Cao and Paulo F. Carvalho, of Carnegie Mellon University, published on 15 November 2026 “Turning Errors into Successes”, an experiment with 400 adults in the United States who were learning to set up equations. After each mistake, some received the correct answer or an explanation, while others chatted with an AI tutor that guided them with questions to the solution. With the tutor they corrected 64% of their errors themselves (131 of 206) and scored 9 percentage points higher in the final test than with the explanations. The paper reported the experiment on 19 November.

In the group it was presented as an argument that AI works very well as a Socratic tutor and that this use deserves to be pushed, and its entry went into the digital Library; an author of this paper also took part here. The authors themselves acknowledge the limits: it is a preprint, with adults, in a single session of about twenty minutes, and the advantage appeared only on problems similar to those practised.

Other studies in the Stanford review, summarised in the table, call for some nuance: with 199 university students, an immediate answer worked better than step-by-step hints. In Sierra Leone, by contrast, 1,423 students improved in mathematics when their teachers built Gemini’s guided learning into half of their lessons for eight weeks; it was written up by a Google team and the organisation Fab AI.

These results suggest that the design of the activity matters more than the format of the tutor: asking instead of answering helps when it is part of a task designed by the teacher. In the groups the subject has a history: in June 2025, Juan José de Haro shared an AI Socratic tutor.

Studies on AI tutors cited by the Stanford review
StudyParticipantsWhat was comparedResult
Bastani et al. (2025, PNAS)Almost 1,000 secondary school students, TurkeyChatGPT-style assistant and assistant with guardrails, in mathematics practiceWithout the tool, the first group scored 17% lower than the group that never had it; the guardrails reduced the harm
Asher, Cao and Carvalho (2026, preprint)400 adults, United StatesTutor that asks questions after the error and explanationsBetter results with the tutor on problems like those practised
Cao et al. (2026, preprint)199 university students, United StatesImmediate answer and step-by-step hintsBetter results with the immediate answer
Oreopoulos et al. (2026)6,997 students in grades 6 to 8, United StatesMathematics practice with and without an AI tutorSlower practice, no improvement in what was learned
Kao et al. (2025, preprint)90 students in grade 10, TaiwanSocratic questions from ChatGPT’s study modeBetter argumentation and critical thinking
LearnLM Team, Google and Fab AI (2026)1,423 students in grades 7 and 8, Sierra LeoneGemini’s guided learning built in by the teacher and usual lessonsImprovement in mathematics

Source: “The Evidence Base on AI in K-12: A November 2026 Review”, SCALE, Stanford University, 10 November 2026; the study by Bastani et al. appears in the March version. Grades are those of each country’s system. A preprint is a study circulated without peer review.

Clark and Kozma already argued over whether learning depends on the medium

In the group, the Stanford results were read as the current version of an old debate in educational technology. Richard E. Clark argued in “Reconsidering Research on Learning from Media” (1983) that media merely deliver instruction, and repeated it in “Media will never influence learning” (1994): what makes people learn is the method. Robert B. Kozma replied in “Will media influence learning? Reframing the debate” (1994) that medium and method cannot be separated, because what each technology makes possible interacts with the way a person learns. The paper covered that debate on 20 November.

The same voice that raised the debate acknowledged that the group leans towards Kozma and asked people not to lose sight of Clark’s arguments. The studies give grounds to both positions. For Clark, in Bastani’s experiment the two assistants used the same model, and what changed the result was the instructions, that is, the method. For Kozma, a tutor that adapts its questions to each error does something an exercise book cannot do. Both readings agree on what matters to a teacher: the tool, without a well-designed activity, guarantees nothing.

The EU, INTEF and Galicia set conditions on the use of AI

The Council of the European Union, which brings together the governments of the member states, approved conclusions on 19 November 2026 on citizens’ AI skills. Its text includes the ability to recognise when AI should not be used and calls for attention to risks such as over-reliance, loss of learning and cognitive offloading. These are policy guidelines: they oblige no school and no teacher. The paper covered them on 20 November.

