Students aren't dodging the work; they believe they've done it. This week's newsletter is about the cognitive debt that arises when AI does the thinking first.
Here is this week's newsletter on artificial intelligence in education, where we gather the articles and news we have selected this week.
In Kristeligt Dagblad, there is a reference by Associate Professor Kristine Bundgaard to a study from Danmarks Evalueringsinstitut (the Danish Evaluation Institute), which shows that one in three upper secondary school (gymnasium) students use AI daily. 40 per cent of AI users are worried that they will become lazy, and 26 per cent fear that it will become harder to immerse themselves in texts. Students can feel that something is slipping.
This week I focus on the question: What happens when students never build the academic foundation they will later need to evaluate their AI results?
In the same article, Thomas Conradsen and Morten Mikkelsen talk about students who use AI to produce summaries of long texts instead of reading them themselves. Senior researcher Jesper Bremholm from Nationalt Videncenter for Læsning (the National Knowledge Centre for Reading) calls it a real threat to the ability to read and analyse, and gymnasium teacher Flemming B. Olsen argues that schools should drop homework altogether and move the work into the classroom.
In the United States, a coalition of 250 American and Canadian experts demands a five-year pause in the use of generative AI in schools from pre-school to 12th grade. Emily Cherkin from the University of Washington captures the core point. The most concerning aspect of AI in school is not that children outsource tasks to the technology. It is that they never get to build the skills they could later choose to let AI handle. The same observation is made by US writing professor Kristi Girdharry, who in The Conversation describes how students cannot tell when they are taking mental shortcuts in their own thinking.
It may therefore sound like a contradiction when Jan Simonsen, Country Manager for Salesforce Danmark, writes in Børsen, that judgement will become the most important competence in a time when AI can solve tasks instantly. AI researcher Ethan Mollick also argues that AI does not remove jobs but shifts what it is valuable to be able to do. Judgement is built through years of working with material, formulations and problems, and a joint study from MIT and Harvard shows how the use of AI accumulates cognitive debt and weakens independent thinking over time. If the foundation crumbles, judgement has nothing to stand on.
AI-Kompetencerådet (the AI Competence Council) reports that 314,000 Danes have built AI skills in a year. But what happens if students never get to build the foundation?
In this week's commentary I return to Emily Cherkin's point. For it is not AI that determines whether students develop judgement. It is us.
Happy reading.
In Kristeligt Dagblad, Thomas Conradsen and Morten Mikkelsen write that gymnasium and university students are increasingly using AI to produce summaries of long texts instead of reading them themselves.
The article is part of the series Fremtidens læsere.
Mikkel Aslak Koudal Andersen, technology understanding consultant at VIA University College, is convinced that the use of AI-generated summaries is very widespread across both youth education programmes and higher education. He believes the education system is groping in the dark, and that one consequence may be more oral exams, because it is easier to assess whether the student has actually understood the material.
Kristine Bundgaard, Associate Professor of Language Technology at Aalborg Universitet, points to a report from Danmarks Evalueringsinstitut showing that one in three gymnasium students use AI daily. 40 per cent of AI users are worried that they will become lazier, and 26 per cent fear that it will become harder for them to immerse themselves in texts. She also points out that AI tools can omit significant parts or hallucinate and thereby produce incorrect information.
Jesper Bremholm, senior researcher at Nationalt Videncenter for Læsning, sees the growing use of AI-generated summaries as a real threat to the ability to read and analyse texts at gymnasium and university level. When students settle for a page of key points instead of reading the text themselves, this can, in his view, put their critical sense under pressure.
In an opinion piece in Gymnasieskolen Flemming B. Olsen argues that schools should break with the traditional homework culture and move tasks into the classroom.
His starting point is the pressure on children and young people’s mental health, but the argument is sharpened by the increasing spread of AI in schoolwork. When students sit alone at home, AI, he says, becomes a tempting shortcut to quick results without real immersion, and this creates a significant learning loss while also making it impossible for the teacher to assess the student’s actual development. His solution is to move the work into the classroom, where the teacher can guide students in the use of the technology and ensure that AI is used as a constructive sparring partner rather than a cheating tool.
Flemming B. Olsen also points out that a homework-free school is a step towards greater social equality, because academic support comes from trained professionals rather than from parents’ resources or access to advanced digital tools at home.
Flemming B. Olsen
In Fortune Catherina Gioino writes that a coalition of more than 250 experts and organisations in the USA and Canada is calling for a five-year moratorium on all student-facing generative AI products in schools from pre-school to 12th grade. The coalition is led by the Boston-based children’s rights organisation Fairplay, and products that do not pass safety tests during the pause should, according to the proposal, be banned permanently.
Emily Cherkin, screen time consultant and instructor at the University of Washington’s Evans School of Public Policy, points out that generative AI is an untested product being given to children in the name of education, without the benefits being documented. Her central point is that the problem is not only that children cognitively outsource tasks to AI, but also that they never build the skills they could otherwise later choose to outsource.
The report refers to a joint study from MIT and Harvard showing that the use of AI accumulates cognitive debt and weakens independent thinking over time, as well as to OECD research indicating that students who use ChatGPT as a study tool perform worse in exams than students without access.
The coalition also points to a structural contradiction in the sector. AI companies themselves exclude minors in their terms of service, yet simultaneously market the products to schools. Anthropic’s terms exclude users under 18 years of age.
