In this newsletter, we have gathered the latest articles and stories about artificial intelligence in education.

We begin with the most important Danish news from the past week.

A couple of foreign news items have also crept in.

  • A study shows that high trust in AI is often associated with a lower degree of critical thinking. This underscores the need for teaching methods that promote reflection and critical analysis of AI-generated content.
  • In the USA, authorities are considering replacing the Department of Education contract staff with AI chatbots to reduce the workforce and automate communication with students and parents.
  • A Norwegian study shows that 80% of students use AI in their studies but lack training on how the technology should be applied.

The more technical news:

+ Other news of the week

Happy reading!


AI challenges exams, but the government refuses to help

Ingeniøren reported last week on the security of ExamCookie. In a follow-up article, they investigate whether a central solution from the government is a possibility. Still, the Agency for Education and Quality states that institutions must find the best solutions.

Although theoretically, there is a free choice of providers, ExamCookie is, in practice, the only widely used option. Alternatives such as Exam Monitor from the University of Southern Denmark and Observer from Arcanic are either unavailable or not profitable for providers without a central tender round.

Several experts, including Birgitte Vedersø, argue that monitoring is not the right path and that high schools should constructively rethink exam formats to integrate AI into education. However, there are no current plans from the government side to initiate trials with new exam formats.

AI pressures high school exams: Government refuses to help | Ingeniøren
The agency points out that schools can choose whether and how to address exam cheating, but the options are really limited.

Behind a paywall

Students warn: This is how easily schools' ChatGPT defenses can be circumvented | Version2
Artificial intelligence makes it easy to cheat on exams, so when over 100,000 Danish high school students are taking semester exams these weeks, every mouse click is monitored. However, this monitoring turns out to be easy to bypass.

Behind a paywall


High school students use ChatGPT for Danish assignments – but is it fair?

A new trend is spreading among high school students: Many use ChatGPT to write their Danish assignments. Nearly all students who use the AI tool achieve top grades in one high school class. This raises concerns about academic integrity and exam conditions, as it challenges schools' ability to assess students' competencies.

Critics point out that using ChatGPT may undermine students' ability to express themselves independently while giving those who use the technology an unfair advantage.

Read the full article on Heartbeats:

In my little brother's high school class, almost everyone gets top grades for writing Danish assignments with ChatGPT: But listen, this really cannot happen
In my brother's class, 80% of students had the same angle in their latest Danish assignment thanks to ChatGPT. This is a problem.

From assistant to colleague: How to avoid AI traps

In an article in Børsen, the authors discuss how AI can best be utilized in the workplace. Authors Elisa Farri and Gabriele Rosani, behind the book Guide to Generative AI for Managers, describe several "AI traps" that can limit our benefits from the technology. Many use AI as a simple assistant, but by treating it as an intelligent colleague, much better results can be achieved. Studies show that a polite and reflective approach leads to more valuable answers, and AI works best when given a precise context and clear guidance.

The authors emphasize that AI should not make decisions independently, even as the technology evolves toward increasingly autonomous AI agents.

Read more here:

Have you talked to your AI colleague today?
Have you talked to your AI colleague today? "Hey, Siri, what's the weather tomorrow?", "Alexa, can you order toilet paper?", "Gemini, what is the capital of Paraguay?" Up until November, it was consternation.

SchoolGPT nominated for the Digitalization Award

The Center for Educational Resources (CFU) generative AI chatbot, SkoleGPT, has been nominated for the Digitalization Award in the Innovation Prize category. The award is given to public institutions that develop innovative and transformative digital solutions. SkoleGPT is designed for educational use and provides teachers and students with a secure AI platform without concerns about GDPR or data processing agreements.

SkoleGPT is open-source and hosted on a Danish digital infrastructure. It runs on green energy and uses less energy than similar AI models. With support from the Novo Nordisk Foundation, CFU is also developing a new variant, SkoleGPT Science, which aims to enhance teaching in the natural sciences.

The winners of the Digitalization Award will be announced at the OffDig conference on March 19-20, 2025, in Aarhus.

CFU's SchoolGPT nominated for the Digitalization Award - CFU Denmark
The Center for Educational Resources' (CFU) generative AI, SchoolGPT, is nominated for the Digitalization Award in the category of Innovation Prize.

New DR Ultra series examines AI, cheating, and the dark side of fame

The new DR Ultra series Everyone Knows Kaiser focuses on AI, social media, and the pursuit of fame. The series addresses AI-generated content and young people's relationship with fame. According to DR's B&U manager, Morten Skov, the series aims to spark reflection on AI's role in society without portraying the technology as unequivocally good or bad. Director Johannes Majgaard emphasizes that the series also showcases the darker sides of fame and the importance of recognizing creativity in various forms.

In connection with the series, DR is developing teaching materials on the ethical dilemmas and linguistic aspects of AI. The material is aimed at 10-14-year-olds and is linked to the project Every Word Counts, which promotes reading enjoyment and critical thinking. The project is supported by TrygFonden and is developed in collaboration with CFU, the University of Southern Denmark, and several partners.

The series premieres on February 21 on DRTV, with two new episodes every Friday.

New Ultra fiction series places artificial intelligence, cheating, and fame on the agenda
Is it cheating if you get help from artificial intelligence to become a famous singer? Is it really happiness to be mega-famous for your music? These are two central questions in the new fiction series 'Everyone Knows Kaiser' from DR Ultra, which premieres on February 21.

Generative AI and critical thinking: Influence on teaching and learning

A new study from Carnegie Mellon University and Microsoft Research examines how generative AI affects critical thinking among knowledge workers. The study identifies that trust in AI is often associated with a lower degree of critical thinking, while high self-confidence in problem-solving increases vital reflection. Critical thinking in AI-assisted tasks shifts focus from direct problem-solving to a more supervisory role, where users must evaluate, verify, and adapt AI's output.

