Didactics 3.0: Education and learning in the era of artificial intelligence

– by Marco Guarracino –

The debate: between enthusiasm and alarmism

The contemporary debate surrounding emerging digital technologies often tends to polarize between enthusiastic positions and technophobic tendencies. However, when the object of analysis becomes the very core of knowledge transmission – education and, specifically, the university ecosystem – we realize that the advent of generative artificial intelligence does not simply represent the introduction of yet another technical tool. We are witnessing a digital and cultural revolution that redraws the boundaries of the very identity of learners and teachers. Until a few years ago, software was conceived and used as an extension of human capabilities – sophisticated tools designed to speed up data calculation, storage, or formatting. Today, the ongoing transition presents us with a radically different scenario: AI has evolved from a mere tool to a "traveling companion" (Floridi, 2023) within digital and educational environments.

This shift introduces a new relational dimension. Students no longer simply query a database, but engage in a dialogue with a probabilistic agent capable of synthesizing, restructuring, and even simulating critical thinking. The impact on the traditional educational structure is profound and, in some ways, traumatic. The university, understood as a historical institution dedicated to the validation and transmission of knowledge, suddenly finds itself exposed, fragile, and methodologically ill-equipped to accommodate this fluidity. There is a clear divide between the exponential speed of technological development and the lengthy, reflexive, and bureaucratic timeframes of academic instructional planning.

As highlighted in a previous article (Guarracino, 2025), the need for a critical alliance between teaching and artificial intelligence is no longer a theoretical option, but a practical emergency that must be addressed. If in 2025 the first cracks in traditional assignment methods were visible, today those cracks have become structural fractures. Most current educational programs still rely on methodologies inherited from the analog era, in which learning is measured through memorization or the production of standardized written texts. But in a context where a generative linguistic model can produce an academic essay that passes traditional exams, the entire learning assessment framework is in crisis (Selwyn, 2024).

It is not simply a matter of curbing plagiarism – an anachronistic response that is already proving to be a failure and conceptually shortsighted – but of understanding that the metrics with which we have historically assessed skill acquisition have been disrupted. If writing a summary or solving a standard problem can be entirely delegated to an algorithm, what are we evaluating when we read an exam paper? Educational institutions are forced to embrace a cultural shift that impacts students' daily behavior even before it impacts teachers' curricula. What is lacking, at a systemic level, is the development of new forms of assessment that evaluate not the isolated final product, but the cognitive process, the research trajectory, and the student's personal reworking (Biesta, 2022). This requires addressing an extremely relevant conceptual and pedagogical issue, linked to an ongoing and increasingly rapid digital evolution.

The methodological challenge lies not in banning AI from the classroom – an attempt that would only create a senseless gap between the real world and the teaching space – but rather in rethinking the pedagogical posture itself. If AI sits beside the student like an omnipresent tutor, the university must transform itself from a place for "delivering answers" to a space for "formulating questions." This initial examination leads us directly to the epistemological heart of the problem: preventing the presence of this new traveling companion from resulting in a progressive and silent atrophy of individual critical capacities – in other words, preventing it from replacing rather than complementing them. 

The Paradox of delegation and new forms of learning

The moment we accept the presence of artificial intelligence as a relational actor in learning processes, we inevitably encounter the most insidious risk of this transition: the automation of thought through a sort of cognitive delegation. The core of the problem lies not in the technical effectiveness of the tool, but in the profound nature of the act of studying. Studying is not simply about accumulating information or producing a textual output that meets the requirements of an exam syllabus; it means engaging in a process of "construction," navigating uncertainty and doubt to arrive at a synthesis that is authentically one's own. When a student outsources the writing of an essay, the creation of a concept map, or the summary of an academic volume to a linguistic model, what is missing is not only the formal authorship of the writing, but the very process of sedimenting knowledge (Rovelli, 2024).

Complete delegation to AI creates an illusion of competence, which is even more dangerous when experienced during the study and learning phase. The user obtains a formally flawless product in a very short time, but experiences a weakening of cognitive control over the content. This new form of delegation must be understood and eradicated at its root, not through punitive prohibitions, but by reintegrating technology into a training program capable of reversing its meaning. The goal is not to exclude the machine, but to ensure that it adds value, stimulates applied criticism of teaching, and strengthens, rather than weakens, memory and analysis. Memory itself, often relegated to mere factual knowledge in the age of search engines, today regains a dramatic centrality: without a solid foundation of internalized knowledge, the individual lacks the logical and cultural criteria necessary to validate, contextualize, and critique the output generated by an algorithmic system (Eco, 2014).

