A quick look at the comfort, learning and danger zone with AI
2025-01-12
Recently, I had the chance to discuss the impact of AI on learning and productivity with a good friend of mine, whose currently working in this field at a major research instiution. Quite a few things from that conversation were really valuable insights and I might or might not share them in the future, but today we are going to talk about the one aspect that resonated the most with me - apply the learning zone model to working with AI...
The learning zone model basically tries to put every task you do into three different zones - the comfort, learning and danger zone - in order to determine its complexity and managability for you. Tasks in the comfort zone are easy for to do, already have a lot of routine and don't offer anything new to learn from - the just to be get done for the most part, some examples include writing summaries, protocol or an e-mail to your boss. Next are tasks from the learning zone that include something new you haven't done before or you don't have much experience with, but still quite managable for you with enough time and ressources and thus offer you the opportunity to learn - examples might be working with a new software framework you haven't used before, writing a new kind of report et cetera. Finally, there are tasks from the danger zone that completely exceed your current skillset or knowledge and which you can't possibility solve on your own without major help - like building a chat app without any coding skills or launching a business without research.
With that out of the way, you might ask: So what? Where's the connection to AI? Well, as we all know AI (specificially generative AI like LLM) is becoming increasingly capable and not using it for certain tasks is already a competitive downside. So when should and when shouldn't you use AI to outsource work? That exactly where the learning zone model comes into play. Generally speaking, everything in the comfort zone should be done using AI tools to speed up the process as you can always correct and mistakes the model does and just speed up the whole process. For the learning zone using AI models is still advisable to some extent to get a good starting point, easy to understand explainations and especially answers to specific questions and examples of how things can be done. Here the AI is already more likely to make mistakes, but you are still familiar enough with the topic to evaluate whether or not you can trust these systems and their usage still provides positive values. This changes in the danger zone, where the tasks get increasingly more complex (also for the AI to solve) and these systems still make a lot of mistakes due to their limited design and reasoning capabilities. Most importantly, you as the human responsible can't evaluate whether the generated results are correct or total BS and thus shouldn't trust a single word of what the AI has generated. This is where the limits of current technology are reached and you should resort to just going the old style route of learning a topic yourself and doing it manually.
Generally speaking, I currently stop using AI once the systems starts to generate increasingly general answers that have little to no relevance to my specific prompt or once the answers just include definitively wrong statements. With that said, I hope you found this quick look at the comfort, learning and danger zone with AI interesting and learned a thing or two. As always, feel free to share thoughts and experiences in the comments down below and have a lovely day...