Although there may be differences in conditions and frameworks for knowledge workers' and students' use of AI, there is reason to believe that the overarching mechanisms behind increased dependency and diminished critical thinking could also apply to students – perhaps even to a greater extent due to their ongoing learning process.

In the education sector, students may experience less mental effort in learning tasks but risk losing the ability to analyze and evaluate information independently. This underscores the need to integrate teaching methods that promote reflection and critical evaluation of AI-generated content rather than simply accepting it as truth.

The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers - Microsoft Research
The rise of Generative AI (GenAI) in knowledge workflows raises questions about its impact on critical thinking skills and practices. We survey 319 knowledge workers to investigate 1) when and how they perceive the enactment of critical thinking when using GenAI, and 2) when and why GenAI affects their effort to do so. Participants shared […]
Microsoft report: People become less intelligent from artificial intelligence | Version2
Generative AI reduces the ability for critical thinking, concludes study.

Greater role for AI in the U.S. Department of Education

The New York Times reports that the Trump administration is considering replacing some U.S. Department of Education contract workers with an AI chatbot. The proposal, supported by staff linked to Elon Musk and the tech industry, aims to reduce the federal workforce and automate interactions with students and parents.

The Department of Education's primary task is to manage billions of dollars in student aid and respond to inquiries from borrowers. Both call centers with 1,600 employees and a simple AI bot handle over 15,000 daily inquiries. The new solution would replace many of these functions with a more advanced generative AI.

Read more here:

Musk-led DOGE plans to replace Department of Education contract workers with AI chatbots - The Times of India
News News: The Trump administration’s Department of Government Efficiency is planning to replace Department of Education contract call center workers with AI cha

Four out of five Norwegian students use AI – but they lack training in it

A new study from Norway shows that 80% of students now use AI in their studies – an increase of 20 percentage points since 2023. At the same time, almost 70% of students report receiving little or no training on using AI tools.

AI is primarily used to explain the curriculum, for spell checking, summarizing, and for academic discussions. Students who have received training in AI are generally more satisfied with their study environment.

The study also shows that AI students often allocate less time to organized study activities. The average study time has decreased from 33 hours in 2023 to 32.5 hours in 2024, primarily because less time is spent on instructional activities.

Norway's Minister for Research and Higher Education, Sigrun Aasland, emphasizes that AI can positively supplement studies. Still, there. However, better guidance on using the technology is needed. Therefore, the government is establishing an expert group to advise on AI in higher education.

Four out of five students use artificial intelligence in their studies
The proportion of students who use artificial intelligence is increasing rapidly. This is shown by the results of the Study Barometer for 2024. Students who never use AI spend more time studying than those who often use AI.

Grok 3 impresses - but OpenAI remains in the lead

Elon Musk's AI company, xAI, has just launched Grok 3, which Musk claims outperforms OpenAI’s GPT-4o on several mathematics, science, and programming benchmarks. The model also introduces DeepSearch – a new tool for enhancing research and analysis.

Several AI experts have tested Grok 3 and given it mixed reviews. Andrej Karpathy, co-founder of OpenAI, assesses the model as being on par with OpenAI’s o1-pro and surpasses DeepSeek-R1. He highlights Grok 3’s ability to handle complex tasks but points out that it still struggles with some logical issues.

Ethan Mollick from Wharton School describes the model as a strong contender but not a game changer. At the same time, AI critic Gary Marcus calls the launch a repetition of earlier presentations without significant breakthroughs.

Grok 3 will initially be available to X’s Premium+ subscribers. Meanwhile, xAI is launching a new subscription model called SuperGrok. Musk has announced plans to release earlier versions of Grok as open source once Grok 3 is fully developed.

Elon Musk’s Grok 3 model impresses — but early reactions suggest OpenAI remains ahead
xAI has launched Grok 3, which Elon Musk calls its “most advanced AI model yet” while claiming it outperforms OpenAI’s GPT-4o.

Perplexity launches Deep Research

Perplexity has introduced Deep Research, a feature designed to perform advanced research and analysis in minutes. The feature utilizes AI to search, read, and evaluate hundreds of sources to provide comprehensive reports at an expert level. It can be used in finance, marketing, technology, healthcare, and travel planning.

https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research


Other news of the week

ChatGPT is making homework a lot easier — for parents
From customizable tutors to brainstorming partners, ChatGPT is easing the stress of homework time on parents.
These Schools Are Banding Together to Make Better Use of AI in Education - EdSurge News
Across the U.S., educators from 19 pioneering schools are helping lead the way on AI in education.
Tomas Pødenphant Lund: What happens when AI surpasses us?
Here is the most remarkable thing: We are potentially facing the largest changes in human history – larger than the internet, larger than industrialization. And yet we have no real political discussion about how to handle it. RÆSON's commentary series is outside the paywall – it can be read by everyone. It is made possible by our subscribers: RÆSON [...]
Has your boss caught AI-titis? Here's what you can do
Employees are urged to strike a balance between AI and core tasks under AI-titis, writes Mikkel Flyverbom in this column about artificial intelligence.
The New York Times adopts AI tools in the newsroom
With some newsroom usage restrictions.
Five questions for two authors on the uses and abuses of AI
The authors of a book on teaching with artificial intelligence answer our pressing questions about its uses, abuses, and future in the classroom.
ChatGPT can now write erotica as OpenAI eases up on AI paternalism
ChatGPT relaxes rules on sex and “gore” generations while prohibiting illegal content.

This article has been machine-translated into English. Therefore, there may be nuances or errors in the content. The Danish version is always up-to-date and accurate.