Dear Giorgia, I am writing to you ... about software

by Enrico Nardelli

(versione italiana qua)

I was prompted to write by your recent address to the General Assembly of Assolombarda, the association representing businesses in Lombardy, in which you announced that you and Minister Giorgetti are working to introduce a tax incentive for software investment in the forthcoming budget law.

I very much welcome the intention to introduce such a measure. For the past twenty years, first as president of the Italian Association of University Professors of Informatics and later as president of the European association bringing together university informatics departments, I have worked to raise awareness across society of the importance of software for growth and development.

This is all the more true of an economy such as Italy’s, which has built its strength on excellence in specialised manufacturing niches. The availability of sophisticated software is now an essential component for competing successfully in an increasingly digital and dynamic global market.

There is, however, one point I believe deserves attention. According to UNCTAD-WTO data, Italy is expected to record a negative trade balance in computer services of approximately $1.1 billion in 2025 (category: computer services, which includes bespoke software development, consulting, maintenance and cloud< services). The deficit is limited relative to trade, and the balance alone does not say how much of the incentivized spending will go to foreign suppliers. The value of imports, about 6.6 billion dollars in 2024, shows, however, that a non-negligible share of our demand is already met by foreign firms.

The risk, therefore, is not that the incentive is wrong in itself, because the first to benefit are the Italian firms that invest, and such investment can lead to an increase in their productivity. The risk is that, without adequate complementary measures aimed at strengthening Italian supply, part of the resulting economic benefit will flow abroad.

To achieve this national objective in a world where digital technology is becoming ever more sophisticated and competition increasingly intense, we must invest in education, both at school and university level, so that Italy can significantly strengthen its long-term capacity to develop, manage and adapt software. Building these skills is not something that can be achieved durably in a few years.

Earlier this year, the Minister of Education and Merit in your government introduced an important school reform under which informatics is being introduced into the curriculum from primary school, starting in the current academic year.

This has been done not only to introduce informatics as basic cultural education, because all citizens must now know what this science that underlies the digital world we live in is, but also to ensure that more young people can go on to choose it as a subject of study at university. If we increase the number of people in Italy capable of writing software, a larger share of investments can be directed toward solutions developed or integrated in Italy.

The reform introduced by Minister Valditara will not be enough unless it is accompanied by a serious investment in teacher education, so that teachers are properly equipped to teach informatics in schools. The United Kingdom, for example, established a national centre in 2019 with £84 million in funding to train teachers. We also need a strong development of degree programs in informatics, which must produce the technicians businesses need, currently in insufficient numbers in our country.

Italy does not need to build another Microsoft or Google. That is not what our country needs. What we need is for every business, even one with just ten employees, to have an informatics specialist capable of identifying and developing software solutions suited to its specific needs, and adapting them as the business grows and its operating environment changes. Every company is different, and each operates in circumstances that evolve day by day. Every business needs a solution tailored to its own needs. However, unless its software is continually evolved by someone who understands the company's operational context, it will not be able to fully meet those needs, because, by its very nature, software is rigid and mechanical and lacks the necessary flexibility.

This remains true even when software is built on artificial intelligence technologies. However sophisticated these technologies may be in certain fields, they do not possess the deep understanding of the real-world dynamics of a specific organisation that human beings have. That is why investing in people's informatics education remains the true strategic priority, even in the age of artificial intelligence.

It is therefore necessary to include two measures in the budget law focused on these two aspects: informatics education for school teachers and strengthening degree programs in informatics.

I am well aware that these are long-term measures, but I believe that you, the first woman to serve as Prime Minister of Italy, have a strong strategic sense and fully grasp the importance of initiatives in this direction for our country's long-term growth.

--
The original version (in italian) has been published by "StartMAG" on 15 October 2026.

For a Realistic Vision of Artificial Intelligence

by Enrico Nardelli

(versione italiana qua)

On Tuesday, June 16th, in the press room of the Chamber of Deputies, a press conference was held to present an open letter to society titled "For a Realistic Vision of Artificial Intelligence", signed so far by more than 350 researchers working in public and private research institutions in Italy and around the world.

The letter, promoted by myself and Walter Quattrociocchi of Sapienza University of Rome, aims to bring the discussion on artificial intelligence (AI) back to more realistic grounds, stripping it of anthropomorphized visions that oscillate between the salvific and the apocalyptic.

We were fortunate to have the kind hospitality of the Vice President of the Chamber, Hon. Anna Ascani, who opened the event by thanking us for the initiative and recognizing the importance of bringing it to a venue where there is a need to understand what must then be regulated. On the subject of AI, the Chamber has been working at various levels for some time, in collaboration with the academic world. There is awareness of the risks, but also of the opportunities that Italian citizens could benefit from, with full recognition that this is an extremely sophisticated technology that cannot, however, replace human beings. It is therefore important, she concluded, the message that this letter is conveying.

Subsequently, Quattrociocchi briefly presented the motivations behind the letter, which is to fill a narrative void created by this frenetic and, in some ways, unexpected advancement of generative artificial intelligence (GAI), which simulates knowledge at the level of the best experts without truly possessing it. A narrative has been created that, in some cases, is divorced from the reality of what is objectively and scientifically verifiable. The intention is not to exclude debate on opinions, but to distinguish between opinions, visions, and experimental evidence—an approach that is fundamental especially when dealing with abstract themes such as intelligence and consciousness. This scientific approach has, paradoxically, become more difficult in recent years, where the dominance of social media in the realm of communication rewards suggestiveness over precision. The letter's objective is to foster the development of a solid understanding of this technological innovation, in collaboration with academia, politics, and the media.

Next, Andrea Cerroni, sociologist of science and technology at Roma Tre University, recalled that technology itself carries within it a vision of the world and a societal model, defined by those who built that technology. Moreover, the data on which these technologies are based are selected by someone who chose them based on their needs and objectives. This is a legitimate process, but one that often lacks transparency. In the specific case of GAI, this can lead to cognitive oligopolies and opaque logic, inducing users to mistake what these systems produce for absolute truth. He also noted that even those who practice science do not produce "The Truth" but rather what seems to be a reasonable explanation of the phenomena we observe.

Then, Andrea Orlandini, from the Institute of Cognitive Sciences and Technologies at the CNR, spoke about the apparent simplification resulting from GAI, which makes even complex fields of knowledge easily accessible to everyone, yet for which people often lack the knowledge to evaluate the reliability of responses that, being generated on a statistical basis, are inevitably prone to contain what are called "hallucinations"—that is, statistically plausible but incorrect sequences. The letter, he added, aims to separate the scientific context, in which different viewpoints—perhaps in disagreement with each other—are presented with an approach based on logical rigor and experimental evidence, from the informational-educational realm which, in a rapidly evolving sector, sometimes leads to erroneous simplifications and misleading exaggerations, often driven by strong economic interests.

Concluding the first round of scientific reflections, Salvatore Orlando, from the Department of Law and Business Economics at Sapienza, who began by reinforcing Cerroni's previous observation, noting that data is not objectively found in nature but is "collected" by a subject, who therefore performs not only a selection on "what" to collect but also decides how and when to gather these elements on which interpretations and theories are then built. He then emphasized, as a legal scholar, that one can speak of true "decisions" only in the presence of a human being or a legal entity. Finally, he recalled the danger of the authorization granted by the Artificial Intelligence Act to platforms to acquire and process sensitive data of everyone in order to make decisions regarding the possible removal of content.

Hon. Mascaretti, while appreciating the initiative and the common-sense reflection on society's perception of GAI, criticized some recent cultural and political initiatives involving the anthropomorphization of this technology, which have received extensive media coverage that does not contribute to a real understanding of its mechanisms. He then presented an interesting and apt comparison about how the spread of the automobile in the postwar period brought a large mass of Italians to discover corners of this beautiful country (and beyond) that they previously could not reach and know, just as the technology of generative artificial intelligence (GAI) enables many people to access domains of knowledge that were previously difficult and thus beyond the reach of mass understanding. He concluded by recalling the commercial interests behind GAI tools, appreciating our initiative to push the media world to address this issue on more realistic grounds, an aspect of extreme importance especially for younger generations.

