Reflective Listening 13 min read

Double Descent: Keep Asking How Do You Know

J

Jared Clark

April 25, 2026

Most conversations that fall apart do so before anyone notices they've started falling. Someone says something confident — "People like that don't change," or "That policy never works," or "My side is just being reasonable" — and the other person either agrees or argues. Both responses skip the most important question: how do you know that?

What I've come to call Double Descent is a practice built around not skipping that question. It's the discipline of asking "how do you know?" and then, when you get an answer, asking it again — and again, if necessary — until you hit bedrock. Until you find the thing underneath the claim that the person can't actually defend with evidence, only with experience, intuition, or inherited assumption. That's bedrock. And bedrock is exactly where real conversations have to start.


What "Bedrock" Actually Means

I want to be precise about this word, because I don't mean it as a criticism. Bedrock isn't a failure. Everyone has it. Bedrock is the layer of belief that doesn't rest on anything more fundamental — the place where the chain of justification stops and something more basic takes over. It might be a formative experience. It might be a moral intuition so deep the person never thought to examine it. It might be a frame inherited from a community, a religion, a political tradition.

When you do Double Descent well, you're not trying to destroy someone's bedrock. You're trying to find it, because once you find it, you're finally talking about the real thing. Everything above bedrock is superstructure — arguments, statistics, policy positions, interpretations. The superstructure is what people argue about. The bedrock is what people actually disagree about. And in my experience, two people who trace their disagreement all the way down to their actual bedrock often discover that what they really disagree about is surprisingly small, or surprisingly understandable.


Why One "How Do You Know?" Is Never Enough

Here's the pattern I see constantly: someone asks a good probing question, gets a slightly-more-specific claim in response, and treats that as the answer. It isn't. The first response to "how do you know?" almost always produces what I'd call a surface justification — the most socially acceptable, pre-packaged version of the belief. It's what the person says when they want to seem reasonable.

Research on epistemic cognition suggests that most people have two versions of their beliefs: the version they can defend publicly and the version that actually drives their behavior. A 2021 study published in Cognition found that people who were asked to explain the causal mechanisms behind their policy preferences — not just state their preferences, but explain how they work — significantly moderated their confidence in those preferences. The first question opened the door. The second question, the "but how exactly does that mechanism work?" question, is where overconfidence started to crack.

That's the logic behind Double Descent. The first "how do you know?" gets you out of the claim. The second one gets you closer to the ground. Sometimes you need a third. Occasionally you need a fourth. But in my experience, you rarely need more than three before you're standing on something that neither party can fully justify with logic — which means you've both arrived somewhere honest.


How to Actually Do It

The mechanics are simpler than they sound, but the practice is harder than it looks.

Start with the claim, not the person. When someone makes a confident assertion, ask "how do you know that?" without any accusatory tone. The goal is genuine curiosity. If you're performing curiosity while secretly hoping to trap them, the conversation will die immediately. People are much better at detecting inauthenticity than they think they are.

Receive the first answer fully before asking again. This is where most people fail. They hear the first response and immediately launch the second probe. That feels like interrogation. Instead, reflect back what you heard — "So what you're saying is that you've seen this happen in X context, and that's what's driving the belief?" — and let them confirm or correct. Then ask again.

The second descent is gentler, not harder. "And how did you come to believe that?" or "Where does that understanding come from for you?" are softer than the first question but go deeper. You're not looking for a logical syllogism. You're looking for a story, an experience, a community, a moment. That's where bedrock lives.

Name it when you hit it. When the person says something like "I just know it's true" or "I've always believed that" or "that's just what I've seen my whole life" — that's bedrock. Don't push further. Say something like: "That makes sense. It sounds like that comes from a really deep place for you." Now you can have a real conversation.


What This Looks Like in a Real Disagreement

Let me give you a concrete scenario because abstract method descriptions are easy to lose.

Two people are disagreeing about whether a certain neighborhood's crime problem is primarily a policing issue or a poverty issue. Person A says: "More police presence is clearly the answer."

A Double Descent practitioner doesn't argue. They ask: "How do you know more police presence would help?"

