How to Think in the Age of Answers

AI is a know-how problem before it is a technology problem. Rheingold builds the point on Doug Engelbart’s 1962 framework, which described human capability as the product of four things working together: language, artifacts, methodology, and training. The artifacts have advanced far beyond what Engelbart could demonstrate. The methodology and training have barely moved. We have extraordinary tools and almost no shared practice for using them well, and the distance between the two is where most of the trouble lives.

We already failed the easier version of this test. When search engines arrived, the job of deciding what was true moved off the publisher or editor and onto whoever typed the query. Rheingold began teaching that skill to college students around 2005, and he watched students treat a high ranking in the results as proof of legitimacy. Schools taught that literacy poorly. AI raises the bar by an order of magnitude inside an attention economy already built to fragment concentration.

The trap is using the model like a search engine. Ask a question, take the answer, stop. Used that way, AI is a faster way to be confidently wrong. Rheingold’s alternative is dialogue, where you keep questioning the answer and value accrues through iteration. He notes that these systems can teach their own use. Tell the model what you are trying to do, ask what a good prompt would be, then ask why it is good, and the tool becomes a tutor in how to think alongside it.

These systems persuade in a way earlier tools did not. Stanford research behind The Media Equation found that people respond socially to media, relating to voices and screens as if they were people even when they know better. An LLM speaks fluently and carries the manner of a confident interlocutor, so its output earns trust faster than it should. Much of what it produces may be sound. Some of it is wickedly wrong. The fluency is what makes the error hard to catch.

The repair is developmental, and begins before dependence sets in. Rheingold worries that a generation could offload the cognitive work of writing and reasoning before building the capacity to do either. His answer is to strengthen the muscle early. Working with ChatGPT, he wrote a series of books to teach a five-year-old to notice his own thinking, form a plan, check why the plan failed, and try again, the scientific method rendered small. The aim is to arrive at these tools with enough judgment to use them as an amplifier rather than a substitute.