The Illusion of Explanatory Depth

At a Glance

1. Quick Summary

Ask someone how well they understand a flush toilet and most will say "pretty well." Ask them to write down, step by step, what happens between the handle going down and the bowl refilling, and the confidence evaporates. That gap is the illusion of explanatory depth, named by Leonid Rozenblit and Frank Keil in 2002. It is strong for mechanisms and weak for facts or plots, because a mechanism lets us mistake familiarity with the surface for knowledge of the insides. The core finding has replicated in preregistered studies and has been extended to political policy, mental disorders, and conspiracy beliefs. The best-known extension, that explaining a policy makes political views more moderate, failed preregistered replication. The cure is also the diagnostic: try to explain it.

Simply About the Bias

This bias is the feeling that you know how something works, when really you only know what it does. You have seen a bicycle a thousand times, so your brain tells you that you understand it. But seeing a thing is not the same as knowing what goes on inside it.

Try this with a friend. Ask them to draw a bicycle from memory, with the frame, the pedals, and the chain in the right places. Most people are sure they can. Then the drawing comes out with the chain hooked to the front wheel, and they laugh, because they had no idea they didn't know. The same thing happens with how your phone finds Wi-Fi, how a vaccine works, or why the sky is blue. You feel like you know, right up until someone asks you to explain it.

Here is the trick. Before you say "I get it," explain it out loud as if you were teaching a younger kid, one step at a time. If you get stuck, that is not a failure. That is the bias showing you exactly where the gap is, so you know what to go and learn.

2. The Science Behind It

2.1. Discovery and History

The illusion of explanatory depth (IOED) was named in a 2002 paper in Cognitive Science by Leonid Rozenblit and Frank Keil at Yale. Keil had spent years studying "folk science," the rough causal theories people carry around about biology, physics, and machines. Those theories are sketchy, yet people never seem bothered by the gaps.

Rozenblit and Keil's procedure was simple. Participants rated how well they understood everyday devices on a seven-point scale, wrote out a full explanation of how one worked, re-rated, tried a diagnostic question, read an expert explanation, and rated again. Across twelve studies, ratings fell at every stage, with the biggest drop right after the attempted explanation. Nobody told them they were wrong. The act of explaining did the work. The same procedure applied to facts, procedures, and film plots produced much smaller drops or none, so the illusion is specific to mechanisms. Their explanation: mechanisms can usually be inspected in the world, so we rarely store them; knowing what a part is for feels like knowing how it works; we almost never explain things, so we get no feedback; and explanations have no natural stopping point.

Later work tested the mechanism (Adam Alter, Daniel Oppenheimer and Jeffrey Zemla's 2010 construal-level account), widened the domain to bicycles, political policies, mental disorders, and conspiracy beliefs, and checked whether the effect holds up. Robert Körner, Astrid Schütz and Lars-Eric Petersen reproduced the core effect in three preregistered studies in 2024, while Jarret Crawford and John Ruscio's preregistered attempts in 2021 did not reproduce the political moderation finding. Steven Sloman and Philip Fernbach's 2017 book The Knowledge Illusion reframed the whole thing: most of what any of us "knows" lives in other people, books, and tools.

Key milestones:

• 2002: Rozenblit and Keil name the effect and show it is specific to explanatory knowledge
• 2006: Rebecca Lawson's "The science of cycology" shows adults cannot draw a working bicycle
• 2013: Fernbach, Rogers, Fox and Sloman extend the effect to political policy
• 2021: Crawford and Ruscio report three preregistered failures to replicate the political moderation effect
• 2024: Körner, Schütz and Petersen replicate the core IOED in three preregistered studies (N = 607)

2.2. Key Researchers

2.3. Landmark Studies

"The misunderstood limits of folk science: An illusion of explanatory depth" (Rozenblit & Keil, 2002)

Yale undergraduates rated their understanding of items such as a zipper, a flush toilet, a cylinder lock, and a helicopter, then wrote a causal explanation for a subset and re-rated. Self-ratings dropped at each step, regardless of training or how familiar the items were. The same procedure produced little or no drop for facts, procedures, or narratives, the strongest evidence that the illusion is about causal explanation rather than general overconfidence.