Person A says: "Because we've seen crime go down in places where police presence increased."

The practitioner reflects and then asks again: "And how did you come to trust that particular interpretation of those cases?"

Person A pauses. Then says something like: "Honestly, I grew up in a neighborhood where things got bad when the police pulled back. I saw it firsthand."

That's bedrock. Not a policy paper. A childhood memory. A thing that happened to a real person in a real place. Now the conversation has something honest to work with.

The other person might have their own bedrock — maybe a memory of aggressive policing that felt like occupation rather than protection. When both bedrock layers are on the table, the conversation changes completely. You're not arguing about policy abstractions anymore. You're talking about two real human experiences that produced two different frameworks for understanding safety.


The Comparison That Matters: Debate vs. Double Descent

One of the reasons I find this practice so compelling is how sharply it contrasts with the way most public disagreement is structured.

Feature Debate Mode Double Descent
Goal Win the exchange Understand the ground
First move Counter the claim Ask about the claim
Listens for Weaknesses to exploit The story behind the belief
Treats strong claims as Targets Invitations
Endpoint One side is right Both sides are honest
What you learn Their argument Their bedrock
Effect on relationship Usually damages it Usually deepens it
Works best when Stakes are low Stakes are high

Debate mode treats the claim as the problem. Double Descent treats the claim as a door. There's something behind it worth knowing — and once you know it, you can stop pretending the argument is about logic when it's actually about history, identity, and experience.


Why People Resist Going Deeper

I should be honest: not everyone wants to do this. Some people are not avoiding their bedrock by accident — they're avoiding it on purpose, because on some level they suspect their bedrock won't survive scrutiny. That's not a moral failure. It's a deeply human protective instinct.

According to Jonathan Haidt's research on moral psychology, people form moral intuitions first and construct post-hoc reasoning second. The reasoning isn't the real source of the belief; it's the defense attorney hired afterward. When you ask someone "how do you know?" twice, you're effectively asking them to notice that their defense attorney might be working for a client they've never actually met.

That's uncomfortable. It should be. The discomfort is the sign that you've gotten somewhere worth being.

What I'd encourage is not to force the descent. If someone gives you their surface justification and clams up at the second question, you can note that you'd love to understand it better someday and leave the door open. The method isn't a crowbar. It's an invitation. Some invitations get accepted later, not immediately.


Double Descent and the Practice of Reflective Listening

Double Descent lives inside a broader practice of reflective listening — the commitment to understand what someone means before you respond to what they said. Those aren't the same thing. What someone says is the claim. What they mean is the whole iceberg beneath it.

When you combine reflective listening with the Double Descent method, the sequence looks like this: you hear the claim, you reflect it back to confirm you understood it correctly, you ask the first descent question, you reflect the answer back, you ask the second descent question, you reflect that answer back, and then — only then — you speak from your own perspective.

Most people skip all the reflection and jump straight to their response. When that happens, no one actually talks to anyone. They just take turns performing their pre-existing positions at each other.

Reflective listening, combined with Double Descent, is the structural difference between a real conversation and two people standing in the same room not talking to each other. If you're curious about the broader practice of reflective listening at WeaveCulture, the work we do on civil dialogue starts from exactly this premise.


The Hardest Application: Descending on Yourself

Here's the thing I don't want to skip over: the most important application of Double Descent isn't on other people. It's on yourself.

Try it. Pick one of your confident beliefs — something you'd assert without much hesitation in a conversation. Then ask yourself: how do I know that? Write down the answer. Then ask again: how do I know that? Follow the thread. Give it two or three rounds of genuine questioning.

What you'll usually find is that your bedrock is some combination of things you experienced, things you were taught, and things you chose to believe because they made sense of something important in your life. None of that makes the belief wrong. But it does make it something other than pure reason — and that distinction matters enormously when you're trying to talk to someone whose bedrock is different from yours.

A 2019 Pew Research study found that 73% of Americans believe their political views are based primarily on facts and evidence, while attributing the other side's views primarily to bias and misinformation. If Double Descent does nothing else, it should at least complicate that picture. Because when you descend on yourself with the same rigor you'd apply to someone you disagree with, you tend to find bedrock where you expected to find bedrock — just your own version of it.


