Our System is Designed to Enable Cheating
Hi Franklin,
I spoke about this Cognitive Debt report at a conference in Virginia this week.
I have a different take. Many students struggle with the cognitive load that writing requires and being able to offload that cognitive load to the AI makes it possible for them to express their ideas but not let sentence structure and grammar prevent them from sharing their ideas.
My oldest daughter has Down syndrome and she has ideas and beliefs, but has been unable to express them in typical situations most people encounter because of her disability. Using AI, she has been able to express her true beliefs and opinions in a way that was not possible before.
AI can take her jumbled typed message and turn it into something that is beautifully written, while still being true to who she is.
While these tools have made people rely on them in negative ways, AI has given my daughter a voice that only those of us closest to her have been able to hear.
Ethan Mollick wisely said something like this: AI is best when you are expert enough to spot the mistakes it is making, but really struggles at the work that made you expert in the first place.
For my daughter, it amplifies her ability and gets to the heart of what she’s trying to communicate.
So, there is an over-reliance that “smart” people are developing on AI, but it still has a place to help others who would struggle otherwise.
AI as chatbots in their current state are just the beginning. They will get better and be more focused on specific tasks. But we will likely need humans to spot check and proof for depth for a while yet.
The study was clear in showing that people who used their brain only to start performed better on the measures we would value as humans (remembering what they wrote and quoting).
The thing to remember is that tasks like writing teach us things as we do the activity. That productive struggle is valuable and we don’t want to do away with it.
There are certain tasks we should just ask AI to do for us, and others that we would be better served to do first, and then have AI give us feedback on improvements. This is a tough thing to ask when so many of our systems, especially schools, are focused on the end product.
Our current educational system is actually designed for cheating. When there is an emphasis on the final product, a short timeline, and incentives to complete the work to check the box, we set ourselves up for AI to be used to cheat. It would actually be surprising to NOT use AI when the incentives are how they are.
Rather than redesigning AI, we need to redesign our systems so that the goal isn’t just to complete our work, because as long as that is the goal, we will forever find ways to do that in the quickest, easiest way possible, even if that is a detriment to us later.