Nevertheless, MagicSchool AI, one of the most widely used K-12 platforms in the USA, is built on Anthropic’s models. The five-year pause is intended to allow time for independent third-party audits, a public register of AI tools in schools and new regulatory frameworks which, according to the coalition, do not yet exist.
Catherina Gioino
In The Conversation, Kristi Girdharry writes about her experiences with generative AI in teaching since 2023. She describes herself as neither for nor against AI and points out that even the most powerful tools are only as good as the learning environments we create around them.
She writes that students often cannot sense when they are taking mental shortcuts in their own thinking. She refers to a study from the British Journal of Educational Technology showing that students who used ChatGPT improved their essay grades in the short term, but did not achieve real knowledge gains. The researchers called the phenomenon metacognitive laziness.
Her role has changed from being a neutral co-learner to being a guide with an academic standpoint. Kristi Girdharry deliberately asks students to write without AI, not as a purity test, but because they can only understand what AI does to their thinking when they know what it can do unaided.
Kristi Girdharry
In an opinion piece in Børsen (the Danish business daily), Jan Simonsen, country manager for Salesforce Denmark, writes that AI is taking over a range of tasks that recent graduates typically handle at the start of their careers.
His point is that this does not make recent graduates redundant, but it does place new demands on how businesses think about roles and recruitment. The route into the labour market is changing, and business leaders have a responsibility to actively create new positions that combine human judgement with AI capability.
He observes that many job advertisements still call for classic competencies such as communication and critical thinking, while newer skills such as AI fluency are not mentioned. According to him, AI fluency is about using AI as a thinking and sparring partner that challenges assumptions and opens up strategic opportunities one would not otherwise have seen.
His core message is that judgement will be the most important competence when AI can perform tasks almost instantly. It is about assessing tone and brand in AI-generated communication, identifying gaps in automated analyses and asking follow-up questions that open up new strategic possibilities.
According to Jan Simonsen, the most valuable employees will not be those who work fastest, but those who know when to follow AI and when to challenge it.
Jan Simonsen, landechef, Salesforce Danmark
In Berlingske (a Danish newspaper), Christoffer Løntoft Baun writes about the service RentAHuman, where AI systems can hire real people to carry out tasks in the physical world. Over 600,000 people are signed up to the service, and the article uses it as a springboard for an interview with AI researcher Ethan Mollick from the Wharton School.
Ethan Mollick believes the debate overdramatizes the development. He rejects the notion that all work will disappear, noting that neither the steam engine, the pocket calculator nor the internet led to fewer jobs.
His central point is that AI shifts what is regarded as valuable competence. It is no longer especially valuable to be able to write very basic code, but that does not destroy the programming profession. It changes what a programmer needs to be able to do.
Christoffer Løntoft Baun
In Børsen (the Danish business daily), the AI-Kompetencerådet (the Danish AI Competence Council) writes that 314,000 Danes have acquired AI skills over the past year. This means that one third of the goal of upskilling 1 million Danes by the end of 2028 has been reached.
According to Lilian Mogensen, chair of AI-Kompetencerådet, around six out of ten upskilling efforts have taken place among digital generalists – that is, employees who use AI in their daily work – rather than among technical specialists.
Lilian Mogensen emphasises that upskilling no longer happens via general introductory seminars, but is instead directly tied to each employee’s work tasks. At the same time, she warns that Denmark risks creating a digital A- and B-team if more groups do not gain access to AI skills.
Thomas Zigler
In this newsletter, I noticed Kristi Girdharry’s article, where she asks students to write the first draft without AI. It’s not because she is opposed to AI, but because students can only sense what AI does to their thinking when they know what they can do without it.
Here we touch on what is creating friction between teachers and students right now: a professional grief. The teacher has known a world without AI and can feel the difference between something that has been thought and something that has simply been fetched.
I have previously written about students taking the easy route. Professor Thomas Hildebrandt put it more sharply in a comment on LinkedIn, than I myself had done. In many cases, generative AI has not removed the hurdle. It has made it possible to convince ourselves and others that we have got over to the other side without that actually having happened. That is a more serious diagnosis. What we are seeing in classrooms is not students cheating their way out of the work. It is students who believe they have done it. They feel they have learned something. They feel they have improved. The prompt comes before the thought, and the result looks like something that has been thought.
This is where cognitive debt arises. Not as an absence of work, but as the illusion of work.
Many now say that judgement will be the most important competence in an AI era. That includes, among others, Jan Simonsen in Børsen and Ethan Mollick from Wharton. But judgement is not just critical thinking as a method. It is the ability to sense that there is anything to think critically about in the first place. That ability does not arise spontaneously the day you need it. It is built up over years of slow reading, bad drafts and returning to what you did not understand the first time.
The most worrying thing about AI in school is not that students outsource tasks to the technology. It is that they never get to build the skills that they could later choose to let AI handle. The debt does not show up on the report card. It does not show up in the assignment. It only reveals itself where judgement has to be used and there is no foundation to lean on.
Think five years ahead. A recent graduate is sitting with an AI-generated draft analysis and has to judge whether it stands up. But how is she supposed to feel that the analysis is shaky if she has never herself written a sentence that is shaky?
That is why it becomes a choice the teacher has to make. Not about whether AI should be in the classroom. It already is. But about when the brain should work alone, and when AI may come in. Some things must be without AI, and some must be with it. It cannot be either/or. That is where the hurdle needs to be raised again, deliberately and didactically, because it will not raise itself.
The question is not whether we should use AI. We do. The question is whether we ensure that there is still something beneath the technology that can evaluate it.