In my remarks, I first took up Hon. Mascaretti's comparison, noting the benefits that GAI can bring—for example, helping people understand a medical diagnosis or a court ruling—but also the risks connected to errors that could be present in these explanations, which are unavoidable because they are mathematically inherent to how these systems construct their output. Additionally, I recalled how younger generations are the most defenseless, by virtue of their lack of experience and incomplete knowledge. As I discussed in my recent article, there is experimental evidence showing that the availability of GAI tools with no restrictions on their use leads students in training to bypass that phase of cognitive effort that is inseparable from the process of acquiring knowledge and skills. Therefore, if we do not want to lose the opportunity for future generations to acquire those high-level cognitive competencies that are increasingly necessary in an increasingly complex digital society, we will need to discuss and reflect on what can be done.

In the second round, Hon. Iaria spoke, appreciating the manifesto's call to explain what artificial intelligence really is, which is not human—a myth that must absolutely be dispelled—adding that it is necessary to learn to use these tools that will be increasingly present in the world of work. Next, Hon. Caso, who reiterated that the crux is understanding this technology and that it is necessary to use these tools to innovate teaching methods and to facilitate learning. He finally discussed the importance of maintaining control of these tools, currently held by private entities and outside the European context. Otherwise, the risk is that they create a new digital divide. This is a fundamental point on which states, even in a dimension of European collaboration, must be able to intervene, considering GAI as a "common good".

Quattrociocchi took up this point about this technology as a common good, while reiterating the importance of ensuring and guaranteeing its reliability. These tools produce output similar to that of human beings, but humans have the capacity to verify their statements, whereas GAI is designed to optimize linguistic plausibility. The problem is that then, even driven by the commercial push of platforms, one jumps from this to the cognitive plane. This is already causing damage in universities, Quattrociocchi stated based on his personal experience, with students producing theses and projects that are beautifully written but which they do not truly understand. He concluded by noting that if GAI does not acquire solid reliability, it will not be able to ensure an economic return.

The conference ended with brief reflections from the other colleagues. Cerroni recalled that GAI should not be seen as a tool that allows one to shorten overall time, but rather to offload repetitive and mechanical activities, focusing on cognitively higher-level work. Orlandini emphasized the need to inform in a positive way, considering that these are very sophisticated tools that acquire new elements revealing new emergent behaviors that must be studied, but contextualized properly. Orlando observed that in the legal field, the interpretation of norms and their hermeneutics are subject to evolution that, however, GAI tools do not see because they always refer back to quantitatively prevailing interpretations which are, however, the oldest ones. From the perspective of legal culture, this is therefore a reactionary stance that severs the connection created between society and law, based precisely on the most modern interpretations.

I concluded by agreeing that these are fundamental instruments of consensus control, which requires that all researchers who have signed the open letter continue to work for the spread of better understanding of them in everyone's interest.

--
The original version (in italian) has been published by "StartMAG" on 18 June 2026.

Generative AI and learning: primum non nocere

by Enrico Nardelli

(versione italiana qua)

In this article, I offer some general reflections on the use of generative AI tools in learning processes, first recalling two important observations that are far too often forgotten in discussions on this topic. The first is that there are many and varied levels of education: primary, lower secondary, upper secondary, tertiary, and vocational. "Students" at these different levels have different capacities, which require tailored approaches. The second is that, in the specific world of schooling, the three most relevant roles – students, teachers, and technical-administrative personnel – face different needs. Certainly, for teachers and technical-administrative staff – adults who have received specific professional training for their roles in school – the use of tools that enhance cognitive abilities can be helpful, as long as it is not excessive, whereas the case is somewhat different for students.

In this regard, it is worth noting that a recent study of 319 knowledge workers, conducted across 5 different countries and 5 different work sectors, based on 936 real-world examples of generative AI use, highlighted a decrease in perceived effort, coupled with a rise in overconfidence in AI. Cognitive work shifted from execution to verification and integration of responses. The study highlighted the risk of an atrophy of critical thinking skills 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: by mechanizing routine tasks and leaving the management of exceptions to human beings, one deprives them of daily opportunities to exercise their judgment and strengthen their cognitive abilities. Consequently, these 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, yielded results consistent with the previous one. The use of generative AI reduces cognitive effort but can undermine critical thinking if it occurs without autonomous reflection. These risks are particularly pronounced among less educated generations – which should prompt reflection especially among those concerned with social inequalities – and younger ones – a topic we will return to when discussing students.

These two pieces of experimental evidence 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 paper, it had indeed been hypothesized that constant and pervasive use of these tools could alter the efficiency of such functions and that therefore interventions would be necessary to counteract these potential negative effects.

And now let us turn to students. As far back as March 2023, in my first article on this topic, I wrote: «Allowing our children to use  [these tools] before their full development means undermining their potential for cognitive growth».

While an adult can evaluate what a generative AI system proposes based on their knowledge and experience, a school student is still developing the very knowledge and skills that would be necessary for them to carry out this verification. Therefore, an adult who chooses to use generative AI is deciding to reduce their cognitive effort and not exercise their abilities, but since they have already acquired the necessary competence, if the abandonment of practice is not excessively prolonged there will be no negative consequences. On the other hand, a student who uses these tools risks permanently depriving themselves of the opportunity to develop fundamental competencies for their cognitive growth, as we shall see below.

In this regard, however, an important distinction must be made among the different types of uses that can be made of generative AI tools. As is often the case with many tools invented by humans, indeed, how one uses the tool makes the difference. A familiar example is 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 available a preprint (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 engineering – some interesting findings. The study involved 52 junior-level software engineers who had to learn to use a code library previously unknown to them in order to complete their tasks. Half had access to a generative AI tool that could also provide them with the correct code, while the others did not have this capability. The first group was slightly faster in completing their tasks, but when subsequently asked about what they had learned, they achieved an average score of 50%, lower than the 67% score achieved by those who had done everything "by hand".

This last result, although based on a limited sample, aligns 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 access to a generative AI tool to study the topic and produce arguments for or against it, while students in the other half could only use a search engine and therefore had to reach conclusions working independently. Those in the first group reported lower cognitive effort but at the cost of lower quality arguments produced.

In addition, apart from the already discussed study involving 666 participants in the United Kingdom, which had highlighted a greater risk in younger people of not developing critical thinking, it is highly relevant to recall a study on mathematics learning involving approximately a thousand upper secondary school students. Although the use of generative AI improves the results obtained in solving exercises, students who used it subsequently performed worse when it was no longer available, thus showing that they 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, another element of interest highlighted by the Anthropic study derives from the analysis that researchers conducted on interaction patterns with the generative AI assistant for engineers in the first group, with inferior performance. Among these, those who reported a score below 40% essentially delegated all or nearly all tasks to AI. Those in this group who instead reported a score above 67% used 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 a situation of this kind becomes the norm, there is a risk of obtaining a permanent reduction in cognitive abilities, even in adults who have developed them. Every faculty, whether mental or physical, remains active only through constant, specific exercise.

Two reflections follow from this.

The first: in work contexts, it is important not only that employees be given the opportunity to use generative AI tools with sufficient time to verify their results and avoid producing "slop" that someone else will have to fix, but also that a clear "corporate practice of generative AI" be disseminated that can mitigate its negative consequences.

The second reflection concerns specifically school students, who generally tend to view school study hours as an obstacle standing between them and more fun and interesting activities (spending time with friends, playing, doing sports, …). It takes a good deal of self-discipline to remain glued to the desk "racking your brains" over topics you cannot understand, when a smartphone offers an AI assistant available 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 would add that in any case this implies greater work for the teacher, who will need to work much harder to prepare appropriate supplementary instructional material if they want their students to derive real benefit from it.