"The science of cycology: Failures to understand how everyday objects work" (Lawson, 2006)

Lawson asked adults to complete a sketch of a bicycle by adding the missing frame parts, the pedals, and the chain. Many could not. A common error was a chain looped around the front wheel, which would make steering impossible. Errors were frequent even among regular cyclists and persisted when participants only had to recognize the correct drawing.

"Political extremism is supported by an illusion of understanding" (Fernbach, Rogers, Fox & Sloman, 2013)

Participants rated their understanding of policies such as cap-and-trade and single-payer health care, and how strongly they supported or opposed each. After writing a mechanistic account of how a policy would work, understanding ratings fell and so did the extremity of positions. A second experiment compared explaining a mechanism with listing reasons: only explaining reduced extremity. Crawford and Ruscio ran three preregistered close replications in 2021 and did not find that explaining increased moderation. Understanding ratings still dropped, so the core IOED held; the downstream political effect did not.

"Power and the illusion of explanatory depth" (Körner, Schütz & Petersen, 2024)

Three preregistered studies with 607 participants tested whether feeling powerful makes the illusion worse. The core effect replicated cleanly for devices, with a medium-to-large drop in self-rated understanding, and independent judges confirmed the written explanations were thin. Power itself had only a very small effect. Construal style was unrelated to the IOED, and a p-curve analysis of the 2010 studies found no evidential value for that link.

2.4. Neurological Basis

There is no established neuroscience of the illusion of explanatory depth specifically. No imaging study has isolated what happens in the brain when someone's "I understand this" rating collapses, and claims to the contrary should be treated as speculation. What exists is a broader literature on metacognition.

Judgments of knowing appear to rest on a sense of fluency rather than a check of stored content. When a familiar object comes to mind easily, the feeling of understanding arrives at once. Writing an explanation forces retrieval of the actual causal chain, and when retrieval stalls the feeling is revised downward. Work by Stephen Fleming, Raymond Dolan and others links accurate metacognitive monitoring to the anterior prefrontal cortex and shows that people vary in how well their confidence tracks their performance. That is a reasonable place to look for individual differences in the illusion, but the connection has not been tested directly.

3. Evolutionary Origins

Why would a mind carry explanations it does not have? The most plausible answer is that, for most of human history, it did not need to.

Sloman and Fernbach's framing is the cleanest. Humans reason about mechanisms, and we do it in groups. No individual in a hunter-gatherer band knows how to make every tool, track every animal, and treat every illness; that knowledge is spread across the community. Each person's job is to know enough to use the shared store and to recognize who knows the rest. A brain that stores a sketch, plus a sense of where the details can be found, is efficient.

The trouble is that "I could find out" feels, from the inside, like "I know." Because mechanisms are usually available for inspection, we treat environmental support as if it were internal knowledge. For a tool you can pick up and examine, that is a harmless shortcut. For a policy or a financial product, it is not.

A second strand connects the illusion to the "need to act fast" family. Confidence makes action possible. An ancestor who paused to verify a complete causal model of a river before crossing it would have been outcompeted by one who felt they understood rivers well enough. Our causal theories are coarse because coarse was always enough. What has changed is the environment: supply chains, vaccines, and monetary policy are far deeper than any person's working model of them.

4. How This Bias Manifests

4.1. In Everyday Life

• Confidently "fixing" an appliance, then discovering mid-job that you cannot picture how the parts fit together
• Explaining to a child how rain forms, and running out of mechanism after two sentences
• Closing a search tab with a stronger sense of knowledge than you retained; Fisher, Goddu and Keil (2015) found that recent online searching inflated people's estimates of their ability to explain unrelated topics

4.2. In the Workplace

• Approving a technical plan because its summary slide makes sense, without being able to describe how the system would produce the result
• Managers who believe they understand a process they supervise until a new hire asks "and then what happens?"
• Meetings that reach agreement because everyone shares a confident vocabulary and nobody has tested whether it points to a shared model

