This week's news: AI boosts teacher productivity, but not everyone is on board

This week we have gathered some of the latest news on artificial intelligence and education. Let’s start with a new study showing that AI can improve teachers’ productivity, but not everyone is making full use of its potential. The study indicates that the biggest gains are achieved when AI is used for strategic planning rather than merely to automate tasks. In the same vein, another study reveals that as many as 94% of AI-generated exam submissions remain undetected, even when evaluated by experienced educators. This naturally raises questions about the integrity of our current assessment systems.

At Viden.AI, we have published the first three parts of our article series on professional grief, and last week we published the article focusing on the sense of powerlessness that can arise when students use generative AI in lessons.

For a broader perspective on AI’s impact on society, we recommend the new documentary “AI på liv och död” on SVT Play. The documentary gives quite a good insight into how AI affects different areas, from the military to healthcare and social media. We hope it will be shown on Danish TV at some point.

AI researcher Serge Belongie from Københavns Universitet (the University of Copenhagen) warns that popular AI tools such as ChatGPT may dilute Danish culture by favouring American values and narrative traditions. At the same time, a study from MIT shows that the use of artificial intelligence can hamper learning if it is used as a shortcut to problem-solving.

When it comes to the use of ChatGPT in the workplace, men are in the lead, while women and older employees lag behind, often due to a lack of training. Researchers urge employers to offer targeted AI training to reduce the gender gap.

Finally, it is worth noting that AI’s environmental footprint is still very large, and it does not appear to be shrinking. AI models require enormous amounts of energy, and estimates suggest that a single ChatGPT query uses as much electricity as 40 mobile phone charges. With data centre energy consumption expected to double by 2026, the tech giants face a significant challenge to their climate targets.

Apologies that the newsletter ended up so long, but happy reading!

AI can improve teachers’ productivity, but not everyone is making full use of its potential

A new study on the use of generative AI among US K-12 teachers shows that AI tools such as ChatGPT can improve both workflows and productivity. The study followed 24 teachers over the 2023-2024 school year and identified three groups of teachers:

1. Teachers who used ChatGPT both to generate teaching materials and to get ideas for lesson plans.
2. Teachers who primarily used ChatGPT to help with concrete tasks such as creating quizzes, but not for planning.
3. Teachers who did not use ChatGPT or other AI in their teaching at all, often due to ethical concerns or a lack of time to get to grips with it.

The findings indicate that teachers who used AI to generate ideas and input for planning experienced the greatest productivity and higher quality of work. By contrast, teachers who primarily used AI to create materials did not report the same productivity improvements. Those who achieved the biggest gains integrated AI as part of their “backward planning”, starting from the learning objectives and working backwards to identify relevant tasks.

At the same time, the study shows that a significant number of teachers either avoid using generative AI due to ethical concerns or because they are unsure about the technology’s role in their teaching. Some were worried that AI would encourage cheating among students, while others lacked the time to familiarise themselves with the new tools.

Read the study here:

Backwards Planning with Generative AI: Case Study Evidence from US K12 Teachers

Backward planning is an effective and efficient operational process when working towards a goal: work backward from the desired outcome to figure out the steps

See all articles by Samantha Keppler

AI in university exams: 94% of AI-generated submissions remain undetected

In another academic study, 94% of AI-generated exam submissions went undetected, even when assessed by experienced educators. The research, carried out at a reputable UK university, tested AI in the education system by submitting 100% AI-written submissions in five undergraduate modules across all years of study in a psychology degree. The results showed that, on average, the AI submissions received grades about half a grade boundary higher than real students achieved. In 83.4% of cases, the AI submissions outperformed a random selection of the same number of real student submissions.

Read the study here:

A real-world test of artificial intelligence infiltration of a university examinations system: A “Turing Test” case study

The recent rise in artificial intelligence systems, such as ChatGPT, poses a fundamental problem for the educational sector. In universities and schools, many forms of assessment, such as coursework, are completed without invigilation. Therefore, students could hand in work as their own which is in fact completed by AI. Since the COVID pandemic, the sector has additionally accelerated its reliance on unsupervised ‘take home exams’. If students cheat using AI and this is undetected, the integrity of the way in which students are assessed is threatened. We report a rigorous, blind study in which we injected 100% AI written submissions into the examinations system in five undergraduate modules, across all years of study, for a BSc degree in Psychology at a reputable UK university. We found that 94% of our AI submissions were undetected. The grades awarded to our AI submissions were on average half a grade boundary higher than that achieved by real students. Across modules there was an 83.4% chance that the AI submissions on a module would outperform a random selection of the same number of real student submissions.

PLOS ONEPeter Scarfe ,

Series on professional grief

At Viden.AI, we are in the process of publishing a series of articles on professional grief, and last week we published an article about the sense of powerlessness that can arise when students ask about or (mis)use generative AI in lessons. We have received quite a lot of positive feedback that many recognise themselves in all the situations mentioned. The next article will be published on 29 August 2024.

