Just Like Mathematics, AI Should Be a General Subject

There is little doubt that artificial intelligence (AI) has become an increasingly integral part of contemporary life, influencing how we communicate, conduct research, prepare reports, write letters and emails, generate images, develop software, and approach problems that would ordinarily require considerable time and specialised expertise. From conversational AI systems that assist with everyday tasks to sophisticated machine-learning models deployed in scientific research and industrial operations, the technology is steadily transforming the relationship between human knowledge and the tools available for applying it. Yet, despite this growing influence, an important question remains insufficiently addressed: who is genuinely positioned to benefit from the opportunities that artificial intelligence presents, and what determines their ability to do so?

My concern is that the benefits of AI are not necessarily distributed equally, even among people who have access to the same tools. At present, those who are already educated, professionally trained, and familiar with the conceptual language of their respective disciplines often possess a considerable advantage over those who are only beginning to acquire such knowledge. A software developer, for instance, is more likely to communicate a project's requirements effectively to an AI system than someone with little or no background in software engineering, not merely because the developer knows how to formulate a prompt, but because the developer understands the underlying concepts, technical terminology, architectural decisions, and constraints that define the problem. Similarly, a researcher who understands research methodology can evaluate an AI generated literature review more critically than someone who has yet to develop the ability to distinguish methodological rigour from superficial presentation.

This distinction is important because the quality of an AI assisted outcome depends on more than access to a sophisticated tool or the ability to type instructions into a chatbot. It also depends on the user's existing knowledge, clarity of thought, familiarity with the subject, and capacity to assess the accuracy and relevance of the response. In many professional contexts, those who understand the language of their disciplines are better positioned to formulate meaningful instructions, recognize inadequate answers, identify errors, and refine the results until they meet the demands of a particular task. AI may therefore amplify an individual's existing capabilities, but it does not automatically supply the foundational knowledge required to use those capabilities effectively.

Unfortunately, a growing number of young people appear eager to move directly from an elementary understanding of a subject to the production of sophisticated outcomes, sometimes overlooking the intellectual development that should connect these two stages. A student who has barely begun learning the fundamentals of a discipline may want to use AI to produce the final product associated with professional competence without first acquiring the vocabulary, principles, and reasoning skills through which that product can be properly conceived and evaluated. Although this approach may occasionally produce impressive results, it can also create a dangerous illusion of competence, in which the appearance of expertise is mistaken for a genuine understanding of the subject.

The solution, in my view, should not be to discourage young people from using artificial intelligence, nor should it be to restrict the technology to those who have already acquired advanced qualifications. Rather, we should begin preparing students early enough to understand what AI is, how it works, what it can accomplish, where its limitations lie, and how it can be applied responsibly to the problems they encounter in their education and future careers. To achieve this, artificial intelligence should be introduced as a general subject at the secondary school level and subsequently developed into a compulsory, discipline-specific course at the tertiary level.

At the secondary school level, AI education should provide students with a broad understanding of the technology and its place in contemporary society. Students should be introduced to concepts such as artificial intelligence, machine learning, large language models, training data, algorithms, and the basic principles underlying the development of AI systems, with the depth of instruction appropriate to their level of education. They should also learn about the practical problems these technologies are designed to address, the circumstances in which they can be useful, and the risks associated with inaccurate outputs, bias, privacy violations, plagiarism, and excessive dependence on automated assistance. Such an approach would help students move beyond the superficial experience of interacting with AI applications towards a more informed understanding of the systems they are using.

My position is reinforced by a recurring observation in conversations with young people and adults about their familiarity with artificial intelligence. When asked whether they know how to use AI, many respond that they do not, or that they are not particularly familiar with the technology. However, when the question is reframed to ask whether they use ChatGPT, Gemini, or similar applications, the responses often change considerably, as people begin describing the different ways in which they employ these tools for writing, learning, searching for information, and completing everyday tasks. This contrast suggests that some people may be using AI without recognizing that their everyday digital experiences are already connected to a much broader technological field.

The problem is not necessarily that these individuals lack the ability to use AI; rather, they may lack a conceptual understanding of what they are using, how the technology relates to other digital systems, and what possibilities extend beyond the particular applications with which they are familiar. Artificial intelligence is increasingly embedded in search engines, social media platforms, digital assistants, recommendation systems, and numerous other services encountered in ordinary life. When people understand AI only through a few popular chatbots, they may mistakenly regard it as a remote, specialized technology rather than a field that already influences many aspects of their daily activities.