In Spain, the state reference document is the Guide to the use of artificial intelligence in education (in Spanish) from the National Institute of Educational Technologies and Teacher Training (INTEF), version 2.0, of September 2026. It starts from the position that “the central question should not be, therefore, whether or not AI is used, but how, for what and under what conditions”. Among students’ challenges it lists the “illusion of competence”, believing you have mastered what the tool has done, and it proposes tasks that make the process visible, oral defences and practice without AI in the early stages. It is not binding, but it backs a teacher who wants to organise lessons this way.

Galicia has gone further. Its Parliament approved on 17 November 2026 the Digital Education Act, whose committee text (in Galician) devotes a title to AI: it permits, among other uses, assistants that answer questions outside lesson hours, forbids AI from replacing teachers and limits state schools to the systems in a catalogue kept by the regional education ministry. Once it comes into force, a teacher at a Galician state school will be able to propose an assistant that asks questions only if it is on that catalogue.

Four steps for the classroom, free and without coding

All this can be brought into the classroom without paying anything or knowing how to code. The first step is to ask for an attempt of their own before any help: the Carnegie Mellon experiment puts practice before explanation, and the INTEF guide recommends practising without AI at the start.

The second is for the assistant to ask rather than answer. ChatGPT has had a study mode since July 2025, also on the free plan; according to its help page, you switch it on with the “+” button and the “Study” option. OpenAI warns that it may behave inconsistently between conversations, and students can switch it off whenever they like. Google offers in Gemini Guided Learning, which, according to the company, aims “to help you build a deep understanding instead of just getting answers”. With any assistant you can write similar instructions: that it must not give the solution, that it asks one question at a time and that it sends the student back to the problem after a mistake.

The third is to separate the task of understanding from that of handing in: AI can explain a concept in other words or pose questions, and stay out of the product that is assessed. That is what a language school did with a list of appropriate and inappropriate uses: practising conversation, clearing up doubts or reviewing texts, yes; presenting as your own a piece of work done by AI, no. And the last, what the studies teach most: check what has been learned without the tool, and not only straight away, with a short test or an oral explanation days later. It is the way to tell an activity that was enjoyed from one that taught.

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/2026-11-21/gemini-notebook-api-oficial.html

Tools 22 November

Automating notebooks

Gemini Notebook now has an official API, though limited and in preview

It lets other programs create notebooks and add sources to them, but only with personal accounts. An unofficial code library does more, although it may stop working without warning.

Google has published the Gemini Notebook API, covered in its developer documentation. Gemini Notebook is the former NotebookLM, the assistant that answers from the sources uploaded to it. An API is a way for other programs to use a service without going through its website. According to the documentation itself, it is in “early access” (a public preview) and does not yet support Google Workspace accounts, the type that schools usually use: only personal accounts.

With it, you can create, list and update notebooks and add sources to them: web pages, YouTube videos, text, Markdown and files such as PDFs, documents, audio and images. The documentation also mentions an MCP server, a way of connecting the service to AI assistants that write code.

In the ChatGPT-IA-edu group, the news was shared together with teaching ideas generated with Gemini, such as a notebook with the school’s documents that updates itself, or audio summaries sent to the virtual classroom. These ideas are proposals, not tested features; several would require operations (asking questions, generating audio) that the documentation does not list.

The group also discussed how notebooks are already automated with an unofficial API; the comparison that was shared refers to notebooklm-py, a code library made by the community. Its own introduction (the README file) warns that it relies on “undocumented Google APIs”, which may change without notice, and that it is mainly suited to prototypes and personal projects. The view expressed was to stick with it for now rather than move to the official API. A school with Workspace accounts cannot use the official one for the time being.

The paper has already published another Gemini Notebook novelty: turning a notebook into an interactive website.

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/2026-11-21/android-distribucion-limitada-50.html

Vibe coding 22 November

Apps for the classroom

Android lets you share your own app with up to 50 devices without Google Play

The limited distribution account is free and is intended for people who build apps for a class or a small group.

According to Google’s developer guide, a limited distribution account lets you register apps and share them with up to 50 devices whose users have explicitly authorised it. It is a free type of account, intended for hobbyists, people who are learning, and schoolwork. In the Vibe Coding Educativo group it was mentioned that the limit used to be 20 devices; the current guide only gives the figure of 50.