Articles in the professional grief series:

• Article 1: When artificial intelligence changes the teacher’s reality
• Article 2: AI’s impact on the teacher’s professional identity
• Article 3: When generative AI creates feelings of powerlessness
• Article 4: to be published on 29 August 2024

New documentary: AI på liv och död

Inga Strümke hosts a new documentary on SVT Play that offers a timely and nuanced insight into how AI affects everything from the military to healthcare and social media. Through a series of examples and interviews with experts, the documentary sheds light on the opportunities and challenges we face as a society.

The documentary focuses on a number of important questions: What does the increasing use of AI mean for us as individuals and as a society? How can we harness the potential of AI while addressing the ethical dilemmas that come with it?

You can stream the programme in Denmark, but unfortunately there are no Danish subtitles.

AI på liv och död

We follow the Norwegian AI researcher Inga Strümke into the world of AI, behind the scenes with the military and to Facebook’s robotics lab. In the wars in Ukraine and Gaza, AI programs sometimes decide which people are to be attacked, and on social media disturbing images are shown to children who cannot stop scrolling. At the same time, AI can save lives in healthcare and in the underground. But researchers now want to set limits on how the tech giants use the new technology.

SVT PlaySveriges Television AB, Stockholm, Sweden

Artificial intelligence may dilute Danish culture

Videnskab.dk has published an interesting article discussing the challenges of generative AI when language models are primarily trained on American data.

Serge Belongie fra Københavns Universitet (University of Copenhagen) points out that popular AI tools such as ChatGPT often favour American culture and values. For example, ChatGPT’s narratives may reflect American moral values rather than Danish or Nordic storytelling traditions. Over time, this can lead to a dilution of cultural nuances that are otherwise crucial to our sense of identity and shared understanding.

Søren Hauberg fra DTU (Technical University of Denmark) notes that the fear of American cultural dominance is not new, but that AI technology amplifies this challenge by tailoring content and recommendations to the individual user, which can lead to cultural echo chambers. Belongie is working on a model designed to preserve cultural differences by offering AI users a global perspective on values and attitudes.

Read the article here:

Disney or H.C. Andersen? Artificial intelligence risks diluting Danish culture

Our culture is threatened by artificial intelligence because the technology is primarily built on American data, says an AI researcher.

Videnskab.dkSee Ida’s profile

Can artificial intelligence hinder learning?

A study at MIT illustrates how the use of artificial intelligence can negatively affect learning. Students were divided into three groups to solve a task in an unfamiliar programming language. One group used ChatGPT, another used Meta’s Code Llama, and the third group used only Google. The ChatGPT group completed the task fastest, while the Google group was slower, as they had to analyse the task in smaller parts.

When students were subsequently tested without aids, the results were turned on their head. Those who had used ChatGPT remembered almost nothing and failed. Half of those who had used Code Llama passed, while all who had used only Google did well. The study indicates that rapid problem-solving with the help of AI can undermine deeper learning and understanding.

However, AI can also be used productively in learning contexts if it is used as a tool that stimulates reflection and provides rapid feedback. The challenge is to find a balance so that AI does not become a shortcut that hampers deeper understanding, but instead serves as a catalyst for learning.

Debate: Does artificial intelligence make us dumber?

Artificial intelligence in the form of large language models, such as ChatGPT, can seem like a superpower and help you deliver a better product. But you must take care not to become dumber in the process.

DNAnders Malthe-Sørenssen

The Impact of AI on Computer Science Education – Communications of the ACM

Esther Shein

Men dominate the use of ChatGPT in the workplace

According to a large Danish survey of 100,000 employees, women are hesitant to use AI tools such as ChatGPT in their work. Younger men in particular are leading the way, while women and older employees lag behind, often due to a lack of training. However, the women express a desire to use the technology, but point to insufficient training as a barrier.

The researchers behind the study, Anders Humlum and Emilie Vestergaard, emphasise that employers can play a central role in reducing the gender imbalance by offering targeted training in AI. At the same time, they warn that technological advances such as ChatGPT have historically exacerbated inequalities unless active steps are taken to include all employees.

In addition to the gender differences, the study shows that the use of ChatGPT can increase both pay and efficiency, but that this requires companies and employees to rethink workflows to fully exploit the potential of AI technology.

Read more below

More men than women use ChatGPT in the workplace

Men are overrepresented when it comes to the use of the AI chatbot ChatGPT in a work context. This is shown by a new questionnaire survey of 100,000 employees in the Danish labour market, led by assistant professor at the University of Chicago Anders Humlum and PhD student at Københavns Universitet Emilie Vestergaard. See also: OpenAI is concerned: People develop feelings for ChatGPT 20 percentage points more likely [...] Read more…

Mobilsiden.dkJohan Landler Janning, Lasse Olsen

He has examined 100,000 Danes’ use of ChatGPT and discovered a marked inequality: “That was something we had not quite seen coming”

Interview: A major new Danish study shows that women are hesitant to adopt ChatGPT in working life. That means it is especially men who reap productivity gains, and this can have major consequences, warns one of the researchers behind the study.

