Education in the Age of Artificial Intelligence: Why automation should not be the only solution?

Education is an area where AI technologies and their impact are currently not fully understood. I argue that perhaps a more appropriate role for AI in education is to provide opportunities for human intelligence augmentation, with AI supporting us in decision-making processes, rather than replacing us through automation[1].

To provide augmentation to humans in education, we need to explore the ability of AI systems to cope with the complex social contexts in education and to serve learners equally and equitably as appropriate. However, whether we will ever attain the level of AI maturity that will enable fully automated systems to be part of everyday activities in education systems at scale is still an interesting question that needs to be further investigated.

It is important to note that most AI in Education research has been on attempts to create systems that are as perceptive as human teachers[2] and has been focusing on designing autonomous tutoring systems. As long as this automation aims to benefit humanity, in other words, take humanity as an end in itself, it can be considered as ethical in Immanuel Kant’s terms and it should be welcomed. However, in education, this can only be true when the needs satisfied are the ones that are endorsed and valued by key stakeholders of education: teachers, learners, and parents. In other words, as long as the key stakeholders of education’s own-reasoned motives are respected and support automation, there can be space for it. After all, in one sense, such automation is an augmentation itself as it provides productivity, albeit productivity is not the best proxy for human benefit.

The real problems with automation start when AI tools used for automation are driven by values that are not shared by the key stakeholders named above and humans are used as a means to achieve an end that doesn’t benefit humanity. Therefore, I would argue that, where we are not sure that humans are an end-in-themselves, we should avoid fully automated AI systems in Education.

It is important to emphasize that there are specific areas in which machines excel to have unfair advantages over human cognitive capacities. These can be considered as perfect intelligence for specific tasks that always make the correct decision in particular contexts. However, most decisions educators take do not have one correct answer to be always found. What expert educators do instead is to identify the best possible action to take based on the information and resources available to them. This best possible solution is not the one that is humanly possible nor it is the one artificially possible. It is the one that is logically possible and reaching to this best possible solution often require a combination of what AI and human intelligence can offer.

Therefore, I would argue that in education, we should leave the complex decision-making in the trusted hands of human educators and other relevant human stakeholders, while augmenting them with ‘artificial intelligence’: automated mechanisms for various rich data collection and processing options. These non-autonomous human-AI hybrid systems might be more valuable for our education systems, where the ultimate purpose is to improve educational outcomes rather than improving the state-of-art in the field of AI[3].

In Education, AI systems should be considered a continuum with regards to the extent they are decoupled from teachers and learners, rather than only an approach to provide automation of their behaviors. I think that identifying the areas in which machines are best suited to complement human cognition is the basis for this ‘subservient’ role of AI. Hence, the next questions for all of us to consider are, what exactly do we want to augment in human cognition in education contexts and how can AI technologies help us achieve it?

Mutlu Cukurova

Institute of Education

University College London