I conclude by noting that this entire sector is still in tumultuous development and, while it is certainly necessary for researchers to study methods and approaches that can be useful in improving teaching and learning, it is wise to proceed in the daily practice of educational systems with the utmost caution to avoid causing damage, especially to younger students. This is also because the consolidation of experimental evidence in the practice of scientific research is slow and sometimes non-linear. For example, a study that received a great deal of attention because it had highlighted a significant positive impact of the use of generative AI tools on improving learning, through a meta-analysis of 51 research articles, was later 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 aphorism primum non nocere is as true in medicine as it is in education.

--
The original version (in italian) has been published by "StartMAG" on 15 june 2026.

Digital Tools and Learning: Plato Wins Again

by Enrico Nardelli

(versione italiana qua)

I had recounted, in a previous article, how the 2023 UNESCO report had sounded a first, dramatic alarm bell by defining Ed-Tech (that is, the use of digital tools in the education system) a "tragedy" with immense costs and virtually no benefits.

On January 15, 2026, before the United States Senate Committee on Commerce, Science, and Transportation, Dr. Jared Cooney Horvath, a neuroscientist specializing in learning and technology, recounted another silent tragedy: that of an entire generation sacrificed on the altar of digital technology.

Horvath, director of LME Global and author of the volume The Digital Delusion, is neither a nor a nostalgic. He is a scholar who has spent years analyzing tens of thousands of studies on the use of digital technology in education.

For over a century we have taken for granted that each generation was more intelligent than the previous one. This is the so-called "Flynn Effect." But Horvath presented chilling data: starting from the mid-2000s, the cognitive development of young people in the developed world stopped growing. In many cases, it has gone backward.

Horvath didn't limit himself to words. He brought before the committee, which conducted a public hearing (here the full video) on the theme "Plugged Out: Examining the Impact of Technology on America's Youth," data from major international studies involving millions of students in dozens of countries. These include PISA (Programme for International Student Assessment), the triennial survey on the level of preparation of 15-year-old students, TIMSS (Trends in International Mathematics and Science Study), a study that analyzes international trends in mathematics and science learning, and PIRLS (Progress in International Reading Literacy Study), an international survey that evaluates the reading skills of primary school students.

The conclusion is uniform: greater daily exposure to digital tools consistently corresponds to lower scores in reading, mathematics, and science. The more time spent in front of a screen, the worse the performance. The data is available within his written testimony.

I know very well that correlation is not causation, but it seems objectively difficult to identify causes other than the massive spread of digital technology in schools. And, in any case, the precautionary principle suggests the advisability of carefully reconsidering the situation.

On the other hand, what Horvath reported is in agreement with many independent studies related to reading and writing with digital devices.

For reading: text comprehension and retention are more solid on paper than on screen, particularly for complex or long texts. Spatial stability, reduced scrolling, and physical interaction with the paper page support memory formation and comprehension. The scientific literature on this point is substantial and convergent. A meta-analysis of 49 studies published in 2024 reaffirmed that students who read on paper consistently achieve higher scores on comprehension tests compared to those who read the same material on screen. Here is an informal description of what is called the "screen inferiority effect."

For writing: taking notes by hand is superior, for long-term learning, to taking notes on a computer. From the pioneering study by Mueller & Oppenheimer of 2014 (with the significant title "The Pen Is Mightier Than the Keyboard") to the recent article by Italian neuroscientists from 2025, the evidence is consistent. While typing on a keyboard encourages literal transcription and superficial processing, writing by hand requires synthesis, comprehension, and reorganization of concepts, and is therefore superior in an educational context.

Horvath examined nearly 400 meta-analyses, covering over 21,000 research studies related to the use of digital technologies for teaching. Most of these achieve results below the effectiveness of normal classroom instruction. In other words: a good teacher who lectures in a traditional way produces better results than almost all digital tools analyzed in the scientific literature.

Only in very limited contexts — adaptive exercises for basic skills and targeted remedial interventions — do digital tools manage to approach significant results. In these cases they work because they support the repetition of exercises in well-defined areas, not because they improve deep learning.

The reasons for this lie in neuroscience. When attention is repeatedly interrupted, three costs emerge: loss of time due to task switching, increase in errors due to cognitive interference, and weakening of memory formation, since learning shifts from deep encoding toward habitual processing.

The problem is that digital platforms are optimized for rapid switching between content, novelty, and continuous attention capture. Even when used for academic purposes, they activate the same behavioral patterns that students practice in recreational screen use: frequent checking, rapid scrolling, multitasking. These digital devices train young people's brains to respond to brief, intermittent, gamified stimuli. Technology doesn't make them smarter. It makes them dependent on the immediate reward circuit.

Deep learning, the kind that forms lasting memory and critical thinking, is built through interaction between people and in the continuous and thorough exercise of one's mental capacities, not in the relationship between a person and a screen. The human brain learns through relationships, socially mediated error, the physical presence of others — that didàskalos that Plato knew to be inseparable from the mathetés. Technology, if well used — and this is a very critical "if" — can enhance this encounter between teacher and student. But it can never replace it.

The fatal error of the last twenty years has been believing that a tablet could be a bridge to knowledge, when in reality it has been a wall that has prevented access to deep cognitive functions. Technology is a tool for experts who want to lighten their work, Horvath reminded us, it is not the way novices learn to become experts. A calculator is an extraordinary tool for someone who already knows statistics; it allows them to externalize the effort to concentrate on strategy. But if you give the same calculator to a student who has not yet internalized the meaning of various methods, you are not helping them: you are preventing them from building the synapses necessary to understand statistics.

The central point of all of Horvath's work, with which I completely agree, is not the rejection of technology, but bringing it back to its true nature: a useful tool, not a substitute. The absolute priority must be learning, which must be carried out in person with human teachers, as we have always done, that is, in an analog way. In this regard he stated, during his testimony: "The secret to learning to use digital technology is not to use digital technology. It's learning analogically, the way we've always learned, and then using technology to facilitate the exercise of cognitive abilities we've already developed."

Yet, many opinion newspapers (especially those that have praised distance learning) ignore this wealth of evidence. They prefer to tell stories of "innovative schools" with interactive whiteboards and classrooms without books. Of course, it's easier to celebrate technological novelty than to face the effort of building true educational relationships and investing resources to have well-prepared and well-paid teachers, the only ones who can make a difference.

Who is thinking about the long-term well-being, cognitive and otherwise, of future generations?

--
The original version (in italian) has been published by "StartMAG" on 26 April 2026.

The Wonders of Generative Artificial Intelligence?

by Enrico Nardelli

(versione italiana qua)

The title is inspired by the novel "Le meraviglie del 2000" (= The Wonders of the Year 2000), not exactly among Emilio Salgari's best-known works, which describes the adventures of two friends who sleep for a hundred years and awaken in a future world (in 2003) where electricity is omnipresent, has made communications instantaneous and automated countless work activities. For them, coming from the early 1900s, it's a beautiful and pleasant experience. But then... if you keep reading, you'll understand for yourselves why I chose it as inspiration.

One of the promises tied to the advent of generative artificial intelligence (GenAI) is that by automating the most repetitive and low-value-added tasks, it would allow workers to have more time for more creative and meaningful activities. Research conducted for eight months, from April to December 2025, at an American tech company of about 200 employees, including through in-depth interviews, partially contradicts this expectation: GenAI tools don't reduce workload, but intensify it. Workers operate at more sustained paces, perform a broader range of tasks, and extend their activity to time slots that were previously reserved for rest, often without anyone explicitly asking them to.

Three main forms of intensification have been identified.

