LLMs in primary education

I would like input on how AI, specifically LLMs, will positively contribute to primary education.

Hi Rick_Olson,

Great question; here’s my two cents (based on long-term close (indirect) knowledge of primary education).

Depending on implementation, many possibilities would appear to exist. After all, anything requiring some kind of linguistic communication could benefit, i.e. separate from other types of communication and activities in primary education.

Thinking about where linguistic communication could be supported by LLMs, the first thing that comes to my mind is ideational support for managers and teachers. LLMs may provide suggestions on how to structure activities, how to organize, perform and report evaluations, how to enrich educational material, how to enhance communication with parents and other stakeholders, and so on. The importance here is that the outputs of the LLMs do not have to be fully reliable, after all, they are just used as ideational machines.

More complicated uses could include using LLMs as question answering chatbots. In my view this is more risky as outputs of LLMs are not always reliable and this may lead to unwanted confusion and possibly conflict. So, having children or parents interact directly with LLMs would appear to be on the risky side. In my view, having children use LLMs directly in text generation could undermine the development of their own linguistic capabilities, but experimenting a bit could be useful to prepare them for the role LLMs are / will be playing in society at large.

LLMs can also provide translations which may be useful in multi-lingual education. So, they may support the teaching of language from the language spoken at home to a language spoken or taught in school. This may be very helpful for children from multi-lingual families and may provide support to teachers and managers in their communication with parents and children.

A final functionality of LLMs is summarization. I guess there could be various uses of this functionality, depending on the need for fully accurate summarizations (which would need to be checked).

Probably there are many more possible uses. I.m.o. the most important aspect is to use the LLMs in a reliable fashion to support activities in a responsible way and avoid confusion, misunderstandings and conflict.

I hope this is useful. I am sure others may have more informed ideas.

Hello Rick, I suggest you watch the TED Talk by Salman Khan last year, and the enterview that Bill Gates makes to Salman Khan. The central idea is that LLMs have the potential to become personalized tutors. The idea is based in Bloom’s research that states that the absolute best way to learn is having a 1 to 1 tutor, but since this is (was) impossible, educational researchers started looking for ways to replicate this finding in a scalable way. It seems LLMs colud become this 1 to 1 tutor figure that could help students improve their learning in a scallable way.