I have the pleasure of teaching students who, on the whole, are curious and eager to learn. This also applies to the methods classes I teach, and methods do have a reputation for being hard. When things get too hard, ChatGPT (and its siblings) can promise an easy way. Sure, code produced by LLM often works (there’s plenty of material on the web to learn from), but what do you do when it doesn’t work? Can you debug it?
Much of the code that I come across in this way is over-engineered, and it has an inconsistent coding style. On the one hand, I don’t care about style: if your code gets the job done, fine. The hard bit remains the interpretation, giving meaning to the numbers. On the other hand, the result can be much more complicated than necessary, additional frustration, and nothing learned — that transfer we’re looking for in education, the ability to solve the problem in a different context?
If you have to start from scratch every time, it’s not going to be efficient…
Published 3 January 2025