4.3. In Business and Marketing

• Investors backing a product on the strength of a demo and a mechanism they cannot restate in their own words
• Strategy decks built around causal claims ("engagement drives retention drives revenue") that nobody has worked through as a chain
• Buying analytics or AI products on a feeling of understanding; Chromik and colleagues (2021) found the same illusion when people are shown explanations of machine-learning models

4.4. In Politics and Media

People rate their understanding of policies highly, and those ratings fall sharply after they try to explain how the policy would work. Whether the drop also moderates attitudes, as Fernbach and colleagues reported in 2013, did not replicate in Crawford and Ruscio's 2021 studies.

Typical manifestations:

• Holding strong positions on policies whose mechanism one cannot describe past the first step
• Debating by trading reasons (which stabilize positions) rather than explanations (which test them)
• Vitriol and Marsh (2018) found that people with a larger illusion of understanding about political topics were more likely to endorse conspiracy theories

4.5. In Healthcare

Zeveney and Marsh (2016) applied the explanation task to mental disorders and found the same pattern: people believe they understand conditions like depression until asked to explain how they arise.

• Patients who feel they understand their medication's mechanism and adjust doses on that basis
• Clinicians who have not explained a mechanism in years and discover, when a student asks, that their model is thinner than they thought
• Health decisions driven by a vivid, coherent, and wrong internal picture (the body "detoxing")

4.6. In Finance and Investing

• Owning products (structured notes, leveraged ETFs, options) whose payoff mechanism one could not draw on paper
• Treating a financial story ("rates go down, so stocks go up") as a mechanism rather than a correlation
• The 2008 pattern: layers of participants who each assumed someone else fully understood the instrument

5. Real-World Case Studies

Case Study 1: The Toaster Project

• Context: In 2009, British designer Thomas Thwaites set out to build an ordinary electric toaster from scratch: mining iron ore, extracting copper, making plastic. A new toaster cost under four pounds.
• What happened: The project took nine months and cost over a thousand pounds. A basic toaster turned out to contain around four hundred parts made from dozens of materials, each requiring industrial knowledge no single person holds. The finished object worked for a few seconds before melting.
• The bias at work: Before starting, Thwaites and everyone he told had the ordinary feeling that a toaster is understood: a wire gets hot, the bread browns. That feeling was a sketch of the function, not a model of the mechanism or its supply chain.
• Consequences: None beyond a book and a TED talk, which is the point. It was a low-stakes demonstration of how much hidden depth sits inside the most familiar objects.
• Lessons learned: Familiarity is not understanding. The toaster is cheap precisely because no one has to understand all of it.

Case Study 2: Structured Credit Before 2008

• Context: Before the financial crisis, banks packaged mortgages into securities, repackaged the lower tranches into collateralized debt obligations, and in some cases repackaged those again. The instruments were sold worldwide, often with top credit ratings.
• What happened: The U.S. Financial Crisis Inquiry Commission's 2011 report documented that many buyers and sellers relied on ratings and on the general story ("diversified pools of mortgages are safe") rather than on a working model of how the cash flows would behave if house prices fell nationally. When they did, the products failed together.
• The bias at work: Participants at every level reported understanding these instruments. The vocabulary was shared, the summaries were coherent, and almost no one had traced the mechanism end to end. The feeling of understanding was supported by the environment (ratings, models, peers) rather than by internal knowledge.
• Consequences: Trillions in losses, bank failures, and a global recession. The illusion was not the only cause; fraud, incentives, and leverage all played their part.
• Lessons learned: For any product you cannot explain tranche by tranche, your understanding rating should be low, whatever your colleagues' ratings are. "Everyone understands this" describes a shared feeling, not shared knowledge.

Historical Example: "I, Pencil" (1958)

In 1958 the American economist Leonard Read published a short essay written in the voice of a pencil. Its claim was that not a single person on earth knows how to make one. The cedar is logged in one place and milled in another; the graphite is mined, mixed with clay, and baked; the lacquer, the ferrule, the eraser, and the glue each come from separate industries, and the people at each stage know only their stage.

