Better training thanks to artificial intelligence: where to start?

Son Thierry Ly, researcher in educational economics and founder of Didask, invites the relevant use of Artificial Intelligence in the field of training

Artificial intelligence and machine learning are on the rise, a craze that is sparing the world of education and training less and less. The promise of further individualization of learning paths through artificial intelligence is leading many education technology specialists to present theAdaptive learning like the future of the sector. But what is it really like? What are the conditions for the relevant use of artificial intelligence in education and training?‍

Putting technology at the service of learning

Indeed, just because it is now technically possible to offer 100% individualized courses does not necessarily mean that this is necessarily desirable from a pedagogical point of view. The main thing is not to multiply algorithms and technical functionalities, but to improve learning outcomes. The individualization of paths enabled by artificial intelligence can be a major vector of progress for learning, provided it is considered as a means and not as an end. To do this, it is necessary to put pedagogy back at the center and to focus on the natural mechanisms that make our brain learn.

Thanks to the results of research in cognitive psychology, we know what is not working: adapting courses according to multiple intelligences, for example. Proposing a diagram to a learner who says he is more “visual” and a voice-over to a learner who says he is “auditory” will be less effective than presenting a diagram AND a voiceover to both learners : this is what cognitivists call the double principle encoding (see Clark & Mayer).

What, on the other hand, could work? How to effectively integrate artificial intelligence into your training courses in order to allow a real and sustainable increase in skills? In particular, two paths deserve to be explored: one relating to learners, the other concerning teachers and trainers. These paths have in common put artificial intelligence at the service of educational effectiveness, rather than considering it as an end in itself.

Adapting courses to the (real) needs of learners

Learning involves assimilating new information. However, the processing of new information by the brain necessarily has a cost, which is called cognitive load (see our article on cognitive load). In order to avoid situations of cognitive overload, which are the main obstacles to learning, it is important to present the right level of difficulty to the learner at all times : not too easy, not too difficult. An artificial intelligence algorithm that offers each learner the concept that they are best able to understand at a given moment would guarantee you better learning results. It would also allow your learners to save time by not spending unnecessary time on topics they already know.

Attention, you cannot deploy this approach to any type of online training. The more relevant your educational content is, the more granular your learning data is, the more an artificial intelligence algorithm will be able to adapt the difficulty to your learnerss. Your courses should therefore have two characteristics to maximize the educational effectiveness of your algorithm:

• finely honed skills with clear educational objectives (see our article on educational objectives)
• numerous exercises to apply these skills (see our article on the testing effect)

Without application exercises, and therefore the possibility of making mistakes, you will not know when your learners are having difficulties. Without finely honed skills, you won't be able to accurately identify the source of their difficulties. This shows how the application of artificial intelligence to education must be based on a demanding vision of pedagogy.

‍

Facilitate the creation of pedagogically effective content

In education, the question of artificial intelligence is often asked from the learner's point of view. However, learning statistics are also a great source of recommendations for content creators.. One could thus imagine a “virtual educational engineer”, a kind of guide for training designers, whose role would be precisely to ensure that online courses meet the best teaching practices.

Based on the data of your learners, the virtual educational engineer could for example warn you when your explanations in text format are too long, when you would be better off inverting the order of the questions in your MCQ, or even when your video would benefit from being replaced by a diagram. Artificial intelligence then becomes the guarantor of sustainable learning.

Again, finely divided content that leaves plenty of room for exercise will give your algorithm the essential bases to work. Artificial intelligence algorithms do not only feed on data: they also need relevant data, and this is where the contribution of pedagogy and cognitive science becomes essential.

By having pedagogical effectiveness as the sole criterion, you ensure that you are not investing in technology for its own sake, but to achieve tangible learning results. After all, what's the point in creating machines that learn if we don't learn better from them?

Video : Son Thierry Ly, researcher in educational economics and founder of Didask, invites the relevant use of Artificial Intelligence in the field of training:

REFERENCES

Clark, R.C., and Mayer, R.E., 2008, E-Learning and the science of instruction: Proven guidelines for consumers and designers of multimedia learning, San Francisco, CA, Pfeiffer.