From these premises, epistemology takes on a completely new and contemporary guise. We must redefine what it means to "know" in an era where information synthesis is just a prompt away. This opens up new avenues and critical perspectives in which the student's personal development and growth are placed alongside the machine, rejecting a sterile substitution of actors. The convergence between artificial intelligence and applied learning must be structured as a dialectical exchange. AI can become a formidable critical mirror, a probabilistic interlocutor with which to test the validity of one's own theses, explore counter-arguments, or map novel interdisciplinary connections, provided that the process remains firmly human.

If we analyze the production of syntheses or conceptual models, we realize that they are extremely powerful pedagogical tools precisely because their effectiveness lies in the learner's effort at abstraction. Delegating this phase entirely means abandoning the cognitive training of critical thinking (Morin, 2015). Teaching must therefore evolve towards tasks that require active mediation. Students – and anyone else involved in studying and learning – are no longer asked to write a linear review, but to critically analyze a summary produced by AI, uncover potential "hallucinations," verify its historical sources, and integrate it with their own experience as scholars. In this way, learning is not subjected to technological innovation, but rather appropriates it, transforming a potential factor of intellectual laziness into an accelerator of awareness and personal growth. This is certainly not an easy or immediate process, but it is nonetheless necessary to offer new teaching and learning opportunities in a changing, sometimes disoriented, and yet highly digitally dependent context. 

Pedagogical experiences, technical demystification, and the crux of teacher training 

To prevent reflection on AI teaching from remaining confined to theoretical speculation, it is necessary to closely observe the state of the art of pedagogical experiments and, at the same time, understand the true technological nature of the tool we are engaging with. Often, in public discourse and even in academic contexts, AI is perceived through an anthropomorphic lens, almost as if it were a conscious entity endowed with intentionality and semantic understanding. This partial misunderstanding generates a counterproductive duality: on the one hand, the idealization of the machine as an infallible oracle; on the other, awe or unconscious rejection.

Today, the generative AI with which students interact – based on Large Language Models (LLMs) – is actually a sophisticated predictive and probabilistic system. It does not understand the deeper meaning of the words it generates; rather, it calculates the statistical probability that a given term follows another within a specific context, based on a massive amount of previous textual data (Bender et al., 2021). Understanding this probabilistic nature is the first, fundamental step in demystifying the tool. When a model "hallucinates" – that is, invents nonexistent facts or bibliographical references – it is not committing a careless error, but is simply performing its primary function: generating a statistically plausible text. It is precisely in this gap, between plausibility and truth, that the irreplaceable space of human intervention and the scientific method comes into play.

At the international and European level, several academic institutions have begun to take the first steps towards forms of active pedagogy that critically integrate AI. Consider the UNESCO guidelines (2023) or the first university frameworks that have overcome the logic of prohibition to experiment with flipped classroom models and process assessment. In some science and humanities departments, AI is being used as a "devil's advocate" in debate sessions, where students must defend a thesis by refuting objections generated in real time by the algorithm. Other teaching experiences involve the critical analysis of outputs: the exercise no longer consists of writing a text, but of correcting, expanding, and validating a machine-generated text, tracing primary sources and verifying the logical coherence of passages.

However, the large-scale implementation of these innovative methodologies clashes with the most critical element of the entire system: faculty training. Too often, training programs for university staff, teachers, and professors turn out to be sterile bureaucratic procedures, where form systematically prevails over substance. This creates glaring short circuits in which training is made mandatory, or even encouraged, but not fully understood, precisely because basic computer science training is sometimes weak or completely nonexistent. Teachers cannot be expected to design innovative teaching methods based on prompt literacy – defined as the ability not only to formulate effective instructions for interacting with machines, but also to critically understand the algorithmic logic underlying their responses – without first having a solid familiarity with the logical structures of digital culture (Rivoltella, 2020).