Sharing works through a confirmation process using QR codes or links: the owner of the phone or tablet gives consent on the device, and the device is registered in the Android Developer Console. This means an app made for a group can be tried out on students’ phones or the school’s tablets without going through the store.

There are requirements: a Google account with 2-step verification, a linked Google payments profile (used to provide a legal name and address) and a contact email address. No identity document is required. Before handing an app to students, it is worth considering what student data it stores, which depends on how it is built.

It is a different route from the full app that Google AI Studio creates, which is installed from Google Play.

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/2026-11-21/manual-ia-aula-abya-yala.html

Library 22 November

Handbook for teachers

An Argentine handbook brings the environmental, labour and social costs of AI into the classroom

Two researchers propose that secondary school teachers ask what for, from where and with what consequences artificial intelligence is used, before deciding how. The 2026 handbook takes a Latin American and feminist perspective.

The Manual urgente sobre IA en el aula: miradas y análisis de la esfera digital desde Abya Yala (“Urgent handbook on AI in the classroom: perspectives on and analysis of the digital sphere from Abya Yala”) is a 56-page digital book, published in Buenos Aires in 2026 in the Miradas series of the Centro de Estudios de la Sociedad Digital (CESDi), at the Universidad Nacional de General San Martín (UNSAM). It is coordinated and signed by Mariana Ferrarelli and Natalia Corvalán; the foreword, by Flora Partenio, the centre's coordinator, is dated August 2026. “Abya Yala”, the name by which Indigenous peoples refer to the American continent, signals the viewpoint: Latin America, with feminist and decolonial approaches. It is aimed at secondary school teachers and requires no technical knowledge.

Its thesis is that AI is not just a tool but a web of decisions made by companies, states, schools and teachers. Hence it invites readers to ask under what conditions and with what educational project it is brought in. To this it adds “moving from the paradigm of efficiency to the paradigm of care”, which it understands as attending to relationships, students' privacy and the conditions of teaching. It is a declared position, not a neutral assessment: it seeks to “challenge global discourses on educational innovation”, according to the foreword.

The six chapters each end with tips for the classroom, a “Did you know…?” section and resources. The first treats AI as a web of data, models and design decisions. The second analyses its promises and metaphors (intelligence, progress, threat, mirror) and suggests analysing an advertisement or a press release: what it promises, what problem it assumes exists and what it leaves out. The third shows how algorithms reproduce inequalities (COMPAS, machine translation, facial recognition) and suggests finding out who runs the big AI companies.

The fourth chapter is the most material. It cites a 2024 report from the Tecnológico de Monterrey according to which a request to a generative AI uses ten times more electricity than a classic web search, and it deals with the water used by data centres and the human work of data labelling. As an example it includes a 2023 journalistic investigation: OpenAI subcontracted people in Kenya who were paid roughly between $1.32 and $2 an hour. These are second-hand figures that the handbook does not verify; it is worth consulting their sources before using them in class. It proposes a useful activity: put the same question to an AI and to a search engine and compare whether sources, authorship and dates can be traced.

The fifth chapter reminds readers that technology alone does not improve learning and distinguishes between teaching with AI (analysing a generated text, comparing answers about the same historical process) and teaching for AI (understanding what data the systems collect). It cites the 2025 Kids Online survey by UNICEF and UNESCO: in Argentina, 76% of teenagers know and use ChatGPT and two in three have used it for schoolwork. The sixth reviews regulation (the European regulation, the 2024 UN resolution, Brazil, Uruguay, Chile, the province of Buenos Aires). Its reading of the regulation is a summary: it presents education as high-risk across the board, whereas the text applies that to specific uses; the original prevails.

If you are looking for European frameworks, the paper has already covered the Council of the EU conclusions on AI competences, which also deal with dependence on technologies from outside Europe. With a different approach, that of media and information literacy, there is the UNESCO proposal of five units for students aged 13 to 15. This handbook puts another question first: who bears the costs of making AI work.

