Back to A Better Way to Think About Tech.
How Should Students Use AI Chatbots to Study?
Cal Newport calls a careful chatbot habit techno-selectionism applied to large language model chatbots: choose uses that promote what you value and avoid harms, instead of using the tool for everything once you start. A listener letter had already drawn the same line for YouTube, Instagram, and TikTok, and for large language models. Over-reliance wastes time, while controlled access can help, and simply using them to answer questions and offload cognitive abilities is harmful. Asking them to quiz you or guide you without the answer is the premise of AI study tools. Careful use supports active recall, more practice questions, and repeated explanations that exasperate human teachers.
For a lot of basic mathematics, a chatbot is a pretty good tutor when the notes do not make sense. Ask why an expectation formula works, or show an answer that failed. Going back and forth on material with a clear right and wrong answer, on something the model was trained on, produces real tutoring. Demanding the answer turns cognitive offloading into cognitive surrender. Working the problem instead leaves you smarter, better on the test, and with a clearer grasp of the information.
A chatbot will also confidently give wrong information, so the subject area matters. Point it toward a page that already has an explanation and ask for a simpler version. It is more likely to follow the structure and get the explanation right. The standard is what makes life better, not what is easiest.
A Better Way to Think About Tech ·
From the transcript
We can apply our techno-selectionism maybe. Let's see. I'm excited about that. All right, here's what Robert said. In your book, Digital Minimalism, in the context of social media, you stressed how important it was to not simply run digital tools on autopilot. Over-reliance on YouTube, Instagram, TikTok, whatever can easily be a time waster, but controlled and responsible access to it can make it very helpful.
All right, so I'll pause here for a second. That's the configuring C I talked about, techno-selectionism, that idea of you can have rules about how you use technologies. It doesn't have to be all or nothing. All right, back to Robert's letter. The same principle applies to LLMs. Simply using them to answer questions and offload your cognitive abilities is harmful.