In his studies on patients with aphasia, Lev Vygotsky observed that, despite their differences, they all shared a common difficulty: a deficit in abstraction. When asked to verbalize sentences like “snow is black” or to deliberately misname the colour of a flower, they seemed blocked by an invisible force. In their world, words and things were inseparably bound, making any departure from reality impossible.[i]

In this disorder, Vygotsky recognized the patients’ inability to free themselves from the concreteness of things and immediate needs through symbolism and imagination—fundamental traits of human evolution. According to him, this lack of freedom from both internal and external stimuli is also present in children, as they have not yet fully developed the network of significations he refers to as “concepts”. The concept is, in fact, a form of generalization upon which thought relies to manipulate the givenness of things and suspend the immediacy of instinct. It serves as the lever that enables imagination to lift the world out of its immediate reality and to invent its void. For this reason, without concepts—without the capacity for abstraction—thought and language become prisoners of primal routines.

These brief introductory remarks help us contextualize the problem of language in the age of Large Language Models (LLMs) like ChatGPT. Recently, there has been considerable debate, particularly in the field of education, over whether these systems are harmful or beneficial to learning. Some argue that there is nothing inherently bad or new about these technologies, viewing them as just another milestone in the relentless march of technological progress. Others believe it is crucial to incorporate them into the classroom to better prepare our students for the job market. After all, technology is neutral—it depends on how it’s used, right?

But actually, no. It is now known that technologies – depending on their functions – have the ability to modulate our perceptions and even influence our cognition. For example, a study on the use of satellites and GPS technologies demonstrated how they negatively affect our capacity for spatial representation.[ii] This confirms, ex adverso, what had already emerged from a famous study on London taxi drivers, who had a significantly larger posterior hippocampus than the average person due to their constant exercise of spatial memory in a complex environment.[iii] In short, technology cannot be considered “neutral.” Like all practices, it alters the plasticity of the brain and the relationship we establish with ourselves, others, and the world. It is therefore legitimate to ask: what happens to our perceptions and cognitions when language is literally dictated to us by a statistical computing system? Answering this seemingly simple question is far from easy. In fact, opinions on the matter are numerous and often contradictory.

It has been argued—and rightly so, in our view—that within an educational system shaped by neoliberal principles (performance and achievement) and tools (metrics and measures), LLMs can only have a detrimental effect on students, who will seek the quickest path to achieve externally imposed instrumental goals (the highest grade with the least effort). The uncritical use of LLMs, to the detriment of the search for language—this blind faith in the new algorithmic Cassandra—is an experience that, willingly or not, almost all educators have painfully endured in recent years.[iv] But in its negativity, this criticism leaves open the proverbial glimmer of hope. If only the principles and conditions within which we are forced to live could be reversed! Was this not, perhaps, Walter Benjamin’s illusion in his analysis of the work of art in the age of its mechanical reproduction? Yet, is it still possible to maintain, as Benjamin did for art, that technology—by altering the function of language—frees it for new forms of perception and expression?

One cannot help but be sceptical. Indeed, the function of artistic and linguistic expression changes as technological and capitalist developments enable their evolution. And as these functions evolve, our practices change as well (consider, for example, the transition from the darkroom to digital photography). The new practices, in turn, modulate and transform our perceptions and systems of significations (a portrait, for Leonardo, endowed the human face with a value that a selfie has forever diminished). Nonetheless, this type of criticism ultimately remains external to the problem we seek to address. In fact, it does not confront the issue of language and its dictation, but rather takes it for granted. It is therefore useful to return to Vygotsky, who framed the question in a radical way.

