Lounge

https://en.wikipedia.org/wiki/Carnegie_Learning

And I am the creator of a Moodle Question type that uses AI/LLM for feedback.

https://github.com/marcusgreen/moodle-qtype_aitext/wiki

I've been in CALL (computer assisted language learning) for 25 years. I've seen new technologies come & go & I've studied the history of IT in education. Generally speaking, much of the research is of poor quality & there seems to be an inherent bias in favour, i.e. the people who conduct the research are often enthusiasts, even evangelicals in favour of IT in classrooms. That's not to say that there isn't good quality research around, there is, but it tends to paint a more modest picture of effect sizes & more caveats than enthusiasts typically speak or write about.

AI is yet another IT tool & it's going through the typical Gartner hype cycle at the moment, & there are some very vocal people with little or no knowledge or experience in education promising a revolution. They're giddy with excitement. Importantly, extraordinary claims require extraordinary evidence. I suspect that it's too soon for that evidence to have emerged yet.

What's more, as always, the method (strategies, techniques, sequencing, evidence-informed principles & practices, etc.) are more important than the medium used (The way that our food is delivered doesn't do much to alter its nutritional value, as the old analogy in education research goes). LLMs are generated from historical data & so will tend to entrench historical teaching methods & so far, the most popular historical methods have had a poor track record of delivering learning gains (e.g. See: https://www.tandfonline.com/doi/abs/10.1207/s15326985ep4102_1). If you prompt any LLM to design lesson plans or give advice about methodology, they return a barrage of edu-myths & poor pedagogical practices. Prof. Paul Kirschner lists 10 of the common edu-myths here: https://randomthoughtsandideas926468149.wordpress.com/2020/03/22/the-ten-deadly-sins-of-education/ & I've seen all of these, as well as assumptions about learning based on these, returned from LLMs consistently.

And, as Larry Cuban has frequently pointed out (https://larrycuban.wordpress.com/), education just doesn't do "revolutions" no matter how much the latest generation of pundits & billionaires would like to believe otherwise. Education changes in small increments &, in many areas, it still has a lot of growing up to do, i.e. becoming less ideological & more rational & evidence-informed & more responsive to the potential benefits of changing how classroom teaching is done, before it can be called a professional field in the way that medicine & engineering are.

Finally, I'm confident that some practical & beneficial uses of LLMs in support of education will emerge. However, I think it's still too early to predict what those might be. I have some ideas of my own but I lack the time & resources to run adequately controlled studies & I suffer from the curse of knowledge, i.e. I know a lot more about my field than the vast majority of teachers in it; what seems simple & straightforward to me is mind-bogglingly complex & opaque to most teachers (See: https://en.wikipedia.org/wiki/Curse_of_knowledge). Only when further independent evidence comes in & when I attempt to train other teachers in how to reproduce what I'm developing will the challenges to practical uses of LLMs in education become clear... at least to me. For example, knowing what a student needs to work on next, knowing the underlying science of how they need to work on it for good results, & then being able to design appropriate prompts to return useful responses from an LLM is something that many teachers are struggling with (generic prompts return generic, ineffective & often counter-productive responses). It takes a lot of specialist knowledge & experience (expertise) to be able to do that, let alone teach others how to do it.

Let's see what unfolds, shall we?