Read wrote the essay to make a point about markets, but it has become a standard illustration of the illusion of explanatory depth. A pencil is about as simple as a manufactured object gets. Everyone feels they understand it. Almost nobody could explain, step by step, how it comes to exist. Sloman and Fernbach draw on it in The Knowledge Illusion: the pencil works because the knowledge is in the community.

6. The Cost of This Bias

6.1. Personal Costs

• Failed repairs and projects started on the strength of a mechanism never actually understood
• Health decisions based on a confident but wrong model of how a treatment works
• Financial commitments to products whose internals were never traced
• Lost learning, because the feeling of understanding stops the search before the gap is found

6.2. Professional Costs

• Technical decisions approved on summaries that nobody in the room could expand
• Expertise that quietly decays because it is never exercised through explanation
• Poor estimates, because the steps that were never traced are the ones that take longest
• Reputational damage when a client or auditor asks the question that reveals the gap

6.3. Societal Costs

• Policy debates conducted at the level of labels rather than mechanisms
• Financial crises amplified by layers of participants who all assumed someone else understood the product
• Susceptibility to conspiracy theories, which Vitriol and Marsh (2018) found to track the size of the illusion about political topics
• Erosion of trust in experts, because the feeling of understanding makes expert caveats look like evasion

6.4. Statistical Impact

• In Rozenblit and Keil's original studies, self-rated understanding of devices on a seven-point scale fell at every stage, with the largest single drop immediately after participants tried to write an explanation
• Körner, Schütz and Petersen's 2024 meta-analysis across their three preregistered studies (N = 607) put the overall pre-to-post explanation drop at about r = .36, a medium-to-large effect
• In Lawson's 2006 bicycle study, a large share of adults, including regular cyclists, produced drawings that could not function as bicycles
• Crawford and Ruscio's 2021 preregistered replications reproduced the drop in understanding ratings from Fernbach and colleagues' 2013 studies but not the reduction in attitude extremity

7. The Hidden Benefits

The illusion of explanatory depth is not simply a defect. It is the felt side of an arrangement that works surprisingly well most of the time.

• It lets us use what we don't understand: Nobody could drive, cook, vote, or take medicine if they first had to build a complete causal model of each. A rough sketch plus confidence is what makes a complex world usable.

• It supports division of cognitive labor: Sloman and Fernbach treat the illusion as the price of specialization. Each person stores a sketch and a pointer to who knows more; the group as a whole knows the mechanism.

• It gets people started: Beginners who felt the full depth of what they didn't know might never open the hood. A little unearned confidence gets people to where they can discover the gaps productively.

• It is easy to puncture: Unlike many biases, this one has a reliable, cheap correction. A few minutes of attempted explanation reveals the gap.

The benefits hold when the stakes are low or the pointers to real expertise are reliable. They fail when the system is deep, the decision is consequential, and nobody in the chain has traced the mechanism.

8. Self-Assessment: Do You Have This Bias?

8.1. Warning Signs Checklist

• You would rate your understanding of common devices (a zipper, a toilet) as high, but you have never actually explained one
• When someone asks "but how does that actually work?", you usually end up restating what it does
• You have strong views on at least one policy you could not describe mechanically past the first step
• After reading a technical article, you feel you understand it, but could not reconstruct the argument without it
• You rarely explain things out loud or in writing; most of your "knowing" is silent recognition
• You assume familiar terms (inflation, encryption, antibodies) come with a working model attached
• You have been surprised to find a gap in your understanding of something you use every day
• You find expert explanations slower and more hedged than necessary
• You close a search tab feeling you now understand the topic, without having retained the steps
• You have never tried to draw a system (a bicycle, your team's workflow) from memory

Scoring:

• 0-2 checked: Low susceptibility
• 3-5 checked: Moderate susceptibility
• 6-8 checked: High susceptibility
• 9-10 checked: Very high susceptibility

8.2. Self-Reflection Questions

1. Pick a device you use daily. Without looking anything up, write four sentences on how it works. How far did you get before the sentences became vague?
2. What political or economic position do you hold firmly? Can you describe the chain of mechanisms by which the policy would produce its intended effect?
3. When did you last explain something technical to another person and discover a gap?
4. Which of your professional judgments rely on a mechanism you have never traced end to end?
5. Think of a time you felt you understood something after a quick search. What do you still remember of how it works?