Without a foundational digital literacy, refresher courses risk becoming a parade of "abstract technicalities" that increase the sense of disorientation and frustration among teachers and, more generally, all training staff. This cultural divide fuels conflict and protective resistance against an AI that is assuming increasingly central roles, but which is inevitably perceived as a threat to one's own authority, not only in teaching. Teacher training must therefore be radically rethought and recalibrated over time: before teaching the use of individual software programs, we must deconstruct the false myths about artificial intelligence, providing adequate basic training on the mechanisms and fundamentals of computer science. Only if understood, dismantled into its algorithmic mechanisms, and consciously guided can AI overturn the current paradigm, ceasing to be a factor of uncritical impoverishment and transforming into an additional resource within a comprehensive and governed educational curriculum.

Intellectual responsibility and methodological horizons

Assessing the impact of artificial intelligence on education requires a shift in approach: we must abandon the logic of technological emergence and embrace that of design responsibility. As we have attempted to demonstrate in the previous sections, AI is not an autonomous entity endowed with objective truth, but rather a relational and probabilistic ecosystem. Consequently, the relationship we establish with it cannot be limited to a blanket delegation, which would inevitably atrophy the critical capacities of the user and the learner, but must be structured as vigilant and mediated support.

The crux of this transition shifts from the merely technical to the ethical and hermeneutic level. In a context where content generation is massive and instantaneous, verifying sources, cross-checking data, and the philological validation of texts become primary teaching skills (Floridi, 2023). The written work, summary, concept map, or code generated through machine interaction cannot be considered finished products to be passively handed over; rather, they must represent the starting point for a process of personal reworking. A key principle – pedagogical even more than legal – must apply: everything produced with the aid of artificial intelligence must remain ours, and the final output falls entirely under our intellectual and scientific responsibility. AI extends our computational capacity, but it does not relieve us of the burden of judgment and signature.

This leads us directly to the issue of the "learning method," which represents perhaps the greatest opportunity for reflection for all stakeholders, from teachers to students and institutions. If interpreted exclusively as a shortcut to circumvent the burden of studying, AI becomes a factor of cognitive impoverishment and isolation. Conversely, if understood as a support tool, it can usher in a radically innovative and effective teaching system. From a perspective of democratizing knowledge, let us consider the inclusive reach of this technology: for the first time, every student with an internet connection can benefit from additional teaching support – a flexible tutor capable of reshaping explanations, simulating tests, and filling gaps, complementing, and never bypassing, the irreplaceable role of the human teacher (Selwyn, 2024).

The challenge of the method is to move from an educational model based on passivity and compliance to an open, potentially infinite system oriented toward active research. The critical alliance between the human and the artificial does not close the doors of knowledge, but opens them to broader horizons, provided that the centrality of personal growth and awareness of one's own cognitive processes is maintained.

In this rapidly evolving context, this article does not intend to put a final word or offer definitive answers, but rather to open new avenues for investigation and experimentation. Profound questions remain that deserve to be explored in future contributions to the magazine: how will the systematic adoption of AI redefine the design of our universities' physical and virtual spaces? How will the labor market respond to a generation of graduates trained within this human–machine convergence? And, above all, how will the very concept of "authorship" evolve in scientific research and academic writing over the next decade? These are complex questions that remind us that the journey into digital culture has only just begun. 

References

• Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM FAccT Conference (pp. 610–623). Association for Computing Machinery.
• Biesta, G. (2022). Riscoprire l'insegnamento. Raffaello Cortina Editore.
• Eco, U. (2014). La bustina di Minerva: Cara tecnologia, ti scrivo per non dimenticare. Bompiani.
• Floridi, L. (2023). L'etica dell'intelligenza artificiale. Sviluppi, opportunità, sfide. Raffaello Cortina Editore.
• Guarracino, M. (2025). Didattica e intelligenza artificiale. DigItalia Magazine. https://digitalia-magazine.blogspot.com/2025/05/didattica-e-intelligenza-artificiale.html
• Morin, E. (2015). Insegnare a vivere. Manifesto per cambiare l'educazione. Raffaello Cortina Editore.
• Rivoltella, P. C. (2020). Nuovi alfabeti. Educazione e culture digitali. Scholé.
• Rovelli, C. (2024). La fiducia nella scienza e il pensiero critico nell'era degli algoritmi. Adelphi.
• Selwyn, N. (2024). Should robots replace teachers? AI and the future of education. Polity Press.
• UNESCO. (2023). Guidance for generative AI in education and research. UNESCO Publishing.