ComputerworldJacob Ø. Wittorff Editor for digital business development and innovation

Chatbots are still straining the climate

Every time an AI model answers a query, it requires large amounts of energy because the task is sent to massive data centres. These centres, powered by thousands of powerful computers, use enormous amounts of electricity and water for cooling. Calculations show that a single query to ChatGPT uses as much energy as 40 mobile phone charges, or the equivalent of an hour of streaming on Netflix.

It is estimated that energy consumption in data centres could double by 2026 and approach the current consumption of all of Japan. At the same time, major technology companies such as Microsoft and Google are postponing their climate targets, as the intensive development of AI demands significant resources. The training of AI models such as Meta’s Llama3 has already resulted in substantial CO2 emissions, and the tech giants generally keep their cards close to their chest when it comes to their climate impact.

This development means that attention is now focused on how to make both AI technology and data centres more sustainable. One solution could be to create less energy-intensive AI models, but this requires a balance between efficiency and accuracy.

Chatbots are bad for the climate: Huge data centres could, within a few years, swallow Japan’s entire electricity consumption

Companies such as Microsoft and Google have postponed their climate targets because of AI. Danish experts want to make consumers aware of the footprint.

DR (Danish Broadcasting Corporation)Anna Danielsen Gille

Five years ago I wrote about the internet’s enormous CO2 emissions. Then AI came along and changed everything

Artificial intelligence will cause the world’s CO2 emissions to rise. But a young Danish student may have found (part of) the solution.

ZetlandTheis Ehler Molin

AI will consume as much electricity as all of Sweden. Now Copenhagen-based researchers have a solution

The artificial intelligence (AI) of the future will require large amounts of energy. That calls for a more climate-friendly alternative, researchers point out in a new study.

Videnskab.dkSee Tanya's profile

Data centres’ water consumption is exploding - now environmental groups are sounding the alarm

The spread of artificial intelligence is resulting in increased water consumption in many of the data centres that help run the large language models, several of which are located in drought-stricken areas.

ComputerworldJacob Ø. Wittorff Editor for digital business development and innovation

Debate: Is artificial intelligence an environmental culprit or a sustainable solution?

Explore AI’s environmental impact and potential for CO2 reduction. Viden.AI delves into AI’s energy use and sustainability aspects.

Viden.AIClaus Scheuer-Larsen

This week's other news

Company with a wild product: The teacher will take on a new role when artificial intelligence marks pupils’ essays

For a Danish teacher heading home to mark 25 essays, the latest product from Ordbogen.com might well sound attractive.

Fyens.dkLouise Holmgaard Nielsen loniefyens.dk

– Universities must embrace the technology

Several students at UiO are worried about AI development, but the vice-rector rejects the concerns: – I think one should be careful about crying wolf.

universitasPeder Reitbakken

Artificial intelligence as a tool in English teaching

Anton Tinghøj Honoré and Jacob Stochholm Wæhrens have been nominated by Lærerprofession.dk for their bachelor project on how, with the help of artificial intelligence, one can develop pupils’ writing skills and also the teacher’s English teaching.

folkeskolenNiels Chr. Sauer

5 good tips: How to use AI as a student without cheating

Be open, honest and experimental if you want to get the most out of chatbots such as ChatGPT and Claude at university.

Akademikerbladet

Researchers want clear guidelines for the use of ChatGPT in higher education

Universities and university colleges must guide students to use this new tool in a verifiable and ethical way, say researchers who have reviewed the topic.

Forskning.noLeif Tore Sædberg

AI pioneers want bots to replace human teachers – here’s why that’s unlikely

Artificial intelligence ‘bots’ may be widely available to teach in the coming years. But will they be effective? An expert on technology in education weighs in.

The ConversationAnnette Vee

The Future is Analog (If You Can Afford It) | TechPolicy.Press

Maroussia Lévesque considers “analog privilege” and what understanding it can tell us about the detrimental effects of AI systems.

Tech Policy PressMaroussia Lévesque

Professor: "I have long been waiting for Trump to tell a lie so big that it has consequences"

New AI technology has made it easier to spread credible lies. Everyone knows that. Precisely for that reason, it has also become easier to deny facts.

PolitikenJens Bostrup Journalist

Dansk Universitetspædagogisk Netværk (Danish Network for University Pedagogy) - SIG Webinar: Subjectification, pedagogy and GAI: Or what do humanities scholars make of ChatGPT?

SIG Webinar: Subjectification, pedagogy and GAI: Or what do humanities scholars make of ChatGPT?

Is it the end of Google soon? A new AI search engine could be a serious rival

The hosts of DR’s (Danish Broadcasting Corporation) AI podcast Prompt have spoken to the Chief Business Officer at Perplexity, which aims to address chatbots’ biggest problem - that they sometimes lie.

DRFrederikke Nyborg