8.3. Quick Diagnostic Scenario

Scenario: A friend is deciding whether to have a medical procedure and asks you, because you have "read a lot about it," how it actually works. You have read several articles and feel you understand it well.

How would you respond?

• A) Give a confident summary of what the procedure does and why it is a good idea → High susceptibility
• B) Start explaining, notice you are describing effects rather than mechanism, and finish anyway with "something like that" → Moderate susceptibility
• C) Walk through it step by step, say plainly where your understanding stops, and point your friend to someone who can fill the gap → Low susceptibility

9. Identifying This Bias in Others

9.1. Behavioral Indicators

• Describes systems fluently in terms of what they do, and stalls when asked how
• Uses correct terminology without being able to unpack the terms
• Offers reasons for positions readily, but mechanisms reluctantly
• Approves plans on the strength of summaries and never asks to see the next level down
• Has never been heard to say "I don't actually know how that part works"

9.2. Conversational Red Flags

Phrases people say when under this bias:

• "It's basically just..."
• "Everyone knows how that works."
• "I don't need to get into the details; the point is..."
• "The way it works is, it does X." (restating the function as if it were the mechanism)

Types of arguments they make:

• Chains of labels standing in for chains of causes ("the stimulus boosts demand, which creates jobs")
• Appeals to familiarity ("I use this every day")

Questions they avoid asking:

• "And what happens at the next step?"
• "Could I draw this?"

9.3. Situational Triggers

• Objects or systems that are seen constantly but never opened
• Technical vocabulary that has become familiar through news and conversation
• Having just read or searched for information on the topic
• Being told that experts understand something; Sloman and Rabb (2016) found that simply learning that scientists understand a phenomenon raises people's ratings of their own understanding of it
• Time pressure, which discourages the slow work of explaining

10. Cognitive Debiasing Strategies

10.1. Immediate Techniques

• The Explanation Test: Before claiming to understand something, write or say how it works, step by step, from input to output. Rate your understanding again afterward. This is the original experimental procedure and it works as a self-check
• The "How, Not Why" Switch: When you catch yourself listing reasons for a position, switch to mechanism: what it does, then what happens, then what
• Draw It: For any physical or process system, sketch it from memory. Lawson's bicycle study shows how fast this exposes gaps
• Name the Stopping Point: Say where your understanding ends: "I know the valve opens; I don't know what makes it close."

10.2. Long-Term Strategies

• Explain Regularly: Teach, write, or talk through mechanisms as a habit. The illusion survives partly because explanation is rare
• Keep a Ratings Log: Record your confidence before and after attempting an explanation. Over months, the typical drop tells you how much to discount your first impression
• Separate Recognition From Production: Test yourself by producing the explanation, not by rereading until it looks familiar
• Build Good Pointers: You cannot understand everything. Invest in knowing who actually does, and check that your "expert" has traced the mechanism

10.3. Environmental Design

• Mechanism Slides: In reviews and approvals, require a slide or paragraph that describes how the proposal works, as well as what it achieves
• "Explain It Back" Norms: Before a decision, have someone other than the proposer explain the mechanism in their own words. If nobody can, the group's understanding is a shared feeling
• Whiteboard Culture: Keep drawing surfaces where decisions are made, and use them. Drawings expose what slides hide
• Reward the Question: Make "how does that part work?" a normal, low-cost thing to ask, and "I don't know yet" a normal answer

10.4. When to Seek External Input

• When the decision depends on a mechanism you cannot trace end to end yourself
• When your confidence comes mainly from familiarity, vocabulary, or a recent search
• When everyone around you is equally confident and equally untested
• When a system is layered (financial products, supply chains, biological pathways) and you can see only your layer

Whom to ask: Find someone who has traced the mechanism, not someone who shares your confidence. An expert's explanation should get more detailed under questioning, not more abstract.