Although originally separate on the ontogenetic level, Vygotsky argues that thought and language are phylogenetically interdependent. In other words, human evolution was made possible by their interaction, even in the absence of an inherent and original connection between thought and word. These two elements coexist and intertwine in the fundamental unit of verbal thought: the concept. The concept, in fact, represents a meeting point where thought takes shape through the pathways of language. We speak our thoughts because, as Vygotsky points out, thought “is not expressed but completed in the word.” [v] In the child, for example, while meaning and language are connected, they follow different paths. The infant’s world is full of meaning but poor in words. Through experience, this world of significance is domesticated through a new linguistic totality: the sentence. In a certain sense, the child moves from the whole of meaning to the whole of language. And the instrument the child relies on to navigate the sea of meanings is the conceptual platform, which over time becomes increasingly vast and populated. It is in this regard that, another authoritative Russian thinker, Pavel Florensky, argued that “words are the eyes of the mind, and without denominations of different kinds, there would be not only no science, but no perception at all.”[vi] Language enables thought to rework the totality of reality’s meaning—something that, in the child, initially manifests as synesthesia, or as a Baudelairean hyper-connectivity between the senses.[vii]

A key point in Vygotsky’s analysis is that the relationship between thought and speech is neither stable nor fixed; on the contrary, it is evolutionary and incremental. Like a plot that grows more intricate over time, the interaction between thought and speech deepens through their evolving interrelation. The adolescents, for example, replace the child’s play with imaginative thought, which enables them to make sense of the emerging complexity of their emotions. Yet this imagination does not discard the child’s world of play; rather, it builds upon its structures and meanings. By following the paths of signification, the adolescent thus learns to harness the original power of language: abstraction. This capacity for abstraction is, in fact, a force that liberates the imagination—not from the concrete, but from the immediate—allowing it to project itself into the future, to actualize its passions, and to rekindle its memories. As Vygotsky has it: “The formation of concepts brings with it, first of all, liberation from the concrete situation and the possibility of creatively reprocessing and changing its elements”.[viii] Ultimately, the relationship between thought and speech boils down to this: the freedom to imagine ourselves in time. Through fantasy, we anticipate the future and, “consequently, also creatively approach its construction and implementation”.[ix]

But precisely because the relationship between thought and language is not original, but historical, our concepts—or “word meanings”—are subject to the forces of experience. In other words, the relationship between thought and language—being a relationship—is always liable to restructure, unravel, or even break down, as we see in certain cases of aphasia. We can therefore return to our question: what happens to the adolescent when language is dictated to them, without that inner work between thought and word that we call the concept? What happens to our students when LLMs replace the activity we normally call “critical thinking”—an activity that is, in essence, nothing other than the abstracting and temporalizing force of verbal thought?

What emerges is a kind of latent aphasia, by default: an inability to think beyond what is dictated or what has already been said. And how many of us, as teachers, continue to be amazed by the creative incapacity of some students to think without guidelines, or by that often-irrational reticence toward so-called “theory”? Atrophying the relationship between thought and word through the dictation of language means fuelling the presentism of emotions, deepening the fracture—already widened by social media—between feelings and our ability to understand them through linguistic abstraction (“today I feel this way, but tomorrow my world could change”). Encouraging our students to use LLMs means turning them into editors of what has already been said and written; for, LLMs, like ChatGPT, are mostly trained—“statistically educated”—on the basis of digital texts collected from the World Wide Web.[x] Putting abstraction to sleep means preventing students from developing, autonomously, the ability to imagine and construct a future—whatever that may be. Through the often-uncritical use of such software, students are thus pushed to become knights of the Great Kingdom of the Statistical Past, which extends through the involuntary repetition of what has already been said and already been written. Ultimately, living under the shadow of a digital and disembodied Babel, which adapts to the goals imposed by the market of desires, means unconsciously surrendering our sovereignty and our democracy to those who now control the times and spaces in which we can experience them.

For all these reasons, we wish to sound the alarm. This is not about demonizing a technology that, after all, can be very useful for those who have already developed a personal and experiential relationship between thought and word. What we seek to defend is the freedom of our children and students to independently construct and experience the conceptual beauty of language—a beauty that, for Max Weber, represented the only form of theory. In other words, we want to enable them to cultivate the intimate capacity to reinvent time and, with it, the world into which we project ourselves.

