Futures in Context
Jannis Kallinikos
Professor at Luiss University and at London School of Economics
Jannis Kallinikos is professor at Luiss University and professor emeritus at London School of Economics. His research focuses on the study of information and communication technologies and the effects their diffusion has upon the functioning of institutions and the economy.
AI & EDUCATION — INTRODUCTION
The nature of artificial intelligence (AI) as a technology and its involvement in education make it necessary to trace its deeper links to learning, knowledge sharing, and communication. AI is not another set of tools adjacent to or on top of the digital and communication devices that mark and sustain the learning process. Its impact extends far beyond whatever specific tasks it is called upon to help accomplish, structure, or monitor. The widespread impression of AI as a single facility or tool masks the numerous, time-ridden technological innovations that have historically enabled mechanical systems to address problems of intelligence, understanding, and cognition. The cumulative and layered nature of these innovations makes current AI a composite, integrated, and infrastructurally embedded (i.e., the Internet, social media, the Internet of Things, other digital technologies and capabilities) ensemble of operations and methods that perform sophisticated intellectual tasks. These operations and methods are autonomously invoked by AI systems, each of which has been trained through exposure to vast data sources to distinguish, compare, and address a wide range of problems. In these qualities as autonomous intellectual technologies, current AI enters education and the learning process by the front door and inevitably becomes intertwined with how knowledge is comprehended, acquired, shared, and used. Its effects are much wider and more profound than they may seem at first glance, and will most probably expand and diversify in the years to come.
The autonomous nature of AI operations has sparked heated debates over whether AI is a friend or foe to humans, an ally in the pursuit of social and personal objectives, or, alternatively, a substitute for the capacities by which such objectives are sought. This dilemma is relevant and applicable to the broader context of learning and education. Transmission of knowledge, professional training, and learning all require building familiarity with the subjects taught, providing illustrative examples, reviewing real or simulated cases, practicing how to define and address problems, and, more generally, facilitating immersion in the nitty-gritty details of what is taught. Far from applying solely to practice-oriented fields, such as engineering or management, a learning process in which student engagement is a sine qua non characterizes all fields and subjects of science (social and hard), and subjects often erroneously labelled as theoretical. Considerable acquaintance with the subject taught constitutes a fundamental requirement for gaining proficiency in it. Widespread testimonies of students’ attitudes vis-à-vis generative AI and large language models (LLMs) suggest that AI can be used in ways that are not entirely consistent with a profound engagement with the problems and theories discussed in the classroom. Some of these uses may in fact stand at cross-purposes with the prolonged trial-and-error learning essential to the acquisition of critical faculties and the exercise of judgment.
In these qualities as autonomous intellectual technologies, current AI enters education and the learning process by the front door and inevitably becomes intertwined with how knowledge is comprehended, acquired, shared, and used.
LLMs could step into this process and serve as a substitute for engaging with class materials, thus bypassing the laborious, time-consuming tasks of defining and addressing problems, reading, reviewing, and assessing large bodies of literature, and comparing ideas, theories, and cases. Like physical training, the lack of continuous, real engagement with class materials can lead to the gradual atrophy of intellectual abilities, critical thinking, and the capacity to judge, all essential attributes of high-quality education and real learning. Why bother reading a text if ChatGPT, Gemini, or any other of the available LLMs can summarize and evaluate it in a few seconds? Worse, why bother learning things when answers are widely available on demand? This uneasiness remains pervasive among professors, teaching staff, university boards, and governing bodies, and has triggered incremental changes in teaching methods and, above all, student evaluation techniques. Slowly and without much fanfare, students’ widespread use of AI and LLMs has been coped with by changing the learning process and the classroom from the inside. Teaching tasks and roles, student assignments, and evaluation methods, among others, are rethought, restructured, or redesigned.
Some of these developments can no doubt be given a far more positive tenor. AI technologies can, after all, be used productively to enhance, rather than impede or undermine, student learning. It seems possible to encourage an attitude among students and academic staff alike whereby LLMs and AI technologies are leveraged as vehicles for discovering literature, helping staff and students access resources and processes that may not have been possible in the old classroom, leading to higher levels of engagement and learning in which critical mastery of class materials is likely. The belief or hope in the positive use of LLMs does not rule out its darker side. Hope is always easier to have than to realize. Whatever the case may be, it is obvious that current AI, either as a positive force or mostly as a bête noire, constitutes an agent of change in higher education and, most probably, across all levels and instances of education.