11. Practical Exercises

Exercise 1: The Rozenblit-Keil Self-Test

• Objective: Experience the illusion directly and calibrate your sense of understanding
• Time required: 20 minutes
• Materials needed: Paper and pen; a list of six everyday devices (zipper, flush toilet, cylinder lock, refrigerator, ballpoint pen, speedometer)
• Difficulty level: Beginner
• Instructions:
  1. Rate your understanding of each device from 1 (no idea) to 7 (complete understanding)
  2. Pick the two you rated highest and write a full step-by-step explanation of how each works
  3. Re-rate both
  4. Read a short expert explanation of each, rate a third time, and compare
• Reflection questions:
  • How large was the drop after your explanation attempt?
  • Which step was the first to go vague?
• Frequency: Once to start; repeat with new items whenever your confidence in your own technical understanding feels high

Exercise 2: Mechanism, Not Reasons

• Objective: Separate the feeling of understanding a position from an actual model of the policy behind it
• Time required: 30 minutes
• Materials needed: A notebook; three policies or decisions you hold strong views on
• Difficulty level: Intermediate
• Instructions:
  1. For each, rate how well you understand how it would work, 1 to 7, and how strongly you hold your position
  2. For the first one, write the reasons you hold your position
  3. Now write the mechanism: what the policy does first, what that causes, what that causes next. Stop when you cannot describe the next step
  4. Re-rate your understanding and your position, then repeat for the other two
• Reflection questions:
  • Did writing reasons and writing mechanism feel different? Which exposed gaps?
  • If your position did not change despite a large drop in understanding, what is it resting on?
• Frequency: Monthly, or before any debate, vote, or decision that matters

Exercise 3: Draw Your System

• Objective: Find the hidden depth in a system you depend on every day
• Time required: 45 minutes
• Materials needed: A large sheet of paper, pens, and access afterward to someone who knows the system
• Difficulty level: Intermediate
• Instructions:
  1. Choose a system: your team's workflow from request to delivery, your home's heating, or how a customer's order becomes a shipment
  2. Draw it from memory, with every component and every arrow labeled
  3. Mark in a different color every point where you are guessing
  4. Show the drawing to someone who knows the system and ask them to correct it
  5. Count the corrections: missing parts, wrong connections, or steps you had collapsed into one
• Reflection questions:
  • Were the gaps in the parts you use most or least?
  • What recent decision depended on one of the parts you got wrong?
• Frequency: Quarterly, with a different system each time

Daily Practice

One Explanation a Day

Once a day, take something you encountered and felt you understood (a news item, a tool at work, an everyday object) and explain how it works to yourself in two or three sentences.

• Suggested duration: 3 minutes
• Best time of day: Evening, reviewing the day
• How to track progress: Keep a running list of the explanations that stalled, and tick them off as you fill the gaps

Weekly Challenge

Teach It Back

Each week, pick one mechanism you rely on professionally and explain it to someone who does not share your background, inviting them to ask "and then what?" until you reach a step you cannot describe.

• Expected outcomes after 4 weeks: A clearer map of where your professional knowledge is real and where it is vocabulary; a lower, more accurate baseline for your confidence ratings
• Journaling prompts for reflection:
  • At which "and then what?" did I stall?
  • Was the gap something I once knew and lost, or something I never had?

12. For Specific Audiences

For Leaders and Managers

• Approval Discipline: Do not approve what you cannot explain back. Ask the proposer to walk you through the mechanism, and ask someone who is not the proposer to do the same
• Beware Shared Vocabulary: A team that uses the same confident terms may still have no shared model. Have people draw the process independently and compare
• Protect the "How" Question: Junior staff often see the gap first and say nothing. Make asking how something works an expected part of every review
• Delegate Understanding Explicitly: Decide who owns the mechanism of each critical system, and verify that they can explain it rather than describe it