The first is task expansion: since with GenAI one can easily access professional skills that one doesn't possess, workers have begun to tackle tasks that were previously performed by others. Product managers and designers started writing code, researchers took on product development activities, and generally employees attempted work they would have previously delegated or avoided. GenAI tools made these tasks suddenly accessible, making possible what many experienced as cognitive enhancement: less dependence on others, immediate feedback, sense of autonomy. The concrete result was a progressive widening of each role's perimeter, with cascading effects: engineers, for example, found themselves having to review and correct GenAI-assisted work produced by colleagues, with a further increase in their own workload.

The second form of intensification described in the research was the progressive blurring of the boundary between work and non-work. Since starting a task with GenAI became almost frictionless — no more "writer's block" — workers began inserting small work activities into moments that were previously for waiting: during lunch, in meetings, while waiting for a file to load. The conversational style of prompts contributed to normalizing this behavior, making it difficult to perceive the activity as "real work." Over time, breaks lost their recovery function and work took on an increasingly widespread and continuous presence throughout the day.

The third form is intensified multi-tasking: GenAI introduced a new work mode where a person simultaneously performs multiple activities — writing code manually while GenAI generates an alternative version, running multiple agents in parallel, resuming long-postponed activities because GenAI can "handle them" in the background. While this creates a sensation of productive momentum, it also causes continuous attention shifting, frequent checking of what GenAI is doing, and a growing number of open activities to follow.

These three mechanisms feed each other in a self-reinforcing cycle: GenAI accelerates certain tasks, speed expectations increase, dependence on GenAI grows, the work perimeter expands further.

The authors warn that what appears to be a productivity gain risks masking a silent and unsustainable expansion of workload, with consequences for cognitive fatigue, decision quality, turnover and burnout. Since the additional effort is voluntary and often experienced as stimulating experimentation, managers struggle to notice it until the damage is already done.

At this point I pick up the threads of Salgari's novel: what happens to the two young men is that the frenetic pace and continuous stimuli caused by the acceleration of human activities induced by electricity lead to irritability, insomnia, mental fatigue, culminating in a real nervous breakdown. What seemed like an unstoppable possibility for progress reveals itself as a physiological and psychological threat. The moral, extraordinarily current, is that technological progress, if not accompanied by balancing actions that take our human nature into account, can become destructive.

We must be well aware of this aspect as generative artificial intelligence tools (useful, but to be used with constant attention) become increasingly widespread. Surely psychologists who deal with digital transformation will be able to contribute to deepening these aspects.

The solution proposed by the authors of the study I cited is not to ask individuals to self-regulate, but to develop a real "corporate GenAI practice": a set of intentional norms and routines that structure the use of these tools. The authors suggest three concrete levers: intentional pauses to slow the pace and prevent silent accumulation of overload; work sequencing, namely norms that regulate when and how activities advance, protecting concentration windows and reducing context switching that is costly from a cognitive standpoint; finally, human grounding, namely the protection of spaces and moments of social connection that interrupt solitary work mediated by GenAI, broaden perspective and feed creativity through comparison between different human viewpoints.

These are practices that become increasingly important as informatics technology becomes more widespread. Despite beginning to talk about techno-stress in the mid-1980s and some countermeasures being taken, e.g. at least in some organizations email is not accessible outside work hours, we are — thanks to smartphones and social networks — constantly connected.

Let's be careful that the further acceleration due to generative artificial intelligence doesn't derail our mental health. What do you think?

--
The original version (in italian) has been published by "StartMAG" on 11 April 2026.

The big illusion is over: Artificial Intelligence is not replacing software developers

by Enrico Nardelli

(versione italiana qua)

Let me say upfront — to pre-empt the usual hasty comments from those who won't read to the end — that I do believe tools based on generative Artificial Intelligence (GenAI, hereafter) are genuinely useful in software development, as they are in many other fields. But we need to be able to tell dreams from reality.

In 2023, Big Tech CEOs were more or less unanimously declaring that GenAI would replace most software developers by 2025. As recently as March 2025, Dario Amodei, CEO of Anthropic, was claiming that within 3 to 6 months GenAI would be writing 90% of all code.

The story went that the future would be "agentic" — that developers would have digital colleagues who never slept and never complained and would therefore take their jobs.

2024 closed with no fewer than 152,000 employees laid off across the tech sector worldwide, while by December 2025 the number of layoffs had reached 123,000 — all of it justified, according to the executives themselves, as "realigning towards an AI-centric future."

2026 has now begun, and there is no trace of the miraculous transformations that until recently were being touted as certainties — as I had anticipated several months ago. There I cited the Generative AI Gap report from MIT's NANDA project, published in July 2025, which had found that despite $40 billion in global investment, a staggering 95% of GenAI pilot projects had failed to produce any measurable economic return. Most organisations were seeing a net impact of zero on their bottom line.

Here are some updates that help explain why.

The GitClear report of February 2025 examined 211 million lines of code modified in the open source repositories used by Big Tech, finding that between 2020 and 2024 there was a near-doubling (from 3.05% to 5.67%) in the proportion of code that is added and then deleted within two weeks — so-called code churn. This is a clear indicator of a progressive decline in the quality of software being produced. The only technological or organisational change of any significance that occurred over the same period is that the proportion of developers using GenAI tools rose from zero to 63%. Correlation is not causation, of course, but this looks very much like a smoking gun. Furthermore, the report also recorded an increase in duplicated code blocks and a decrease in code reuse. These too are signs of declining quality. A December 2025 study by CodeRabbit, analysing 470 pull requests on the GitHub platform (used by more than 100 million developers worldwide to share and collaboratively develop code), found that GenAI-generated code produces on average 1.7 times more issues than code written by humans.

Stack Overflow, following a survey involving nearly 50,000 developers across 177 countries, reported that the proportion of developers favourably disposed towards GenAI tools fell from over 70% in 2024 to around 60% in 2025. The main source of frustration, cited by 66% of respondents, was that GenAI-produced solutions are "almost right, but not quite" — which leads to wasted time fixing errors in the code generated this way. That said, 69% of them acknowledged that GenAI does increase productivity in software writing. Several studies carried out in 2024 found that this increase ranges from 10–20% for senior developers to 35–40% for junior ones, who are better placed to benefit from a system that has knowledge of virtually all the software ever written in the world.

These findings are corroborated by studies conducted by the Stanford software engineering productivity research group involving more than 100 developers at major technology companies, which showed that the productivity gains in software development achieved through the use of GenAI tools are on average between 10% and 20%. Using them uncritically for software development, however, risks being a very costly mistake in the long run. While the volume of code produced increases by almost 40%, its quality decreases, requiring additional time for correction. And these are just the short-term consequences. Over the medium or long term — that is, in the context of the adaptive maintenance of software systems that have been in operation for some time — the situation risks becoming explosive.

There are no results yet pointing in this direction, given how little time has passed, but two significant findings emerge from the Stanford group's studies. The first is that without strong discipline in keeping a company's codebase structured and organised in a clear and clean manner, productivity gains evaporate or even reverse. The second is that productivity gains are greatest for new, low-complexity projects, while they diminish considerably for mature, high-complexity ones.

Throughout 2025, the wonders of vibe coding were extolled — the approach to software development in which a developer interacts with GenAI in natural language to "materialise" a complex software system by describing its overall vision. The problem is that this approach is fine for a demo but not for production-grade systems, because — like a sandcastle — it does not hold up solidly over time. This is the so-called GenAI technical debt: the future costs that accumulate when, in the rush of software development, one takes advantage of GenAI's code generation speed without bothering with thorough checks and verification.

During 2025, an approach emerged — conceptually known for some time, though now gaining traction — called specification-driven development, in which the developer uses GenAI tools starting from a specification — that is, a high-level description of the desired behaviour of the entire system — defined by the developer and progressively refined in ever greater detail, also with the support of GenAI. In this way, one partially overcomes the main limitation of these tools, which are based on statistics rather than symbolic representation: their inability to represent concepts that abstract away from the literal context under examination. With this approach, the software developer transforms into a specification developer. It is too early to say where this will lead, however, not least because the problems afflicting GenAI systems will not be overcome until the statistical approach is coupled with the symbolic one. The need to review AI-generated code in order to be confident it can be trusted therefore remains, at least until that integration occurs — and no one can reliably say when that will be. Which means that graduates in informatics and informatics engineering will always be in high demand — but that is a separate conversation.