What Changes When You Hit Bedrock Together

There's a moment in a Double Descent conversation that I find genuinely moving, when it happens. It's when both people realize that they've been arguing about the tip of the iceberg while sharing, often, a much more similar depth below the waterline.

Two people arguing about immigration policy might both have bedrock beliefs about family, belonging, and safety that are nearly identical. Two people arguing about criminal justice might both have bedrock experiences of feeling unprotected — just by different things. The arguments look like opposites. The bedrock, surprisingly often, shares the same shape.

That doesn't mean the policy disagreement disappears. It doesn't. But it changes what the disagreement is about. Instead of "you're wrong and I'm right," it becomes "we've had different experiences that led us to different conclusions, and maybe we can work with that." That's a conversation worth having.


Three Citation-Ready Observations on This Practice

One: Double Descent is not a rhetorical technique — it is an epistemic practice. The goal is not to win an argument or expose a flaw. The goal is to find the actual foundation of a belief so that disagreement can become about something real rather than something constructed.

Two: Most disagreements are not about the claims people make — they are about the experiences and intuitions that produced those claims. The first "how do you know?" surfaces the argument. The second one surfaces the person.

Three: Reflective listening without descent produces sympathy but not understanding. You can mirror someone back perfectly and still have no idea what's actually underneath what they're saying. The descent is what gets you there.


A Note on Where This Fits in a Culture of Civil Dialogue

WeaveCulture is built around the premise that civil dialogue is not primarily a skill — it's a set of practices, and practices require structure. Double Descent is one of the most important structures I know for moving a conversation from performance to genuine exchange.

Gallup's research on American polarization found that 77% of Americans believe the country's divisions are at a crisis level, but 84% also say they personally want to do something about it. The gap between wanting civil dialogue and knowing how to do it is enormous. Double Descent is one way to begin closing that gap — not by teaching people to be nicer, but by giving them a method for getting to something true.

You can explore more about how WeaveCulture approaches the practice of civil dialogue and genuine belonging at weaveculture.org.


FAQ: Double Descent and the "How Do You Know?" Practice

Q: What is Double Descent in the context of civil dialogue? A: Double Descent is a reflective listening practice in which you ask "how do you know?" at least twice in a conversation — once to surface the reasoning behind a claim, and again to find the deeper experience, intuition, or assumption that actually drives the belief. The goal is to reach the "bedrock" level of a person's conviction, where real understanding becomes possible.

Q: How many times should you ask "how do you know?" A: In most conversations, two to three descents are enough to reach bedrock — the layer of belief that can't be justified with further logic, only with lived experience or inherited assumption. Beyond three, the method can feel like interrogation. The signal that you've arrived is when the person says something like "I just know" or grounds their belief in a personal story rather than an argument.

Q: Is this method manipulative or confrontational? A: Only if it's done with a hidden agenda. Double Descent works — and builds trust — when the curiosity is genuine. The goal is never to expose or trap; it's to understand. When the practitioner is authentically interested in the other person's bedrock, the conversation tends to deepen rather than escalate.

Q: Can you use Double Descent on yourself? A: Yes, and in my view this is actually the most important application. Applying the "how do you know?" sequence to your own confident beliefs tends to reveal that your own bedrock is more experiential and inherited than purely rational — which is valuable information before you try to use the method on someone else.

Q: What's the difference between Double Descent and the Socratic method? A: The Socratic method is designed to expose logical contradictions and lead the interlocutor toward a pre-determined truth. Double Descent is not designed to lead anywhere predetermined. The goal is to surface the actual foundation of a belief — not to dismantle it, but to understand it. The posture is fundamentally different: curiosity rather than refutation.


Last updated: 2026-04-25

Jared Clark is the founder of WeaveCulture, a platform dedicated to building communities that practice civil dialogue, reflective listening, and genuine belonging.

J

Jared Clark

Founder, WeaveCulture

Jared Clark is the founder of WeaveCulture, a platform dedicated to building communities that practice civil dialogue, reflective listening, and genuine belonging.