If all this seems too speculative, a look at a contemporary work might help us concretize the meaning of what has been discussed so far. In the television series Adolescence (2025), the fractures between thought and word, language and world, emotion and reason—fractures we’ve already discussed—are portrayed with striking force. A 13-year-old boy, perhaps a victim of cyberbullying, kills a young girl but is unable to truly experience the gravity of his action, overwhelmed by the immediacy of the anger fuelled by his virtual ghosts. But this individual tragedy also serves as a metaphor for a broader social crisis. The worlds of adults and adolescents, in fact, are grounded in incommunicable significations: in the media world of youngsters, dreams and passions are formed, only to be hijacked by models that impose desires and hatred; in the world of adults, embodied language struggles to comprehend this virtual yet very concrete power wielded by social media. Through the exploration of these disconnected worlds, literally fragmented by the crumbling of meanings, the series confronts us with the ultimate problem of sense: how is it possible to kill at thirteen? Why? What is the meaning of such lacerating and unexpected violence? These questions, which haunt both the victim’s parents and those of the perpetrator, remain unanswered because, ultimately, the materiality of the act arises precisely from that abyss created by the separation between thought and word, imagination and language, emotion and reason.

And if we do not want our societies to end up resembling the father of the thirteen-year-old protagonist—sucked into the void of guilt opened by the senselessness of the act, in anguish over a tragedy whose meaning has been annihilated by the very technical means we use to communicate—it is essential to educate our students in the freedom to think through the tortuous and carnal experience of language.

Notes

[i] Vygotsky, L.S. (1998) ‘Imagination and Creativity in the Adolescent’, in Rieber, R.W. (ed.) The Collected Works of L.S. Vygotsky. Volume 5: Child Psychology. New York: Springer, pp. 151–166.

[ii] Hejtmánek, L., Oravcová, I., Motýl, J., Horáček, J. and Fajnerová, I. (2018) ‘Spatial Knowledge Impairment After GPS Guided Navigation: Eye-Tracking Study in a Virtual Town’, International Journal of Human-Computer Studies, 116, pp. 15–24. https://doi.org/10.1016/j.ijhcs.2018.04.006.

[iii] Maguire, E.A., Gadian, D.G., Johnsrude, I.S., Good, C.D., Ashburner, J., Frackowiak, R.S. and Frith, C.D. (2000) ‘Navigation-Related Structural Change in the Hippocampi of Taxi Drivers’, Proceedings of the National Academy of Sciences, 97(8), pp. 4398–4403. https://doi.org/10.1073/pnas.070039597.

[iv] See, e.g., HEPI (2025) ‘Student Generative AI Survey 2025’. Available at: https://www.hepi.ac.uk/2025/02/26/student-generative-ai-survey-2025/ (Accessed: 11 April 2025).

[v] Vygotsky, L.S. (1987) ‘Thought and Speech’, in Rieber, R.W. and Carton, A.S. (eds.) The Collected Works of L.S. Vygotsky. Volume 1: Problems of General Psychology. New York: Plenum Press, p. 251.

[vi] Florensky, P. (2007) Il simbolo e la forma. Scritti di filosofia della scienza. Turin: Bollati Boringhieri, p. 28.

[vii] See, e.g., Maurer, D., Gibson, L.C. and Spector, F. (2013) ‘Synesthesia in Infants and Very Young Children’, in Simner, J. and Hubbard, E. (eds.) Oxford Handbook of Synesthesia (online edn, 16 Dec. 2013). https://doi.org/10.1093/oxfordhb/9780199603329.013.0003. (Accessed 11 April 2025).

[viii] Vygotsky, ‘Imagination and Creativity in the Adolescent’, p.163.

[ix] Ibid., p. 165.

[x] See OpenAI (n.d.) ‘How ChatGPT and our foundation models are developed’. Available at: https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-foundation-models-are-developed (Accessed: 11 April 2025).