Pages

Monday, March 13, 2023

ChatGPT Will Power Personalized Learning

I don't get into the business of predicting the future often, and predicting tech future seems particularly pointless as it's an area dominated by aspirational marketing rather than actual prediction.

But I think I've found a use for ChatGPT and its brethren.

The problem with personalizing learning has always been capacity.

Earlier personalized learning systems (think that SRA reading box from your elementary days) were really personalized pacing systems. Everyone traveled the same path, but at their own personal pace. Subsequent attempts to personalize learning have attempted to create multiple branching paths, but the problem remains capacity and inventory. Do you create a worksheet about subject-verb agreement problems focused on 1920's automobile design on the off chance that some day, you may have the student who needs exactly that? You do not. 

But generative text bots can help fill this gap.

I cannot tell you how many hundreds of worksheets I created in my career, and I don't even want to think about the hours I spent on them. Custom made for classes (this class is having trouble telling adjective and adverb clauses apart, so let me bang out a worksheet focused just on that). A small cast of recurring characters with improbably names (so that I would never be targeting current or future students). The years I started every class with a three-sentence editing exercise. 

I've tried ChatGPT at generating basic grammar and usage worksheets. It does fine (sometimes, for no apparent reason, it even throws in an answer key). Even if I were taking time to edit the sentences to more carefully match what I needed (and fit my particular classroom sensibilities), ChatGPT would still be a time saver.

Likewise, I have no doubt that various purveyors of computerized personalized learning products are at this very moment figuring out how to incorporate the new generation of generative textbots to provide the kind of personalization they always claimed they were already doing. Scan the student response, generate a new exercise. Do I think that having software teach a child this way is ideal? I don't. Do I think we've arrived at the point where it can be done? I do. It's cheap, quick, easy, and solves one of the central problems of computerized personalized learning.

Check back later, when standardized test manufacturers start using chatbots to reduce the costs of coming up with test items. 


No comments:

Post a Comment