And then there is the ever-present elephant in the room when it comes to informatics systems: security.

The Veracode October 2025 report on the security of GenAI-generated code reveals that 45% of it contains at least one of the ten most critical vulnerabilities identified by OWASP (Open Worldwide Application Security Project). For Java programs the situation is even worse. This sheds light on another important point: since GenAI tools have been trained on all existing code, and Java systems contain — for partly historical reasons — far more vulnerabilities than those written in other languages, the code that GenAI generates for this language is correspondingly more flawed.

Finally, the most damaging effect of all: the decline in hiring for entry-level software development roles. Because companies believed GenAI could handle junior-level tasks, hiring for these profiles collapsed. In 2024, the 15 largest Silicon Valley companies hired 25% fewer people for positions requiring less than one year of experience, while the reduction at start-ups was 11%. In its wake, we will likely also see a decline in enrolments in university degree programmes and upper secondary school diploma courses centred on informatics.

A suicidal choice, at every level. If companies do not hire inexperienced young people, they will never be able to develop experienced leaders. If young people do not learn to develop informatics systems with their own minds, they will never be able to manage those developed by GenAI.

Companies that have understood how to use GenAI are deploying it to help people be more effective in their work, freeing them from repetitive, low-level tasks. Those that are blindly relying on GenAI to get rid of new hires or pay them less will be forced to think again — and it will hurt.

The pendulum always swings back. What do you think?

--
The original version (in italian) has been published by "StartMAG" on 16 February 2026.

Four reasons to scale back expectations for generative Artificial Intelligence

by Enrico Nardelli

(versione italiana qua)

«Artificial Intelligence (AI) is going to reproduce human intelligence. AI will eliminate disease. AI is the single biggest, most important invention in human history. You've likely heard it all—but probably none of these things are true».

This is the opening of a special issue titled Hype Correction, subtitled It's time to reset expectations, published in December 2025 by the Technology Review of the Massachusetts Institute of Technology (MIT), one of the most authoritative sources on scientific and technological research in the United States and worldwide.

I had anticipated back in April 2025 that things were heading in this direction, a trajectory now confirmed by the analysis of such a prestigious publication.

You certainly remember how the launch of ChatGPT-3.5 (Generative Pre-trained Transformer) at the end of 2022 captured the world's attention, prompting both private and public entities worldwide to invest heavily in the LLM (Large Language Model) technology that underpins Generative AI (GenAI) systems. Many believed this was the path toward achieving Artificial General Intelligence (AGI), an intelligence similar to human intelligence but even more powerful, versatile, and tireless, which would free us from labor and find solutions to all our problems. However, after three years of continuous promises, the launch of ChatGPT-5 in August 2025 – perceived as merely incremental rather than revolutionary – began to make the exaggeration behind that vision increasingly evident.

In October 2025, I pointed out the first signs of a course correction already underway.

Summarizing what has emerged from various reports and studies conducted in 2025, there are four key elements that explain the current state of affairs.

1. LLM-based systems are not the path to AGI. This has been stated, among others, by Yann LeCun, who served as Meta’s chief scientist until November 2025 and left the company precisely because he disagreed with its continued insistence on LLMs. Ilya Sutskever, former Chief Scientist and co-founder of OpenAI (the company behind ChatGPT), has observed that LLMs' ability to generalize – that is, to extract general principles and apply lessons learned during training to new situations – is far more limited than that of human beings. In January 2026, The Atlantic (one of the oldest and most prestigious magazines in the US) published an article declaring "Large language models don't ‘learn’ – they copy," based on work by researchers from Stanford and Yale, who succeeded in getting four of the most widely used GenAI systems to reproduce nearly entire books or very large portions of them. A scientific review by researchers from Caltech and Stanford, published in January 2026, highlighted how even the most recent models, even those presented as "capable of reasoning," actually have significant problems reasoning correctly.
2. GenAI systems remain prone to hallucinations (i.e., making things up) at a rate estimated between 15% and 25% — an unacceptable level for most consequential decisions and interactions, in both personal and professional life. This is especially troubling because, unlike a human being, these systems are incapable of self-correcting through experience. Any ordinary worker may make mistakes at first but usually learns and improves. This does not and cannot happen with GenAI systems, precisely because they are based on an essentially statistical learning of language, namely on how frequently words appear near one another, and they lack causal reasoning capabilities. The special issue cited at the beginning observes how surprising it is that this approach managed to create artificial systems that produce human-like expressions when prompted with any question, but the fact that we perceive them as intelligent is our own projection. See my first two articles on the subject from March 2023 and April 2023.
3. For routine tasks, GenAI systems can outperform the average person, but because they fail to deliver expert-level performance reliably in real-world contexts, they have not managed to drive meaningful productivity gains at the enterprise level. We were misled when we saw successive versions of these systems pass professional qualification exams, but it later became clear that such performance was largely due to having “memorized” all available test materials in those fields rather than to any genuine understanding of their core concepts. In the words of Andrej Karpathy (inventor of the popular term vibe coding, which we will return to in a future article), these are "versatile but shallow and error-prone" tools capable of helping ordinary people accomplish things they would otherwise need an expert for (such as getting the gist of a legal or medical document), but not easily integrated into a productive workflow.
4. Certainly, the majority of people now use GenAI systems daily, both personally and professionally, but in most cases they do so free of charge, given that at least a dozen companies make them available. The upshot is that, after a cumulative $600 billion in investment between 2021 and 2025, there is still no viable business model — and this is prompting investors to rethink their positions. It is no coincidence that talk of a bubble began circulating in 2025, including from prominent industry figures such as Sundar Pichai, CEO of Alphabet (Google's parent company), in November of that year. Daron Acemoglu, 2024 Nobel laureate in Economics, analyzed the influence of the entire AI sector on the US economy through 2035 and concluded that only about 5% of tasks will be effectively performed by AI, and GDP will increase by only 1.1% to 1.8%. One capability still lacking is the ability, given a specific work situation, to reliably provide context-dependent information to solve emerging problems. Indeed, in January 2026, a Washington Post article reported that «economic data shows the technology largely has not replaced workers», and the Remote Labor Index analysis, conducted jointly by the Center for AI Safety and Scale AI, confirmed Acemoglu's predictions for now: on average, only 2.5% of jobs posted on a platform offering paid tasks to independent workers were successfully completed by leading GenAI systems. Additionally, also in January 2026, a survey by Apollo Global Management (one of the world's largest investment management firms) of CFOs (Chief Financial Officers) showed that the majority of them in 2025 «are seeing no impact from AI on labor productivity, decision-making speed, customer satisfaction or time spent on high value-added tasks».

None of this means GenAI tools are useless — far from it. They do augment our cognitive capabilities, provided we scrutinize their outputs carefully. They are extremely useful for carrying out routine tasks in areas we already master (so that we can correct any mistakes). A very recent example is their use as evaluators of the scientific rigor of theoretical computer science papers, which according to 81% of authors helped increase clarity and readability. They will certainly continue to improve, although to achieve major leaps in quality, it will be necessary to integrate them with systems based on a symbolic approach, and it is far from clear when this will happen.

There is still a long road ahead. What do you think?

--
The original version (in italian) has been published by "StartMAG" on 16 February 2026.

Informatics as a scientific discipline officially enters Italian schools: now the real challenge begins

by Enrico Nardelli

(versione italiana qua)

A historic step has been taken for the Italian school system. With the publication in the Official Gazette of January 27, 2026 of the new National Guidelines, informatics officially enters primary and lower secondary schools as a subject of study. This is no longer a matter of basic digital literacy or learning to use a computer – it marks the introduction of computer science as a fully-fledged scientific discipline.

