Intelligence Augmentation: a technological revolution and paradoxical blessing

The Emperor's new AI clothes

De Volkskrant reported in March 2025 that Estonia is making AI integral to its education curricula, preparing schoolchildren for the age of artificial intelligence. What artificial intelligence actually is remains undiscussed, but fortunately, the Estonian Ministry of Education is going to collaborate with large tech companies, so that should be fine, or maybe not. This report is exemplary of how we think and write about AI. When it comes to artificial intelligence, I am increasingly reminded of Andersen's fairy tale about the emperor's new clothes, which are so wonderful that, according to the tailors, only smart people can see them. Many people talk about it, but few bother to explain what it is and why it is so important.

It is undeniable that technological developments, perhaps even a revolution, are underway. They are having a major impact on our society and, consequently, on education and learning. In order to assess these developments and their impact, it is necessary to examine what new clothes the emperor is actually wearing. Just like the emperor's wardrobe in the fairy tale, AI has acquired a status of inviolability. This is partly due to the name we have given it. AI stands for ‘Artificial Intelligence’, which is essentially a characteristic, but we talk about it as if it were an entity, and one that is intelligent and has more intelligence than the species that calls itself homo sapiens. 

In our language, AI is becoming increasingly anthropomorphic, perhaps even theomorphic. GenAI, often referred to in debates about AI, is not intelligent in a human way, but rather a statistical concept that is very good at estimating what a suitable answer to a question is. Critics also refer to GenAI applications as stochastic parrots. Very useful, but GenAI is still not an entity that knows what the answer is. Large Language Models do not generate their answers based on knowledge, but on probability distributions. They do not know that ‘two’ means twice as much as ‘one’, but they are very good at estimating that after ‘two’, the finite verb form of a sentence must be written in the plural. This makes it a very useful tool for writing and improving texts. 

As a tool, a term such as Intelligence Augmentation (IA) would be more appropriate, a term preferred by some information scientists. IA takes AI off its divine pedestal and helps us to think about how we can use it in education.

The inward-looking university

In so-called higher education, we approach intelligence augmentation in a quite myopic way. We seem to be primarily concerned with whether students can complete their assignments and tests too easily. To put it bluntly, we seem to be primarily concerned with whether our testing still discriminates in the right way and how much time it takes to devise a new, supposedly fraud-proof method of testing. We sometimes seem to forget to ask whether what we used to test is still relevant in this IA era. 

This is typical of a university that places itself outside society; we seem to forget to ask how universities can contribute to society; in this case, how do we ensure that the students who leave our institutions with a degree are actually prepared to contribute to that society, even if that society changes and, due to the ongoing development of information technology, requires different, new skills? 

Preparing students for a new reality

As educators and teachers, we must ask ourselves what is needed to prepare students as well as possible to contribute to and participate in this changing society. It is essential to realise that IA is more than merely large language models; it not only influences the writing of essays and master's theses, but also has and will continue to have a profound impact on all kinds of processes in society and, consequently, on the competencies expected of our graduates. This concerns matters that we can address with GenAI developments (i.e. reporting conversations, chatbots in healthcare), but also all kinds of subject-specific processes and insights. 

With the help of machine learning models, the dynamics of knowledge and prediction will change. Unlike traditional analytical models, which are limited to equations with a small number of dimensions, machine learning models can take countless dimensions into account. They ‘learn’ and predict based on diverse aspects, including those that transcend disciplinary boundaries, which remain hidden from the analytical eye or are too complex for conventional statistical methods.

Another consequence of the development of IA is that we no longer need to memorise facts; after all, we can have this information whispered to us in the blink of an eye by stochastic parrots, which makes it all the more important that we, as users of this readily accessible knowledge, are well equipped to assess the value of the information provided and apply it in a relevant context.

The ongoing development of IA systems will therefore require different skills from our students, who will be tomorrow’s professionals. This should be a reason for educators to consider the necessary adjustments to learning outcomes, rather than clinging frantically to the old ones and worrying about testing those old learning outcomes. Of course, new learning outcomes will be accompanied by adjustments to learning and teaching methods. They will also require new ways of evaluating and testing acquired skills, knowledge and competencies. So, there is work to be done, but perhaps IA can help us to put together this new education system.

Building on what we already want

Does this mean that everything needs to be redesigned and reorganised? No, I don't think so! Generally speaking, as educators, we already wanted to train students to become professionals who are able to deal with information critically. I think that if we were to draw up a general final attainment level for all our bachelor's programmes, this goal would come close to what we would want to achieve. 

But in the constructive alignment process that should follow, we have been a bit lazy. In developing learning methods, we were often not so eager to develop a critical style and we quickly fell back on knowledge transfer, on explaining how the world and our various disciplines work (which is not unimportant, by the way). We found testing critical thinking particularly difficult and time-consuming; a multiple-choice approach was much more efficient. Now, IA technology is forcing us to think seriously about what is needed to train critical professionals, simply because IA is increasingly taking over the primacy of factually reproducible knowledge. What remains for us, homo sapiens, and other biological intelligence is the domain of critically approaching information, playing with it and creating with it. 

Returning to the question of whether everything needs to be redone: critically handling information requires knowledge, an overview of facts, and understanding of principles and mechanisms. These remain necessary in the IA era. However, the facts will no longer be the (learning) goal in themselves, but will serve transformative learning. 

In educational practice, this will mean that we will have to focus on creative, transformative problems that require critical reflection. We can think of knowledge applications that require a higher level of creativity that goes beyond the fairly straightforward application of knowledge as we often incorporate into our PBL cases today (e.g., develop a strength training programme for a cyclist and a runner). For these types of questions, one correct answer can be devised based on physiological insights. For transformative learning, we need to come up with issues that force students to weigh up different areas of knowledge and interests. 

This requires reasoning and a real understanding of the underlying mechanisms and theories. This means that these are cases that cannot be solved based on knowledge from a single academic discipline, but rather cases that take on the character of a wicked problem, in which the interests and insights of various stakeholders and disciplines must be weighed. These should be authentic problems, not contrived ones.

The human edge: creativity and empathy

GenAI cannot yet reason or weigh up knowledge and interests. In these domains, where creativity and empathy are of great importance, homo sapiens still has the upper hand, for the time being. 

The paradox of this story is that IA does not make us dumber or threaten our creativity. Instead, it invites us to learn more transformatively and allow our innovative power to flourish.

By Hans Savelberg, Professor of Evolving Academic Education at the Faculty of Health, Medicine and Life Sciences of Maastricht University.