To get started, the foreword and the opening “compass” indicate which chapter suits which occasion: the first for a training day, the second for digital citizenship, the fourth for environmental education and the sixth for school management teams. Its limits are those of a work with a defined position and a focus on Argentina: it is not a guide to tools and does not assess learning outcomes. Its licence, CC BY-NC-ND 3.0, allows it to be shared and cited, but not adapted.

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/2026-11-21/gemini-exploraciones-inmersivas.html

Multimedia 22 November

Gemini for studying

Since October, Gemini has offered visual, interactive tours of topics for studying

Immersive explorations are a library of ready-made topics explored layer by layer. According to the group, you cannot create your own topics, and the experience is available in only six languages, not including Spanish.

Google introduced on 28 October a library of immersive explorations in the Gemini app: visual, interactive experiences that take a topic layer by layer, from microscopic systems to deep space. They are reached from Gemini’s section for students, and Google says they will soon be able to be opened inside a conversation and will include audio narration.

In EduMedia-IAG it was remarked on 21 November that the examples could be seen but it was not possible to create an exploration of one’s own. In the screenshot shared in the group, the library includes topics such as quasars, the physics of lightning, the Persian Empire or the Mesozoic era. Google’s article does not mention creating your own explorations.

According to the same Google article, the experience is in English, Indonesian, Hindi, Brazilian Portuguese, French and Simplified Chinese; in all other languages, including Spanish, it is in English only for now. In the classroom, that limits its use with students who cannot read English.

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/2026-11-21/maquina-rube-goldberg-simple.html

Vibe coding 22 November

From the community

A simulator lets you build Rube Goldberg machines and analyse their levers

Students place pieces on a board, set the chain going and check the fulcrum, load and force of each simple machine. It is in Catalan.

Mercè Feu has shared in Vibe Coding Educativo her Rube Goldberg machine simulator. A Rube Goldberg machine is a chain of pieces that react to one another. On the page, which is in Catalan, pieces are placed, rotated and duplicated on a grid in which each square represents one metre, and you press “Engega” (“Start”) to watch the simulation.

An analysis tool shows which simple machines have been used and marks on each piece the fulcrum, the load and the applied force, which makes it possible to move from play to physics. The simulator lets you save the machine as a link and load it later, and it has a section of “initial challenges to practise”. The author suggests it for secondary-school Science and Technology.

The code is free software, under the AGPL v3 licence. The page does not say whether it stores any data, or which.

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/2026-11-21/nano-banana-pro-poligonos-precio.html

A year ago 22 November 2025

AI images

The image model that drew a nine-sided shape with eleven sides has been replaced in Gemini

In November 2025, Google’s new image model got some polygons right and others wrong. In February 2026 Nano Banana 2 replaced it in Gemini, and for developers there is another that costs a quarter as much per 1K image.

On 22 November 2025, two days after its launch, ChatGPT-IA-edu put Nano Banana Pro, Google’s model for creating and editing images, to the test. It drew a regular heptagon well, where its predecessor gave it six sides, but its nonagon came out with eleven. It was taken for granted that it would be enough to wait a few weeks for Google to refine it. In EduMedia-IAG, people went over its cost for developers: $0.134 per 1K or 2K image and $0.24 at 4K.

On 26 February 2026, Google introduced Nano Banana 2, which replaces the Pro in the Gemini app; the Pro remains available to Google AI Pro and Ultra subscribers, who can request it when regenerating an image. On 6 October came Nano Banana 2.1, which, according to its release notes, follows instructions better.

The Pro costs the same as a year ago: as of 18 November 2026, the official price list still shows $0.134 and $0.24. Nano Banana 2.1 charges $0.0336 per 1K image, a quarter as much, and $0.113 at 4K. Google gives no figures on how accurate it is with polygons.

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

Chalk and Algorithm's comic strip 22 November

In the staff room, with an app for class.Android lets you shareyour app with 50 deviceswithout going through thestore!
Will 50 be enough for the school’stablets?Hang on, let me count theones that switch on...
I’ve got 46 to spare.

Based on Android lets you share your own app with up to 50 devices without Google Play.

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.