45 Teaching AI

At this point we consider the teacher, with respect to AI, savvy enough to use AI safely and in a way that adds value to the education process. The teacher may also want to share with their pupils some insider knowledge, or explain how some tool works. But that doesn’t give the teacher the role and task of teaching AI just yet.

Nevertheless, the question will be raised at some point. Is there a case for educating everyone to or about AI? And in such a case, what should be taught? Who should do the teaching? How much more will the teacher need to learn?

What we have learnt from teaching coding

Ten years ago, most European countries reached the conclusion that teaching children how to use a computer wasn’t good enough and that it was necessary to teach code (or sometimes, with more ambition, computing and informatics)^1,2. The arguments used then are probably valid today for artificial intelligence:

• Coding is as useful as writing and counting
• Many human activities benefit from coding
• Coding is related to other necessary skills such as problem solving

So coding was introduced into schools, but with variable success3. Insufficient resources were allocated to the human aspect of training the teachers. There was a complicated problem here – training the teachers too well could lead to their abandoning the teaching profession to work for the computing industry, where salaries are much higher! Reports from Informatics Europe and other organisations all show this (but there are exceptions, of course).

Training teachers has been a complex task in all countries, By 2023 the results are still heterogeneous. In most countries the feeling is that there are not enough properly trained teachers. This makes it especially complex to envisage training teachers to AI, at a level sufficient for them to teach AI (rather than teach with AI).

AI Literacy

The first goal could be to introduce some form of AI literacy in schools. But there is no agreement yet on what this literacy should comprise. Do we want to explain how AI works or just the results of AI? Does literacy consist of just understanding AI? What about the capacity to adapt and create? These questions need to be addressed. Perhaps, in order to know what should be taught in a course of AI literacy, the first question should be, what do we want to achieve?

AI literacy will allow people to differentiate between magic and science. In order to consider a new AI solution and have some intuition as to how it works (and not just what it does), practical training will be needed. Pupils and students will need to be able to test systems and know how these systems work.

Paradigms

AI isn’t only about algorithms. There are many human aspects, and questions need to be thought out. For example, most AI methods will rely, to a certain extent, on randomness. This may seem strange for techniques that are supposed to help us make some drastic decisions (or, in a growing number of cases, like that of the stock exchange, which enforce these decisions directly).

Yet if AI is going to play a key role in the future, should we not at least start?

In a report for Unesco in 2018⁴ it was suggested that the following five issues, mostly absent today in the education system, will need to be addressed:

1. Even if using the tools seems not to require direct coding, the reasoning behind the AI tools follows the rules, which can be learnt through coding.
2. Randomness matters. AI makes mistakes, and these mistakes are in many ways unavoidable. They can be due to the quality of the data or of the sensors; they will also be due to the statistical nature of the algorithms which are used, Most AI algorithms do not aim to be absolutely correct.
3. The world is no longer deterministic. This is a consequence of the above point, but the consequences are specific, as this is where we understand that an AI system can provide us with different, sometimes contradictory, answers to simple questions. Reading Alan Turing’s 1950 paper⁶ gives a lot of insight into these questions.
4. Critical thinking is essential but it’s necessary to know how to use the right tools. AI tools are getting better at creating fakes – images, videos and now texts. Soon perhaps we will have fake lectures. Common sense alone is not sufficient for us to make informed decisions when it comes to deciding if an image, a voice or a text is fake.
5. We cherish our values – analysing the world, making moral decisions and questioning why we spend time studying or working.

These values need to be scrutinised, considering AI’s progress.

The grey zone of truth is growing wider every day. Experience is perhaps not going to be valued when AI is able to refer to collective experience and crunch the numbers.

Understanding these issues, or at least enquiring, is a necessity.

Curricula and Frameworks

There are few AI curricula for K-12 and their teachers, available at the end of 2023^4,5. Unesco has started to survey these and present them⁸.

Unesco is key to education worldwide. Because Unesco is concerned by the Futures of Education⁹, it takes a special interest in AI for and in education. It provides documents for policy-makers and teachers on AI, education and ethics, and the use of generative AI in education. In 2023 Unesco experts have been working on documents describing what the competences for teachers and students should be¹¹. The final version is due for release in 2024. The 2023 version proposes aspects which balance technological questions and those more related to social sciences or, in the cases of teachers, to professional development issues. Even if coding isn’t immediately necessary, it would seem to be a skill needed for a better understanding of AI.

Coding AI

Coding, or programming, as an activity, has been promoted in most European countries since 2012. In 2023, the European Union supported the teaching of informatics in Europe.

But since the advent of generative AI and its expected impact on education¹⁰, the usefulness of learning to code has been questioned. Can’t we just let the AI perform this task for us? Or, on the contrary, since many jobs in the future will depend on AI, should we not learn to code in order to better use AI?

The main reason for learning coding is that a teacher or a pupil could be able to use AI in computer programs. There are a number of tasks involved with coding AI. Building models usually is part of data science and machine learning. A good coder can take a dataset, clean it without distorting it, and use it to extract rules and patterns through machine-learning algorithms. The programmer can specify the meaningful attributes or let the algorithm classify raw text or images. Some languages, such as Orange, are good at this. In other cases, a programmer will choose to use a general-purpose language such as Python.

¹  Royal Society (2012). Shut down or restart? Report of the Royal Society. 2012 https://royalsociety.org/topics-policy/projects/computing-in-schools/report/T.

² Académie des Sciences (2013). L’Académie des Sciences : L’enseignement de l’informatique en France – Il est urgent de ne plus attendre. http://www.academie-sciences.fr/fr/activite/rapport/rads_0513.pdf

³ Informatics Europe (2017).  Informatics Education in Europe: Are We All in the Same Boat?

⁴  Colin de la Higuera (2018). Report on Education, Training Teachers and Learning Artificial Intelligence. https://www.k4all.org/project/report-education-ai/

⁵ Touretzky, D., Gardner-McCune, C., Martin, F., & Seehorn, D. (2019). Envisioning AI for K-12 : What Should Every Child Know about AI ? Proceedings of the AAAI Conference on Artificial Intelligence, 33, 9795-9799. https://doi.org/10.1609/aaai.v33i01.33019795

⁶ A. M. Turing (1950)—Computing Machinery and Intelligence, Mind, Volume LIX, Issue 236, October 1950, Pages 433–460, https://doi.org/10.1093/mind/LIX.236.433

⁷ Howell, E. L., & Brossard, D. (2021). (Mis) informed about what? What it means to be a science-literate citizen in a digital world. Proceedings of the National Academy of Sciences, 118(15), e1912436117. https://www.pnas.org/doi/abs/10.1073/pnas.1912436117

⁸ Unesco (2022) K-12 AI curricula: a mapping of government-endorsed AI curricula. https://unesdoc.unesco.org/ark:/48223/pf0000380602

⁹ Unesco (2023). Artificial intelligence and the Futures of Learning. https://www.unesco.org/en/digital-education/ai-future-learning

¹⁰ Unesco (2023). Guidance for generative AI in education and research. https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

¹¹ Unesco (2023). AI Competency frameworks for students and teachers. https://www.unesco.org/en/digital-education/ai-future-learning/competency-frameworks