(versione italiana qua)
In this article, I develop some general reflections on the use of generative AI tools in learning processes, first recalling two important observations that are too often forgotten in discussions on this topic. The first is that the levels of education are many and varied: primary, lower secondary, upper secondary, tertiary, and vocational. The "students" at these different levels have different capacities, which necessitate different approaches. The second is that – in the specific world of schooling – the three most relevant roles, students, teachers, and technical-administrative staff have different needs to address. Certainly, for teachers and technical-administrative staff, who are adults who have received specific professional training for the functions they perform in school, the use of tools that enhance their cognitive abilities can be useful, provided there is no excess, whereas the case is somewhat different for students
In this regard, I recall that a recent study of 319 knowledge workers, conducted in 5 different countries and 5 different work sectors, concerning 936 real examples of generative AI use, revealed a decrease in perceived effort, coupled with the development of excessive confidence in AI. Cognitive work shifted from execution to verification and integration of responses. The risk that the study highlighted is that of an atrophy of critical capacities if this use is not balanced by appropriate awareness. This is the well-known risk called "the irony of automation," highlighted in the early 1980s by Lisanne Bainbridge, according to which by mechanizing routine tasks and leaving the management of exceptions to humans, one deprives them of the daily opportunities to exercise their own judgment and strengthen their cognitive abilities. Consequently, such abilities weaken and people find themselves unprepared precisely when exceptions arise.
Another recent study conducted on 666 participants aged 17 and older in the United Kingdom showed results consistent with the previous study. The use of generative AI reduces cognitive effort but can weaken critical thinking if it occurs without autonomous reflection. These risks are greater especially for the less educated generations – which should spark reflection particularly among those attentive to social inequalities – and younger ones – a topic we will revisit later when discussing students.
These two experimental findings confirm what had previously been theorized about the effect of using generative AI tools on higher-order cognitive functions, such as reasoning, problem-solving, planning, and metacognitive monitoring. In that article, it was indeed hypothesized that constant and pervasive use of these tools could alter the efficiency of such functions and that interventions were therefore necessary to counteract these potential negative effects
And now let us turn to students, regarding whom, as early as March 2023, in my first article on this topic, I wrote: «Allowing our children the use of these tools before their complete cognitive development means impairing their possibilities for growth on the cognitive level».
Whereas, in fact, an adult can examine what a generative AI system proposes to them on the basis of their knowledge and experience, a school student is still developing those knowledge and skills that they would need to perform this verification. Therefore, an adult who chooses to use generative AI is deciding to decrease their cognitive effort and not exercise their abilities, but since they have acquired the necessary competence, if the abandonment of practice is not excessively prolonged there will be no negative consequences. Conversely, a student who uses these tools risks depriving themselves forever of the possibility of developing fundamental competencies for their cognitive growth, as we shall see further on.
However, an important distinction must be made among the types of use that can be made of generative AI tools. As is often the case with many tools invented by humankind, indeed, how one uses the tool makes the difference. An example that everyone knows is that of the knife, which can be used for offense or defense or to make certain activities more effective.
Some researchers at Anthropic (the company that developed Claude, one of the most widely used generative AI tools) recently made a preprint available (that is, a preliminary technical report not subject to the peer review process that characterizes scientific publications) that highlights – in the specific field of software programming – some interesting data. The study involved 52 junior-level software engineers who had to learn to use, in carrying out their tasks, a software code library previously unknown to them. Half had a generative AI tool at their disposal to help them, which could even provide them with the correct code, while the others did not have this option. The first group was faster (though not by much) in completing their tasks, but when subsequently asked about what they had learned, they obtained an average score of 50%, lower than the score of 67% obtained by those who had done everything "by hand".
This last result, though based on a limited sample, agrees with previously published scientific research that has reported the risk of a decrease in the ability to understand new topics related to the use of generative AI tools.
In one study, 91 university students were asked to examine a socio-scientific topic unfamiliar to them. In this task, half of them had a generative AI tool at their disposal to study the subject and produce arguments in favor or against it, while students in the other half could only use a search engine and therefore had to reach conclusions by working autonomously. Those in the first group reported lower cognitive effort but at the price of lower quality arguments produced.
In addition, apart from the study already discussed concerning the 666 participants in the United Kingdom, which had highlighted in younger people a greater risk of not developing critical thinking, it is highly relevant to recall a study on learning mathematics involving about a thousand upper secondary school students. Although the use of generative AI improves the results obtained in performing exercises, students who used it subsequently showed lower performance when it is no longer available, thus showing that they have acquired such skills to a lesser extent. This reduction in performance, however, is smaller if the generative AI tool is "constrained" to operate in a mode that prevents it from providing complete answers, but allows it only to provide suggestions and prompts prepared by the teacher.
In this direction, a further element of interest highlighted by the Anthropic study derives from the analysis that researchers conducted on the interaction patterns with the generative AI assistant for the engineers in the first group with lower performance. Among these, those who reported a score below 40% substantially delegated most or all tasks to the AI. Those in this group who instead reported a score above 67% used the AI to have their own independently produced work evaluated and to understand how certain mechanisms worked. The researchers clarified that there is no causal connection between the different approach used and the different score obtained, but the association between the two is certainly interesting to consider.
It is clear that when time is short, we all tend to use available shortcuts. However, if such a situation becomes the norm, there is a risk of obtaining a permanent decrease in cognitive abilities, even in adults who have developed them. Every faculty, mental or physical, once acquired is maintained active only through constant and specific exercise.
This leads to two reflections
The first: in work contexts, it is important that not only employees are put in a position to use generative AI tools with sufficient time available to verify their results and avoid producing "garbage" that someone else will have to fix, but also that a clear "corporate generative AI practice" be disseminated that can mitigate its negative consequences.
The second reflection concerns specifically school students, who generally tend to experience school study hours as an obstacle standing between them and more fun and interesting activities (being with friends, playing, doing sports, …). It takes a good dose of self-discipline to remain glued to the desk "racking one's brains" over topics that one cannot understand, when on one's smartphone is available an AI assistant that can provide complete answers.
Since the student will inevitably tend to take shortcuts, it is important that they have access only to constrained generative AI tools. While this may be feasible in the classroom, it seems impossible to achieve when the student is at home. I add that this in any case implies more work for the teacher, who will need to work much harder to prepare appropriate supplementary teaching material, if they want their students to derive real benefit from it.
I conclude by observing that this entire sector is still in tumultuous development and, while it is certainly necessary that researchers study methods and approaches that may be useful in improving teaching and learning, it is good to proceed in the daily practice of educational systems with great, very great caution to avoid causing harm, especially to younger students. Also because the consolidation of experimental evidence, in the practice of scientific research, is slow and sometimes non-linear. For example, a study that was discussed extensively, since it had highlighted a major positive impact of the use of generative AI tools on improving learning, through a meta-analysis of 51 research articles, was subsequently retracted by the journal that published it due to «discrepancies in the meta-analysis … that undermine confidence in the validity of the analysis and the conclusions derived from it».
The adage primum non nocere holds true in medicine as it does in education.
--The original version (in italian) has been published by "StartMAG" on 15 june 2026.
Nessun commento:
Posta un commento
Sono pubblicati solo i commenti che rispettano le norme di legge, le regole della buona educazione e sono attinenti agli argomenti trattati: siamo aperti alla discussione, non alla polemica.