teaching and ai: don’t do marketing, do research

I was yesterday at a presentation of a brand new AI-based teaching platform, intended for secondary schools. I came there because researching how to teach in the era of AI – I should say in the era of chatbots – is an emergency. There’s a lot of litterature on this coming out lately, and its quality is uneven, let’s say. I am trying to participate to several initiatives. Among them, the BloomingAI project. I read a lot lately on the topic of AI and history teaching.

Yesterday’s presentation of this new project was disturbing. The vocabulary was marketing-oriented, not research-oriented (but, well, why not, it was a launching event, not a research event). All the clichés I have seen in the worse articles I have read were mentioned, first-degree. «We believe the teachers have still a role to play» – but the student can ask an animated caracter to answer questions generated by AI (so, possibly without quality control). Clichés, because, though there is no litterature proving them, you regularly see them popping up in conversations or (low-quality) articles: “rightly used, AI can help us teach better”. Maybe, but it’s absolutely not proven. And, for now, it’s rather the opposite that is proven.

Demonstrating during a marketing event that your platform is doing better than «traditional» teaching is obviously quite easy: you just have to define «traditional teaching» as something with such low standdards than anything you do will help improve it. And it’s basically what happened yesterday: one of the use case was history. If you define teaching history as teaching a series of dates and events, anything you do will improve it. The only problem is: we have decades of pedagogical and history research behind us that prove we don’t teach only dates.

We also have our personal and collective experiences. I changed how I teach since chatgpt went out. Not to integrate AI, but to be sure the basic historians skills are effectively learned by my students. At the bachelor level, I implemented a flipped classroom pedagogy, so that I am sure students learn how to understand, analyse, comment a historical documents without AI. At the master level, we investigate with my students how chatbots are considering the historical past. And they are all the more able to do this, that they have learned to make the critique of a document dealing with the past without AI.

I am not an AI sceptic. Nor am I an AI apologist. I am a technophile luddite: I test technologies, use bricolage (if not braconnage) to implement them in my research if I judge they can help me improve my research. My aim is that researchers – as well as teachers and learners – remain the master of their future, which, with the current surge of (sometimes sloppy) AI innovations is not an easy task. But rigorously testing those technologies is a pre-requisite – a pre-requisite to elaborate a sound critique of those technologies. And as long as this sound critique has not been made in the case of teaching, don’t do marketing. Do research. Otherwise, you’re just a techosolutionist raising expectations that will never be met.

And don’t bother me because I’m using hyphens – I loved using cadratins before chatgpt even existed in the mind of Sam Altman.