This reform represents a highly significant milestone in a journey that began twenty years ago with the conference Informatics: Culture and Society. Informatics is being integrated into the curriculum through two complementary channels: the more abstract and conceptual aspects have been woven into Mathematics, while the more concrete and applied elements find their place within Technology. Although this solution does not establish informatics as a standalone subject—which would have required a more complex legislative process—it now provides the legal framework needed to offer students a comprehensive education in the scientific principles underlying the digital world.

The shift in perspective is a radical one. As the official documents make clear, the goal is not to produce passive users of technology, but informed citizens capable of critically understanding how the computing systems that permeate every aspect of contemporary society actually work. The reform aims to develop fundamental skills in young people, not just to use digital technology, but to understand, evaluate, and, when necessary, create it.

This turning point aligns with broader European Union (EU) directives. In November 2023, the Council of the EU published Recommendation C/2024/1030, calling on member states to improve the provision of digital skills training. With this reform, Italy aligns itself with countries such as the United Kingdom, which introduced compulsory computing education as early as 2014, and responds to the strategic need to close the skills gap in an era shaped by artificial intelligence (AI).

The rise of generative AI itself raises new questions. Some might wonder whether, in a world where systems capable of writing software autonomously are already available, teaching this subject still makes sense. The experts' answer is unequivocal: even in the most advanced scenarios, AI tools will only be able to support — not replace — the people involved in developing and managing software systems. Strategic direction, understanding context and objectives, and the ability to adapt systems to changing needs will remain human tasks. These require that symbolic and abstract understanding of the world that only humans possess, and the conceptual and technical ability to develop complex digital systems that only the study of informatics can provide. We will return to this issue in a future article.

The risk, however, is that a lack of public understanding of these dynamics could lead policymakers to withdraw their support for computer science education at the very moment when this competency is becoming most critical. For this reason, cross-party support for this reform is essential to ensure the continuity of a project that will take at least a decade to fully bear fruit.

The new National Guidelines also address the ethical and social challenges associated with technology. The text explicitly stresses the need for students to acquire «an understanding of how systems based on digital technologies function, together with an awareness of their possibilities and limitations, so as to grasp the enormous opportunities for improvement and development they offer society while preventing them from becoming instruments of exclusion or oppression». It reaffirms the fundamental principle that «human beings must maintain control over decisions based on computer systems that can have a significant impact on people».

The real challenge now lies in teacher training. Unlike other disciplines, informatics is not part of the cultural background of most Italian teachers, who studied it neither in school nor at university. The British experience is instructive: after introducing compulsory computing education in 2014, the United Kingdom had to acknowledge in 2017 that much of the curriculum guidance remained unimplemented due to a shortage of adequately trained teachers. The response was the creation, in 2019, of the National Centre for Computing Education, funded with £84 million and continuously refinanced to this day, despite cuts to public spending.

Italy has had access to funds under the National Recovery and Resilience Plan (PNRR), but organizational and time constraints prevented their use in establishing an efficient system for training teachers in informatics education. Strong political will is therefore needed to ensure that this reform does not remain a dead letter. The academic informatics community, which has been bringing the foundations of the field into Italian schools for many years, is ready to contribute. We recall the numerous initiatives already in place, such as the Programma il Futuro project in collaboration with the Ministry of Education and Merit, and the structured research and training activities carried out by the National Laboratory "Informatics and School" of CINI (National Inter-University Consortium for Informatics).

The road will be long and not without obstacles, but the first, fundamental step has been taken. taly has finally recognized that, in an increasingly digital society, the knowledge of the scientific principles of computer science is neither a luxury nor a narrow technical specialization, but rather an essential component of every citizen’s education. Only in this way will it be possible to shape generations capable not only of using technology, but of understanding its implications, assessing its consequences, and maintaining democratic control over systems that profoundly shape our lives.

--
The original version (in italian) has been published by "StartMAG" on 3 February 2026.

Informatics in schools: we’re there ... almost.

by Enrico Nardelli

(versione italiana qua)

The Council of State has just issued a favorable opinion on the New National Guidelines for early childhood education and the first cycle of schooling (i.e., primary and lower secondary school), in the version published by the Ministry of Education and Merit on July 7th. In September, the Council of State had requested clarifications—interpreted by some as a rejection, and by others as politically motivated obstruction. As I wrote when the initial draft was released in March 2025, which was then followed by a broad and in-depth consultation with social partners, the New National Guidelines represent a revision of the document originally issued in 2012 by the Ministry of Education, Universities and Research (Ministerial Decree 254 of November 16, 2012). That 2012 document establishes the common reference framework for educational provision, which each school then uses to define its own specific curriculum.

This is therefore a moment of real satisfaction for those who, like myself – with the support of outstanding colleagues, both Italian ones at the CINI National Laboratory "Informatics and Schools" and European ones within the international coalition Informatics for All – have been working for over a decade toward this goal. We are grateful to the Minister and the Administration for making it possible.

Why, then, the "almost" in the title? Because, as our British colleague Simon Peyton Jones – the architect of the introduction of computing education in the English school system – explained at the conference Teaching Informatics in Schools, held on October 19, 2023 at the Accademia dei Lincei, without a well-prepared teaching workforce in informatics, introducing the subject into the school curriculum risks even being counterproductive.

In fact, unlike any other subject taught in school – with the exception of the small number of teachers in a handful of specialized upper secondary tracks—none of the current teaching staff have ever studied informatics during their educational or university career. There is therefore no broadly shared cultural foundation to build on for professional development, as is the case, for example, with mathematics, which is taught in primary school by teachers who, although not specifically trained as mathematicians, have nonetheless studied the subject throughout their thirteen years of schooling.

With this in mind, as part of the Programma il Futuro project — which has been informally spreading the fundamentals of informatics across almost all Italian schools for 11 years — we will be hosting a webinar, led by our talented colleague Violetta Lonati, focused specifically on the curricular changes introduced by the inclusion of informatics in the New National Guidelines.

Looking ahead, I believe that the experience of the United Kingdom, as described by Peyton-Jones, points to the right path. After introducing compulsory computing education in 2014, it became clear that the curricular guidelines largely remained on paper due to the lack of adequately trained teachers. Consequently, in November 2018, the British government allocated £84 million over four years to fund an institute for teacher training in informatics education. This led to the 2019 establishment of the National Center for Computing Education, which developed teaching materials and training courses for teachers at all school levels and has been consistently re-funded ever since, most recently in 2025, despite public spending cuts.

This will therefore not be an easy task, and it will require strong political will, hopefully bipartisan, in the country’s strategic interest, since it is a process that will take at least a decade to become fully operational.

I am confident that such will can be found.

That said, while I understand the cautionary remarks expressed in the Council of State's opinion regarding the rationale for including informatics to address the challenges of digital transition: "the proposed approach could lead to an implicit and uncritical acceptance of the phenomenon as inevitable in its current form" (p. 27) and "the significant challenges arising from these transitions and their contested social and economic sustainability" (p. 28), I would like to offer some reassurance that this is not what is being proposed.

Having participated in the drafting of the New Guidelines as a member of the sub-commission responsible for integrated mathematical-scientific-technological education, I note that the text of the document is very clear on these aspects. With regard to the teaching of informatics in primary school, it states: "The acquisition of foundational elements of informatics enables pupils to begin developing, through exploration and experimentation, the cultural perspective this discipline offers, complementary to all others. In addition, it promotes a safe and responsible use of digital technologies" (p. 66 – my translation). The document further adds that in lower secondary school "... students move from a purely operational skill to a more critical and reflective understanding of the implications of technological choices and ... reflection skills on the social impact of information technologies are developed" (p.66 – my translation), clarifying that "Informatics provides an additional way to enrich the description of natural and artificial phenomena with a different perspective" (p.67 – my translation).