For Parents and Educators

• Age-Appropriate Explanation: "Sometimes we think we know how something works just because we've seen it a lot. The way to find out is to try explaining it. If you get stuck, that's great, because now you know what to learn next."
• Ask "And Then What?": When a child explains something, ask for the next step rather than correcting. The child discovers the gap and owns the question
• Draw Before Reading: Have students sketch how they think a system works before the lesson, then compare with the real thing after
• Model It Yourself: Say "I thought I understood this, and I just realized I don't" in front of children. It teaches that the gap is normal and findable

For Healthcare Professionals

• Patient Understanding Checks: Patients often rate their understanding of a condition highly. Ask them to explain it back in their own words; Zeveney and Marsh's work on mental disorders shows how large the gap can be
• Your Own Mechanisms: Clinical models decay when unused. Periodically explain the mechanism of a common treatment to a student and notice where you reach for a label instead of a step
• Health Messaging: Messages that give people a mechanism they can repeat (how a vaccine trains the immune system) compete with the simple wrong story better than messages that rely on authority

For Financial Professionals

• The Tranche Test: For any product you sell or hold, be able to describe its cash flows and failure modes layer by layer. If you cannot, your understanding rating should be low regardless of ratings or peer confidence
• Client Explanations: Ask clients to explain a product's mechanism before they buy, and treat a confident restatement of the pitch as a warning sign
• Narrative Versus Mechanism: Business stories ("growth drives margin") feel like understanding. Translate each into a chain of concrete steps before relying on it

13. Interactions with Other Biases

Biases That Amplify This One

Biases That Counteract This One

Common Bias Chains

Mere Exposure Effect → Illusion of Explanatory Depth → Overconfidence Effect → Escalation of Commitment

Daily contact with a system makes it familiar (Mere Exposure Effect). Familiarity is read as understanding (Illusion of Explanatory Depth). The felt understanding supports confident decisions (Overconfidence Effect). When the decision runs into the depth that was never modeled, the decision-maker doubles down rather than revisiting the model (Escalation of Commitment).

Interrupting the chain: The cheapest break is at the second link. A five-minute attempt to explain the system before the decision converts felt understanding into a measured one. Once a confident decision has been made in public, the later links are much harder to undo.

14. Cultural Perspectives

Cross-cultural research on the illusion of explanatory depth is thin, and anyone claiming a clean cultural pattern is ahead of the evidence. The original studies used American undergraduates. Körner, Schütz and Petersen's 2024 replication used German-speaking samples. Carlos Gaviria and Javier Corredor (2021) found the effect for historical knowledge in Colombia and noted that social desirability shaped how willing people were to admit gaps. The mechanism, confusing familiarity with internal knowledge, has no obvious reason to respect borders.

Where culture plausibly matters is in the surrounding behavior: how freely people rate themselves highly in the first place, how comfortable they are saying "I don't know," and how strongly deference to experts is built in. Cultures that prize confident speech may show larger initial ratings and therefore larger drops; cultures where self-effacement is the norm may show smaller ones. None of these are established findings for this bias specifically.

15. Myths and Misconceptions

16. Expert Insights

"People feel they understand complex phenomena with far greater precision, coherence, and depth than they really do." — Leonid Rozenblit and Frank Keil, "The misunderstood limits of folk science" (2002)

"Political extremism is supported by an illusion of understanding." — Philip Fernbach, Todd Rogers, Craig Fox and Steven Sloman, title of their 2013 paper in Psychological Science

Paraphrasing Steven Sloman and Philip Fernbach in The Knowledge Illusion (2017): we each know far less than we think, and we get away with it because knowledge lives in a community. The illusion is what it feels like to draw on that shared store without noticing where our own heads end.

Paraphrasing Rebecca Lawson on the bicycle study (2006): an object seen every day and used by many of the participants still could not be drawn in working form by a large share of them, which suggests that frequent exposure leaves people with a remarkably sparse model of how a thing operates.

17. Key Takeaways

1. Explaining is the test: The illusion is the gap between how well you feel you understand a mechanism and how far you get when you try to explain it step by step. The test is also the cure.

2. It is specific to mechanisms: Rozenblit and Keil showed the drop for devices and causal systems, and little for facts, procedures, or stories. Familiarity with what a thing does is mistaken for knowledge of how.

