Some thoughts from Angelo Hankins, Bassetti Foundation collaborator and first year student in Global Sustainability Sciences at Utrecht University

While the emergence of artificial intelligence (AI) is transforming many facets of modern society, I find its impact on education—particularly on students— uniquely profound. As a high school student during the early stages of AI development, I witnessed firsthand how these tools began to reshape our academic practices. High school students, still in the process of developing their academic research and writing skills, are on the whole very receptive to developments in AI technology. Can studying this demographic provide an early insight into how AI will influence education in the future? By examining the adoption of AI tools in schools, can we forecast how AI will reshape educational practices, research, and the dissemination of knowledge, while also addressing challenges it poses?

My first encounter with the technological phenomenon came in around 2017 in the form of the grammar checker Grammarly and later the much more powerful Quillbot, which rewrites text to be more concise or more formal, depending on the user’s preferences. In the years following I became increasingly acquainted with the Large Language Model (LLM) ChatGPT, during a period that marked a sharp increase in the development and adoption of AI.

Talk about the development of AI and the effects it will have often takes the form of a metaphorical ‘blanket’ that will cover all aspects of society (compromising jobs) and drastically change the way society functions. I find the blanket analogy to be incorrect however, a nailbed being a more suitable metaphor, with society unevenly impacted by the effects of AI. The same is to be said with education, some parts of education will change drastically while others will largely remain the same.

There have been attempts to let AI completely take over schoolwork, using Quilbot to rewrite existing articles and submitting them as one’s own, and using ChatGPT to write assignments. Yes, there are tasks in which AI can write an assignment which would score highly, but this requires an assignment that is easily repeatable, and graded not on one’s ability to reason and create rational arguments but to repeat and outline established facts and truths, simply rearranging them in the form of an essay or a paper.

Following this line, I would argue that AI is less of a threat to education than that of the inability of the educational system to adapt to this rapidly developing societal change. The nature of education needs to change to adapt to AI developments. Assignments and curriculum can no longer prioritise approaches that grade a student’s ability to recount facts and events but must grade based on conclusions which lie outside the domain of AI. At least currently, while LLM’s are capable of writing impressive texts, their reasoning skills and related ability to derive new conclusions from pre-existing information are below the standard that would be expected from a high school student.

This is by no means a new argument. In fact, it shares many facets with the notion of Bildung brought forward by von Humbold in the early 1800’s. He presented a theory for an educational system which broadens one’s insights through a diverse curriculum, combining “traditional education” with for example research. Undoubtedly, given that these findings were written centuries before the emergence of AI, we may want to focus on some of his musings rather than others in ways that reflect our modern educational system,  for example the aim to stimulate critical thinking.

Ultimately, I consider this change in focus for education to be positive for the development of human society. A curriculum which leads its students to both reason in ways AI is unable to and to make considerations outside of the reign of AI will lead to better skills in making and substantiating informed opinions.

This does not mean that education will be untouched by AI. AI has drastically changed my life as a student, but it has been the AI tools that are specialized and trained to execute a very specific task that have had the largest effect, rather than the generic large language models such as ChatGPT and its Chinese counterpart DeepSeek.

To offer an example, Google’s NotebookLM has fundamentally changed my practices as a student, allowing me to train a model on the readings and presentations for specific tasks or exams. A question which previously required searching for an answer within various significant chapters can now be answered in a matter of seconds, including the relevant citations. AI has also improved its summarising capacities, with Recall a crucial tool for quickly condensing large works into concise but very detailed conclusions.

But ultimately, these are merely tools that increase my productivity while carrying out processes that I already learned to do manually in the past. Will these (old-school) skills be lost for the future generation though?

In the near future, I believe that education will maintain a structure which is similar to its current form, with exams remaining largely unchanged. Essays and writing assignments will need to adapt to push students to reason outside of the reigns of AI, but will still be present within the educational curriculum.

(Image produced with Dall-e and Photoshop AI tools)