This is one of the strongest arguments for introducing AI as a general subject in secondary schools. Education should not merely teach students how to operate a particular application; it should help them understand the broader principles behind the technologies they encounter, thereby enabling them to adapt as individual products evolve and new applications emerge. A student who understands the basic ideas behind machine learning and language models will be better prepared to evaluate unfamiliar tools, ask informed questions about their capabilities, and recognize that the technology extends far beyond the products currently attracting public attention. More importantly, this understanding can encourage students to approach AI with informed confidence rather than either uncritical enthusiasm or unnecessary apprehension.

At the tertiary level, however, a general introduction would not be sufficient. Although students across disciplines should understand the foundations of artificial intelligence, they must also learn how to apply it within the intellectual and professional traditions of their chosen fields. Consequently, universities and other tertiary institutions should introduce compulsory AI courses tailored to the specific requirements of their respective programmes, ensuring that students acquire not only general technological literacy but also the practical competence required to use AI meaningfully in their future professions.

For example, students of mathematics should learn how AI can support mathematical modelling, numerical computation, optimization, statistical analysis, and the exploration of complex mathematical problems, while also developing the ability to verify computational results and identify reasoning that is mathematically unsound. Students of law should be trained to use AI in legal research and document analysis while paying particular attention to the reliability of sources, the interpretation of legal provisions, and the ethical obligations governing professional practice. Medical students should understand how AI may support clinical decision-making and medical research without losing sight of the need for professional judgement, patient confidentiality, and rigorous verification. Similarly, students of engineering, education, economics, agriculture, and the humanities should be taught to identify the applications most relevant to their fields and to assess the limitations that accompany their use.

A central component of such training should be the development of effective prompting skills within each discipline. Prompting is often presented as though it were a universal technique that can be mastered independently of the subject to which it is applied, but this perspective overlooks the extent to which meaningful instructions depend on domain knowledge. An individual cannot consistently formulate precise, technically adequate instructions about a subject whose fundamental concepts and terminology remain unfamiliar. The quality of a prompt is influenced by the user's ability to define the problem, specify relevant conditions, identify the expected outcome, provide appropriate context, and recognize when the resulting response fails to satisfy the requirements of the task.

For this reason, students should be taught to formulate prompts using the concepts, terminology, methods, and standards of their respective disciplines, rather than relying on vague instructions and expecting the technology to compensate for every gap in their knowledge. Such training should go beyond memorizing prompt templates or collecting supposedly effective commands from the internet; it should cultivate the intellectual discipline required to communicate a problem accurately, interrogate an AI generated response, request meaningful revisions, and independently evaluate the final result. The objective is not simply to make students more proficient users of AI applications, but to develop professionals who can integrate the technology into their work without surrendering the judgement and responsibility that their professions demand.

Indeed, the emphasis on disciplinary knowledge becomes even more important when we consider that artificial intelligence can generate responses that appear convincing while containing factual inaccuracies, flawed reasoning, or unsuitable recommendations. Without an adequate foundation in the subject, a student may lack the means to distinguish a correct answer from a plausible but defective one. The ability to generate a response is therefore not equivalent to the ability to establish its validity, just as producing a mathematical solution is not the same as demonstrating that the solution is correct. Effective AI education must consequently preserve the importance of foundational learning, critical thinking, independent reasoning, and the responsible application of professional knowledge.

The proposed approach would also help address a broader educational challenge: the tendency to treat technological competence as a substitute for subject-matter competence. Artificial intelligence should instead be understood as a tool through which knowledge can be applied, extended, and, in appropriate circumstances, developed more efficiently. Students must first acquire sufficient familiarity with the intellectual foundations of their fields to understand what they are asking AI to do and why a particular result should be accepted, rejected, or revised. In this sense, AI education should strengthen disciplinary learning rather than displace it.

Just as mathematics is taught across educational disciplines because quantitative reasoning is relevant to a wide range of human activities, artificial intelligence deserves a place in general education because intelligent digital systems are becoming increasingly influential across society. The comparison does not imply that AI and mathematics serve identical purposes or should necessarily be taught in the same manner; rather, it highlights the importance of preparing students with foundational knowledge that remains useful regardless of their eventual career paths. At the secondary school level, this means cultivating broad AI literacy, while at the tertiary level, it means developing specialized competence through structured, compulsory courses that reflect the practical and ethical demands of different professions.

Ultimately, the objective should be to ensure that access to artificial intelligence does not become an advantage enjoyed primarily by those who have already acquired the education and professional vocabulary necessary to exploit it. If young people are introduced to the technology early and subsequently trained to apply it within their chosen disciplines, they will be better positioned to participate meaningfully in an increasingly AI-enabled world. They will also be more capable of distinguishing genuine expertise from the superficial appearance of competence, using technological assistance responsibly, and adapting to developments that may substantially alter the nature of work.