It is therefore clear that informatics education is grounded, from the outset, in the fundamental cultural dimensions of the discipline, precisely with a view to equipping students with a broad understanding of the complex social dynamics of contemporary life, and that the social criticality of digital technology use is explicitly emphasized. And that’s not all.

Since informatics appears in the New Guidelines partly as a more conceptual component embedded within Mathematics and partly as a more technological component within Technology, the description of the latter discipline clearly enunciates the critical issues to watch out for. Quoting verbatim from p.81 (my translation): «It is essential that pupils and students understand how systems based on digital technologies function, alongside their possibilities and limitations, so that they can grasp the enormous opportunities for improvement and development they offer to society while preventing them from becoming tools of exclusion or oppression. The teaching of informatics will therefore take into account interactions with civic education goals, emphasizing the importance of interacting with others respectfully, especially when using digital platforms. Students must learn how to identify and report problems in online social interactions, develop an understanding of the value of data – both at a personal and a broader level – and how the collection and processing of large amounts of data affects society. And it remains a guiding principle in the first cycle of education that human beings must retain control over decisions based on computer systems that may have a significant impact on people».

In short, the path is clearly marked: the person remains at the center of everything.

--
The original version (in italian) has been published by "StartMAG" on 16 November 2025.

The slop of generative AI

by Enrico Nardelli

(versione italiana qua)

I am writing this article prompted by a very recent post by my excellent colleague Walter Quattrociocchi. It is – to be honest – yet another in a long series where he explains on social media the dangers posed by generative artificial intelligence tools (GenAI, hereafter) for those who lack a solid understanding of the subject they are asking about. In his post, Walter recalls two recent episodes of sloppiness resulting from the use of these tools.

The first involves a print newspaper article published including the text of that standard closing question GenAI tools offer at the end of their response, asking whether the user would like to expand or adjust the text for a different publication context (e.g., a popular science piece rather than an academic one).

The second is that a well-known international consulting firm included fabricated citations in a—highly paid—government report and was, consequently, forced to partially refund its fee.

Neither example is new: the inclusion of AI-hallucinated cases in legal briefs has been happening in the United States for at least a couple of years, and forgetting to remove boilerplate phrases generated by GenAI in scientific articles has also been happening with some regularity for quite some time.

What Walter rightly observes—and I completely agree with—is that the issue is now systemic, widespread across all sectors and at every level. In his words: "We have delegated judgment to a statistical language engine, mistaking textual coherence for factual truth. This is the heart of the epistemic problem: the confusion between linguistic consistency and epistemic truth." A point that resonates with what I wrote in April 2023, warning that the main danger of GenAI tools is that «They have no real understanding of the meaning of what they are producing, but (and this is a major problem on the social level) since they express themselves in a form that is meaningful to us, we project onto their outputs the meaning that is within us», and recalling, two years later, that «showing competence with the words that describe the world is not the same as having competence about the world itself».

Two recent studies offer further evidence of how GenAI tools fall short of their potential and why, therefore, great caution is needed instead of the uncritical praise hailing them as something everyone must absolutely know how to use.

The first is the State of AI in Business 2025 report by MIT, which aptly speaks of a “GenAI Divide”, the gulf between the high rates of corporate adoption of these tools and the low return on investment observed so far. According to the report, 95% of companies are seeing no return on their investment, despite widespread use of the tools themselves. At the individual level, there is an increase in productivity, but this is not translating into corporate profitability.

The causes of these failures are a lack of robustness in supporting work processes, a failure to learn context, and a misalignment with daily operations. The judgment on scalability is particularly interesting: «the core barrier to scaling is not infrastructure, regulation, or talent. It is learning. Most GenAI systems do not retain feedback, adapt to context, or improve over time». I had already addressed this difficulty three years ago in my book La rivoluzione informatica (published just as the first such systems were gaining traction). Writing more generally of "cognitive machines" (the term I use for all computing systems), I noted that they "lack both the flexibility and adaptability to change their mode of operation as contextual conditions shift." It is gratifying to see one's own assessments hold up over time.

The 5% of companies that are succeeding are doing so by asking vendors to tailor these tools to the specific needs of their workflows and their own proprietary data (as I have always advised those who have privately asked me how to use GenAI within their companies). They are also evaluating effectiveness not based on generic benchmarks (like passing standardized exams in various disciplines) but by considering the actual output produced.

A subsequent study by a team of researchers from Stanford and BetterUp Labs further investigated this situation. Specifically, 1,150 US workers from various sectors were interviewed to understand how GenAI output is used at work. It emerged that while some use these tools to refine work they have already produced, many use them to churn out low-quality output that then has to be corrected by someone else downstream. The term "workslop" is gaining currency – where "slop" is an informal word for watery, unappetising food – to describe AI-generated content that looks passable on the surface but lacks the substance needed to move a task forward effectively. More formally, we might call it "shoddy work," and the result is that corporate productivity suffers.

In this study, 40% of the workers interviewed revealed they had received "slop" in the last month that was practically unusable. While the phenomenon primarily concerns peer-to-peer interactions, in 18% of cases it also occurs hierarchically, in both directions. The most serious long-term impact is on trust between colleagues, which is eroded whenever someone receives shoddy work. As with all previous technological waves, what is needed first and foremost is a clear identification of the strategy through which these tools can improve work processes, followed by the selection of technology capable of supporting that strategy and appropriate workforce training. In short, GenAI does have potential; it all depends on how it is used.

The Roman orator Marcus Porcius Cato, known as Cato the Censor, is credited with the saying «rem tene, verba sequentur», i.e., master the subject matter, and the words will follow, an admonition about the importance of thoroughly mastering a topic to be able to argue its cause effectively. Similarly, we could say in this case: "rem tene, mens mechanica te sustinebit," where "mens mechanica" refers to what I call "mechanical intelligence" rather than the more common "artificial intelligence".

Only if you know your subject well will GenAI be able to enrich and enhance what you do. This is why I believe the foundations of any area of knowledge must still be acquired, as they always have been, by “the sweat of one's brow”. Only then will we be able to use GenAI tools in the best way possible.

--
The original version (in italian) has been published by "StartMAG" on 12 october 2025.

Strolling through informatics #33 – Governing digital transformation in the national interest (part two)

previous #32 –––         

by Enrico Nardelli

(versione italiana qua)

We began reflecting in the previous post on the need for serious political action to govern digital transformation, analyzing the infrastructure sector, the data management problem and the case of education. We resume and conclude with this post, which is also the last in the series: I therefore take this opportunity to thank IRPA's Digital State Observatory for their hospitality, inviting those who wish to explore the topics covered in this "stroll" in greater depth to read my book The Computer Revolution.

In Public Administration (PA), we continue to emphasize the importance of digital transformation, without understanding that generals without adequate armies can accomplish nothing. In the absence of a significant number of hires of high school and university graduates in informatics, PA efficiency and effectiveness will not improve. And while I'm confident that "digital soldiers" emerge well-trained from our schools and universities, which prepare high school and university graduates in informatics who distinguish themselves wherever they go in the world, I have some doubts that public management is generally capable of understanding the importance and implications of using informatics in organizations. In many, too many, ministries and public entities, people have remained stuck with a vision of informatics as "siloed" mechanization of individual data processing functions, disconnected from a global and process-oriented vision of relationships between the organization itself, its interfaces in the PA network and citizens. All complicated by a traditional view of automation, according to which once the execution of a function has been entrusted to a machine, the problem can be considered solved. In the case of informatics this is not so, the true informatization of services is always under maintenance (see the post How to address digital transformation), because reality itself is in continuous evolution. But if you don't have "in-house" people capable of carrying out this maintenance, if you must continuously resort to external suppliers, costs and times balloon intolerably.