Upon closer meditation, these developments may come as no surprise. The original impetus for developing AI back in the 1950s was the dream of automating human intelligence capabilities and allowing mechanical systems to autonomously perform intellectual tasks. More generally, automation is the prevailing motive of modern technological history. Automation gives and, at the same time, takes away things from humans. It helps humans achieve objectives that would have been impossible to reach without its support, but step in their place, reframe their skills, and restructure their life patterns.
These may seem like utterly relativistic observations that avoid confronting the central question concerning whether AI technologies and the use of LLMs in education make the learning process better or worse. While straightforward, non-ambiguous answers to these fundamental matters are welcome, their complexity makes them rather tricky. Easy promises that claim quick fixes to these complex issues should be viewed with suspicion. In the areas where they are adopted, AI techniques interfere with established, human- and culturally based patterns of cognition, with poignant implications for learning and education more widely. AI-based object recognition goes a long way toward reframing aspects of perception and sense making. Natural language processing, which underlies LLMs, impacts upon and redefines language use, language understanding, and communication. It is therefore important to reiterate that AI is associated with massive changes that reorganize the human sensorium (e.g., image, sound, and face recognition), reconstruct analytic processes (reasoning and intelligence), and bring about new cognitive habits and interaction patterns.
Placed against this backdrop, AI technologies and LLMs represent a paradigm shift that challenges the very core of the learning process. It is for these reasons that such complex, infrastructurally embedded technologies can seldom be limited to mere tool use, while the outcomes of their intertwinement with learning practices are uncertain and often emerge along the way. While we have so far primarily referred to the context of higher education, these developments most likely extend to primary and secondary education as well. As AI continues to expand and engulf human intelligence and communication habits, it is hard to imagine, let alone predict, the exact paths its consequences for learning and education will take. As in many other occasions, the story of AI in education is most probably one of mixed blessings.
Some of these issues could perhaps be cast in an interesting light by the historical analogy of writing and, later, printing, and the contrast with the oral worlds which writing and printing came to remold. Their differences to AI notwithstanding (there is no autonomous capacity in these older technologies), writing and printing are, like AI, intellectual technologies. They are linked to cognition, imagination, and narration, as well as to memory (recording), communication, and knowledge sharing. Much has been written about the legacies that writing and printing have left in our traditions and societies, which cannot be reviewed here. It would perhaps be useful to note that, like AI, writing and printing have been met with enthusiasm and suspicion at the same time. In Phaedro, Plato expressed ambivalence toward writing and its effects on memory and cultural practice. In the years following World War II, Marshall McLuhan argued in his then much-acclaimed Gutenberg Galaxy that writing and printing give rise to linear patterns of thinking that may enhance certain capacities for reasoning but impair associative imagination and make memory fragile.
AI is associated with massive changes that reorganize the human sensorium, reconstruct analytic processes, and bring about new cognitive habits and interaction patterns.
On the other hand, it is equally hard, if not harder, not to see that writing, and especially printing, have been important means for shaping cultural imagination, promoting scholarship and knowledge, fostering the writing of novels and plays, and producing many other pervasive cultural goods of our times. It is likely that McLuhan’s worries may seem sweeping and exaggerated only because it is hard for us today to envision what has been lost, replaced, or superseded in this long historical process, and what traditions and cultural practices have been irretrievably relegated to oblivion. The mixed blessings of technologies and the cultural practices they support are never immediately evident. New learning targets, capabilities, and skills are not simply placed on the side or on top of existing ones. Most of the time, they go hand in hand with the restructuring of priorities, considerable social forgetting, and unlearning that tend to go unnoticed.
Furthermore, the patterns by which AI redefines the world of work and the broader economy are, by necessity, refracted through university education, impacting what is taught and, critically, how and by whom. The practice of mass education that universities entered into decades ago may not be sustainable in the future. At any rate, assessing the educational impact of AI may require a broader perspective on the viability of the prevailing vocational model of knowledge, research, and education, and the drift away from the Humboldtian origins and ideals of higher education. The current institutional structure of education (vocational and commercially oriented) may not withstand the ebb and flow of AI and the strong pressures it exerts to rethink the relevance of university curricula and the very process of credential procurement. All these are likely to drive shifts in academic roles, career paths, and priorities.
Against this background, it is reasonable to assume that the outcomes of AI for learning and education will remain ambiguous, likely to evoke both hope and fear.
