Education in the age of artificial intelligence

AI applications in education

A way to reduce educational inequalities?

One of the advantages of AI in education is the development of inclusive educational materials which integrate technologies to assist students with disabilities, such as speech transcription to help the hearing impaired or audio reading for the visually impaired. 

In developing countries, AI is also being used to close the literacy gap between low- and high-income students. For example, in Brazil, the Letrus programme is used by some 200,000 pupils and students to learn how to write. In sub-Saharan Africa, where EdTech (educational technologies) is already very present, the use of AI compensates for the lack of teacher training, like the Teacher.AI chatbot in Sierra Leone. 

While they can reduce inequalities in access to education, these solutions could exacerbate others, such as the new ‘digital divide’: no longer concerning the material availability of technologies, which is now commonplace, but education in digital technology at home, which is very unequal depending on social backgrounds. However, like overexposure to screens, harmful use of AI – such as a lack of critical distance from algorithms – also penalises the most disadvantaged. 

Finally, the educational tools currently marketed are based on algorithms that were not designed for educational purposes: large-scale language models, most often American, trained from data reflecting Western cultural norms and containing social, racial, and gender biases. As such, they risk aggravating existing prejudices through the content taught or ‘dropout’ identification systems. 

Personalisation at the heart of debates

AI’s core promise is personalisation of learning. As such, a host of adaptive tutoring platforms offer courses with personalised schedules, courses, exercises, and assessments. This ‘techno-solutionist’ mirage is denounced by sociologists, while schools desperately need more human resources. 

Educational sciences point out that this is an old promise with limited scope: the adaptive approach is currently solely based on performance criteria. It therefore risks creating a vicious circle and worsening cultural inequalities: a student who lacks self-confidence will process tasks more slowly and will be offered simpler tasks by the algorithm. 

Furthermore, the tools currently available do not really personalise exercises: they simply classify students with reference to typical paths, thereby tending to standardise learning... rather than individualising it! By reducing education to a series of tests and measurable results, as if learning consisted of accumulating information, these tools also risk privileging the most quantifiable subjects and neglecting the collective and social aspect of teaching and the individual needs of learners. Public research is therefore striving to design ethical AIs that support learners at their request, within the framework of an explicit dialogue, instead of automatically guiding them and delineating their path by prejudging their choices and level. 

Support for teachers?

AI is often presented as a useful aid to teachers, because most applications are supposed to accomplish some of their work, such as organising activities, designing lessons, and marking homework. AI can also be used to automate tedious, administrative, or educational tasks, such as marking. This promise of time saving, brandished since the existence of the first digital tools, is rarely kept: due to a lack of training, use of these tools has, to date, rather led to... an increase in working time. 

That being said, several tools can prove valuable, such as anti-plagiarism software, which will be improved with successive versions of ChatGPT. Generative models can also help write lessons, slideshows, and quizzes. A US study revealed that 50% of teachers already used ChatGPT in their work, including 10% on a daily basis! 

In any case, the tool must truly support the teacher, not replace them. Certainly, AI can contribute to designing lessons, but it is the teacher who sets the educational goals and controls the content, because the data feeding most algorithms is neither reliable nor recent. It is also the teacher who explains reasoning that algorithms do not always provide, their internal mechanisms remaining a ‘black box’ even for their designers. While waiting for transparent models, generative AI is a great opportunity to sharpen the critical thinking of young people, for example, by asking them to identify the limits and biases of ChatGPT. At university, some teachers are already using ChatGPT for this purpose. 

What school for the future?

If history is anything to go by, it is likely that AI will be incorporated into the education system without calling into question its founding principle: the teacher-student relationship. That said, the gradual privatisation of education could change the situation, with some researchers fearing that the use of ‘smart’ digital tools could lead to fewer teachers. 

One thing is certain: learning will be shaken up. A such, a new discipline is emerging: education in the functioning and use of AI, such as writing clear and unbiased requests (the famous 'prompts'). AI could also promote collaborative learning – by facilitating the formation of communities or working groups – or replace end of year exams with a continuous, fairer assessment of students, thanks to progress tracking systems. Beyond that, personalised companions could accompany individuals even after they start working, for lifelong learning. 

Finally, AI invites us to rethink educational practices by focusing on critical thinking rather than memorisation and assessment based on reasoning rather than on results. The most optimistic even predict that AI will help develop some of the noblest human qualities: critical thinking, sensitivity, imagination, and intuition. But to do this, we will have to teach everyone how to use AI wisely: by freeing ourselves from – and not by locking ourselves into – a reproduction of the world.