In the private sector we still have a cultural deficiency, discussed here, which has so far held back the productive system. But there's also another important consideration. Our industrial fabric is made up mostly of micro and mini enterprises, which have made creativity, flexibility and speed the key to getting from a defeated and destroyed country after World War II to the world's seventh industrial power. These don't use digital technologies or use them little not because – as some say – they are pervaded by "amoral familism" or haven't understood their competitive advantage potential, but because overly rigid computer systems would constitute the molasses that would kill them quickly.

The typical Italian industrialist – not at all stupid and who has made flexibility and speed of change his strategic weapon – has intuited this and stayed away from it: primum vivere! He understood that the only good informatics is "personalized" informatics that accompanies the company flexibly, as if it were a person, but which is able to work without getting tired and without making mistakes, as unfortunately happens to people. And this cannot be obtained simply by buying "turnkey" solutions, as happened with previous technological innovations, but is obtained only if you have available technicians capable of developing such personalizations as the need arises. While entrepreneurs have always been agile in modernizing production lines whenever they glimpsed a market opportunity, in the case of digital automation they have prudently remained on the sidelines, understanding the danger of its rigidity that risked making them waste economic resources.

The problem therefore is not only the culture of businesses, but also the vision of politics. Providing public funding for digital transformation of businesses can be a positive measure, but if it then materializes mainly in buying foreign machinery and solutions, it's difficult for it to contribute to the country's revival.

Alongside defining a development model designed in the interest of the entire country's growth, a massive program of support for training active workers should be implemented, which cannot be limited to the necessary end-user digital skills but should give them the conceptual and cultural tools necessary for Sistema Italia to be competitive in this area.

All this shouldn't be so difficult to understand, for those in politics. And I know very well that politicians – contrary to the common belief that wants them to be freeloaders – are super-busy at all levels, from local to national. I fear, however, that we're neglecting a bit too much a sector that is strategic for the country's future. This doesn't mean letting technicians take charge. On the contrary, in every era and country, those who wanted to bring technicians to government, presenting them as priests of impartiality, have always actually wanted to remove from the people (the demos) the power (the cratos) to control the government's work. They have therefore moved in an anti-democratic way. Digital technology is no exception to this. Computer systems are not intrinsically neutral. Digital transformation is not an absolute guarantee of efficiency and effectiveness. Technology must be, as always, at the service of politics to implement this or that decision.

Politics must decide how to govern this digital transformation, well knowing that epochal changes of this magnitude are not realized in the few years of a legislature and with unrealistic plans. This is why a cross-cutting agreement on a country development plan in light of the ongoing computer revolution is necessary, one that finds everyone in agreement, at least on some fundamental guidelines. We therefore need politicians who care about democracy and our country's future, who are capable of listening to what technology has to offer, of understanding what the possible social impacts might be, and of composing a synthesis of different social classes' needs in everyone's interest.

Strategic long-term state action is needed, similar to that carried out in the post-war period with IRI to rebuild an industrial fabric that would support the country's development: a strategic plan for "digital reconstruction" is necessary. I recall that IRI, with its public-private partnership model, was admired in Europe as a "third way" for growth, between market and nationalization.

I'm convinced that, in all parties, there are people of good will and great political capacity. Together, we can make it.

[[The posts in this series are based on the Author's book (in Italian) La rivoluzione informatica: conoscenza, consapevolezza e potere nella società digitale, (= The Informatics Revolution: Knowledge, Awareness and Power in the Digital Society) to which readers are referred for further reading]].

--
The original version (in italian) has been published by "Osservatorio sullo Stato digitale" (= Observatory on Digital State) of IRPA - Istituto di Ricerche sulla Pubblica Amministrazione (= Research Institute on Public Administration) on 21 May 2025.

Strolling through Informatics #32 – Governing digital transformation in the national interest (part one)

previous #31 ––– next #33

by Enrico Nardelli

(versione italiana qua)

As we discussed in the post "It's easy to say digital", the digital dimension is now closely intertwined with all dimensions of our existence and defines a social space that can be built according to different visions. For example, we recalled in the previous post how we've moved from a vision of society where role models for young people were individuals with careers that could provide good economic satisfaction for the majority of people, to a vision where these models are people of great talent (and great luck, because unfortunately talent alone is often not enough) with very high earnings, who however constitute a small minority of society.

The point is deciding which vision of society we choose. This is a typically political task, given that political activity (in the sense that Aristotle gave it as a natural component of being human) is precisely the way to decide what objectives to pursue and how to do so, within the communities where humans habitually gather, to meet everyone's needs.

In the same post "It's easy to say digital" I quoted Evgenj Morozov who wrote in his book "The Lords of Silicon": «for today's mass party, not taking care of its responsibility regarding digital matters is equivalent to not taking care of its responsibility for the future of democracy itself». Over the past ten years this statement has become even more true while, at least in Italy, politics – ever since we began talking about the "information society" in the 1990s – has largely chased fashionable terms, playing around with operational aspects, granting funding and loans for acquiring "turnkey" digital systems that after six months or a year require significant updating expenses. Instead, we should invest intensively in training internal personnel who can contribute to developing digital solutions and make them evolve in harmony with changing commercial scenarios. This is an essential element for relaunching the country's economic development, all the more relevant as services gain weight in the economy. Not to mention that in some strategic sectors, handing over "the keys to the house" to a stranger is hardly the smartest solution! While in industry the cost of production facilities is a percentually significant item directly proportional to production volume per unit of time, in digital service delivery the necessary "means of production" consist essentially of the brains of people who develop and update the software for their management, and are substantially independent of the number of users. Platforms like Uber and Airbnb are the most striking proof of this.

But services delivered through digital systems allow for flexibility and speed of adaptation at contained costs only on condition that one is able to make them evolve "in-house" or nearly so. Only this way can we obtain computer systems that can ensure increased productivity for companies as they evolve. For our production system, the challenge is on this terrain, which isn't a field reserved only for the "big players," since informatics – when used well – allows for advantageous exploitation even of very small market niches.

It's useless to think we can go back. Society will become increasingly digital: in this world, digital infrastructures constitute what a nervous or skeletal system is for mammals. Their control must be in our hands so that our country doesn't become a colony. We need to understand that digital space, like the solutions connected to it, must be managed in the nation's and its citizens' interests. We need politics that returns to exercising on behalf of the people that sovereignty which Article 1 of our Constitution assigns to them, without delegating it to Big Tech companies that have virtually unlimited reserves of resources to pursue their own interests.

We must maintain control over citizens' digital data, which in the digital world are the equivalent of the citizens themselves. Control of this data is what provides power in digital society. We must avoid leaving this control in the hands of private entities pursuing business objectives that are perfectly legitimate, but cannot override the common good. We must develop the country's capacity to create informatics services and systems. We've never lacked the will to work and creativity, and software development doesn't require heavy capital investments. However, it does require investment in training, which must begin from afar, starting in school. We should have started the day before yesterday, but – in this case it's truly so – it's never too late. How do we think we can manage Italy's digital transition if we don't provide all students with basic education in informatics? Every citizen, when seeing any machinery, no longer thinks it's some "devilish contraption" because they studied those basic scientific principles in school that allow them to understand there are no "miracles" in technology. What are we waiting for to do the same for the cognitive machines of the digital world?

Fortunately, it seems that at least in schools, with the revision of National Guidelines for the first cycle of education, we'll begin doing something concrete. It's been 11 years since we began, in collaboration with the Ministry of Education, with the Program the Future project to spread training on basic informatics concepts in schools, but it's truly a case of "better late than never!"

We'll conclude our analysis in the next post.

[[The posts in this series are based on the Author's book (in Italian) La rivoluzione informatica: conoscenza, consapevolezza e potere nella società digitale, (= The Informatics Revolution: Knowledge, Awareness and Power in the Digital Society) to which readers are referred for further reading]].

--
The original version (in italian) has been published by "Osservatorio sullo Stato digitale" (= Observatory on Digital State) of IRPA - Istituto di Ricerche sulla Pubblica Amministrazione (= Research Institute on Public Administration) on 14 May 2025.