Generative AI and the Risks to Learning

At what age is it appropriate to use GenAI tools? To do what? And what kind of AI, general-purpose or educational? These are difficult questions to answer, as it really depends on what GenAI tool is concerned and for what purpose. For example, many non-generative educational AI tools such as intelligent tutoring systems are used in elementary school for the learning of mathematics and domestic language; those systems, which have been proven to be effective for learning, have become “hybrid” and can now utilize educational GenAI interfaces to make the tools more conversational. To the extent that they are effective and safe, there is no reason to discourage their use in primary education. However, should elementary school pupils be encouraged to do their homework, if any, with a general-purpose GenAI? Probably not. Most developers forbid the use of those tools below 13 in principle, but new models are being released for younger children. Is it possible to use general-purpose GenAI tools pedagogically with younger children? Possibly yes. GenAI will be increasingly part of the fabric of the digital content to which younger children are exposed, and having a teacher work with children on the AI generation of images to nurture a class discussion can be a sound pedagogical activity for example. While there is still little evidence on the inherent age-appropriateness of general-purpose tools, for learning this may relate to users’ level of proficiency (and desired proficiency) in the tasks performed with GenAI tools as much as to their age. Some level of mastery of the tasks delegated to GenAI tools (foundational knowledge and skills) is necessary to exercise critical thinking and judgment on the quality of their outputs, a prerequisite to learn something from using it. This is true whatever the student’s age.

Another important risk is the weakening of the essential human relation in education. While GenAI (and AI in general) present many opportunities for the personalization of education, if not properly used GenAI could also lead to more impersonal education, with an increasing number of tasks automated or delegated to AI chatbots with almost automatic human endorsement. Learning is a social activity where students acquire knowledge about a subject but also learn about social interactions at school and in society, skills that will be essential for their future working life and that need to be mediated by humans (teachers and other adults). Impersonal education can be desirable for learning in some cases though, for example when students are anxious to speak a foreign language and prefer to practice with a conversational agent. But an education in which both students and teachers outsource their tasks to GenAI tools may become the ruin of education, turning an enterprise of human development into a cycle with no human engagement.

The use of (general-purpose) GenAI may enhance performance in a study task without leading to knowledge acquisition. Learning requires the active and often social work of students’ mind. Much of the art of curriculum design and pedagogy is to create tasks that will elicit student reasoning and give them opportunities to advance it. None of this can happen if the student avoids cognitive effort or thinking. For example, translating a text into a foreign language with an LLM is unlikely to correlate with improved mastery of the foreign language. This has also been shown with the use of general- purpose GenAI as an adaptive learning system in mathematics or for improving one’s written text. What is in principle the product of learning (the proficient achievement of a task) may no longer reflect actual learning if GenAI is involved. This does not imply that one cannot learn by using general-purpose GenAI tools, or that educational GenAI tools have the same effect, but the use of GenAI tools should be guided by clear pedagogical objectives or educational purposes. Jurisdictions may consider informing and sensitizing parents and caretakers to this risk of using general-purpose generative AI tools when supporting their children, as well as the students themselves.