Can You Explain a Zipper? Why AI Makes You Feel Smarter Than You Are

Summary

The video examines the gap between feeling that you understand something and being able to explain it. AI tools can widen that gap by giving people polished explanations that feel like knowledge—even when the user has not built an understanding of their own.

The zipper test and the illusion of explanatory depth

The host invites viewers to rate how well they understand how a zipper works, then asks them to explain its parts and operation step by step: what the teeth do, how the slider moves them, and why the closed sides stay together. After trying to explain it, viewers are asked to rate their understanding again.

In a Yale study from 2002, people showed the same pattern across familiar everyday objects: they initially rated their understanding highly, but their confidence fell after they tried to write an explanation. Researchers call this the illusion of explanatory depth. Familiarity with using something can feel like understanding how it works, even when we cannot explain its mechanism.

The video’s central distinction is that feeling knowledgeable and being able to give a reasoned explanation are not the same. Familiarity and ease of processing can make something seem clear, but they do not provide justification.

Knowledge is not always something we can put into words

The host cautions that failing the zipper test does not mean a person knows nothing. Some knowledge is practical or tacit: people can ride a bike, work skillfully with materials, or notice that something is wrong without being able to explain exactly how they do it.

This distinction is especially important for knowing how versus knowing that. A person may know how to perform a skill without being able to describe it. The video focuses instead on claims and ideas—such as how an economy works or whether a policy will have a particular effect—where explanation is an important test of understanding.

Explanation, justification, and borrowed knowledge

Drawing on an ancient account of knowledge discussed in Plato’s Theaetetus, the host describes knowledge as true belief plus justification. Believing a correct answer by accident is not the same as knowing it; justification means being able to give reasons for the belief.

At the same time, most human knowledge is learned from other people rather than personally verified. We rely on teachers, books, scientists, and shared expertise. This dependence is normal and necessary. The risk is forgetting which ideas we understand ourselves and which we are relying on the wider network of knowledge to supply.

The video also describes research in which people explained how policies they felt confident about were supposed to work. Trying to explain the mechanisms reduced confidence and made participants’ views less extreme. The exercise encouraged greater humility about what they knew.

Search engines and AI can make external knowledge feel like our own

A 2015 study described in the video gave participants a question they could search online, such as why golf balls have dimples. Those who searched later rated themselves as more knowledgeable about unrelated subjects, too—even when their search had not produced the correct answer. The act of accessing information seemed to create a sense that the knowledge was in their own heads.

The host argues that AI chatbots can intensify this effect. A large language model produces fluent, organized answers, but reading an explanation is not the same as being able to construct or defend it. Because the answer sounds complete, it may feel like understanding even when the user has not worked through the reasoning.

A more recent study, as described in the video, compared people learning a practical topic such as gardening through a chatbot or ordinary web searches. Participants then wrote advice for a friend. The chatbot group’s advice was generally shorter, less specific, and less useful to readers—even when both groups saw the same facts and even when the chatbot supplied links.

The proposed explanation is that searching across sources introduces useful friction: readers compare information, notice discrepancies, and assemble a picture for themselves. If that work is skipped, the understanding may remain with the tool rather than becoming the user’s.

Why showing your work matters

The host recalls being required to show every step in school math. A correct final answer alone could have come from guessing, copying, or luck; the working showed whether the student understood the material. The video argues that this principle matters beyond school.

Now that AI can produce answers, presentations, and strategy documents quickly, the answer itself reveals less about the person who produced it. It is increasingly valuable to explain how a conclusion was reached: what evidence was considered, what alternatives were rejected, where uncertainty remains, and what might change one’s mind.

That process is closely related to critical thinking: weighing evidence, questioning assumptions, and noticing what is missing. The host argues that critical thinking develops through effortful practice—reading difficult material, comparing conflicting sources, and trying to explain ideas. Regularly outsourcing or skipping those activities may mean missing opportunities to build the skill.

Practical takeaway

The video does not argue that AI or search tools are inherently bad; it recognizes that they can help with research, drafting, and proofreading. Instead, it recommends using them without confusing access to an answer with personal understanding. A useful self-check is:

• Can I explain this idea in my own words without looking at a browser tab?
• Can I show the steps and evidence behind my conclusion?
• Do I know which parts I understand and which parts I am relying on others or tools to supply?
• What would make me revise my view?

A gap in one’s explanation is useful information: it shows what may need further study.

Speakers and sources featured

• Sonam Pelden — host and narrator.
• Plato and Socrates — discussed in connection with the Theaetetus and the idea of knowledge as true belief plus justification.
• Michael Polanyi — referenced in connection with tacit knowledge, or knowing more than one can express.
• Gilbert Ryle — referenced in connection with the distinction between knowing how and knowing that.
• Rozenblit and Keil (2002) — the Yale research on the illusion of explanatory depth.
• Fernbach, Rogers, Fox, and Sloman (2013) — research on explaining policy mechanisms and shifts in confidence or extremity.
• Fisher, Goddu, and Keil (2015) — research on online searching and the feeling of knowing.
• Melumad and Yun (2025) — the AI-chatbot-versus-web-search learning study described in the video.
• World Economic Forum employer survey — cited in connection with the demand for critical thinking.