3. The core effect is solid: It has replicated for over two decades, including in Körner, Schütz and Petersen's three preregistered studies in 2024 with 607 participants.

4. The political extension is unsettled: Explaining a policy reliably lowers confidence in one's understanding of it. Whether that moderates attitudes, as Fernbach and colleagues reported in 2013, did not reproduce in Crawford and Ruscio's 2021 preregistered replications.

5. Reasons are not explanations: Listing why you hold a position recruits arguments and leaves the position intact. Describing how the thing would work exposes the gaps.

6. The environment feeds it: Familiarity, shared vocabulary, hearing that experts understand, and recent internet searches all raise the feeling of understanding without raising understanding.

7. It is a community feature with individual costs: We never think alone, and the illusion is the price of relying on shared knowledge. The fix is to know where your own model stops and to build reliable pointers to people who have traced the rest.

18. Further Resources

Academic Papers

• Rozenblit, L., & Keil, F. (2002). The misunderstood limits of folk science: An illusion of explanatory depth. Cognitive Science, 26(5), 521-562.
• Fernbach, P. M., Rogers, T., Fox, C. R., & Sloman, S. A. (2013). Political extremism is supported by an illusion of understanding. Psychological Science, 24(6), 939-946.
• Crawford, J. T., & Ruscio, J. (2021). Asking people to explain complex policies does not increase political moderation: Three preregistered failures to closely replicate Fernbach, Rogers, Fox, and Sloman's (2013) findings. Psychological Science, 32(4), 611-621.
• Körner, R., Schütz, A., & Petersen, L.-E. (2024). "It doesn't matter if you are in charge of the trees, you always miss the trees for the forest": Power and the illusion of explanatory depth. PLOS ONE, 19(4), e0297850.
• Lawson, R. (2006). The science of cycology: Failures to understand how everyday objects work. Memory & Cognition, 34(8), 1667-1675.
• Fisher, M., Goddu, M. K., & Keil, F. C. (2015). Searching for explanations: How the Internet inflates estimates of internal knowledge. Journal of Experimental Psychology: General, 144(3), 674-687.
• Vitriol, J. A., & Marsh, J. K. (2018). The illusion of explanatory depth and endorsement of conspiracy beliefs. European Journal of Social Psychology, 48(7), 955-969.

Books

• Sloman, S., & Fernbach, P. (2017). The Knowledge Illusion: Why We Never Think Alone. Riverhead Books.
• Thwaites, T. (2011). The Toaster Project: Or a Heroic Attempt to Build a Simple Electric Appliance from Scratch. Princeton Architectural Press.

Book Chapters

• Keil, F. C., Rozenblit, L., & Mills, C. M. (2004). What lies beneath? Understanding the limits of understanding. In D. T. Levin (Ed.), Thinking and Seeing: Visual Metacognition in Adults and Children (pp. 227-249). MIT Press.
• Keil, F. C., & Wilson, R. A. (2000). Explaining explanation. In F. C. Keil & R. A. Wilson (Eds.), Explanation and Cognition (pp. 1-18). MIT Press.

19. Summary Card

20. Glossary of Terms Used

21. Discussion Questions

For book clubs, classrooms, or self-reflection:

1. Pick an everyday object in the room. Everyone rates their understanding of it, then one person tries to explain it. How far does the group get, and whose rating changes most?

2. Fernbach and colleagues found that asking for reasons and asking for explanations produce different effects. Why might listing reasons for a position leave it untouched while explaining the mechanism shakes it?

3. Sloman and Fernbach argue that the illusion is the natural price of a community of knowledge. If that is right, is the goal to reduce the illusion, or to improve our pointers to the people who actually know?

4. The political moderation finding did not replicate, but the drop in understanding did. What would you need to see to be convinced either way about whether explanation changes attitudes?

5. Search engines and AI assistants make explanations instantly available. Fisher, Goddu and Keil found that searching inflates our sense of what we know. How should students, professionals, and voters use these tools without losing track of where their own understanding stops?

সর্বশেষ আপডেট: 15 অক্টোবর 2026