THE MACHINERY OF LOGIC
A Complete Guide to Rational Decoration
How the Brain Uses Reason to Protect What It Already Believes
What follows is not advice.
It is not a guide to thinking more clearly. Not a logic textbook. Not a system for winning arguments or spotting fallacies at dinner parties.
It is mechanism.
The actual machinery running beneath the experience of reasoning. The circuits that arrange evidence to support conclusions already reached. The architecture that deploys logic as a weapon in conversation while disguising it as a search for truth. The deep structure that makes what people say follow one set of rules and what people do follow another.
Most people believe they reason their way to their positions. That they observe facts, weigh evidence, apply logic, and arrive at conclusions.
The sequence is reversed.
The conclusion comes first. The reasoning follows. The logic you experience is not the engine driving toward truth. It is the press secretary explaining a decision already made by someone who does not speak to the public.
This document is that press secretary, exposed.
Nothing more.
What you do with it is your business.
PART ONE: THE REVERSAL
Trust Is Not a Feeling
In 1966, Peter Wason gave people a simple task.
Four cards on a table. Each has a letter on one side and a number on the other. You can see: A, K, 4, 7.
The rule: If a card has a vowel on one side, it must have an even number on the other.
Which cards do you need to flip to check whether the rule is being followed?
Most people say A and 4.
The correct answer is A and 7. You need to check whether the vowel has an even number (A). And you need to check whether the odd number has a vowel on the other side (7). The 4 card is irrelevant. An even number can have anything on its other side without violating the rule.
Fewer than 10% of university students get this right.
These are people who passed logic courses. People who can solve syllogisms on exams. People who would describe themselves as logical thinkers.
The task is elementary deduction. And they fail it.
But here is the part that matters.
Cosmides and Tooby gave the same logical structure a different surface. Instead of letters and numbers, they used a social scenario: “If a person is drinking alcohol, they must be over 21.”
Cards showing: drinking beer, drinking soda, 25 years old, 16 years old.
Which cards do you flip?
Nearly everyone gets it right. Beer and 16.
The logic is identical. The performance inverts completely.
What This Reveals
The brain does not have a general-purpose logic engine that evaluates propositions neutrally.
It has a cheater-detection module. A social contract enforcer. A system exquisitely tuned to catch people who take benefits without paying costs.
When the problem is framed as abstract logic, the module has no grip. Performance collapses to near chance. When the same problem is framed as someone potentially cheating a social rule, the module activates and performance becomes nearly perfect.
This means something specific.
The capacity for logical reasoning in humans did not evolve to find truth. It evolved to navigate social contracts. To catch free-riders. To enforce reciprocity. To argue.
The feeling of “being logical” is not the feeling of correctly processing information. It is the feeling of the social reasoning module running at high confidence. These are not the same thing, and confusing them is the source of most of what goes wrong when humans try to think.
THE TWO SYSTEMS PEOPLE CONFUSE
WHAT PEOPLE THINK LOGIC IS:
┌──────────────────────────────────────────────────┐
│ │
│ Observation → Evidence → Reasoning → Conclusion │
│ │
│ (Truth-seeking process) │
│ │
└──────────────────────────────────────────────────┘
WHAT LOGIC ACTUALLY IS:
┌──────────────────────────────────────────────────┐
│ │
│ Conclusion → Search for supporting evidence → │
│ Arrangement of reasons → Feeling of having │
│ "figured it out" │
│ │
│ (Justification process) │
│ │
└──────────────────────────────────────────────────┘
PART TWO: THE PRESS SECRETARY
How Reasoning Actually Works
Jonathan Haidt published a model in 2001 that overturned the standard picture of moral reasoning. But the mechanism he identified operates far beyond morality. It operates in every domain where humans believe they are thinking logically.
The model: Intuition comes first. Strategic reasoning comes second.
Not sometimes. Not in weak thinkers. Not when emotions run high. Always. In everyone. Including people who have studied logic. Including people reading this sentence who are already constructing reasons why they are the exception.
The brain receives input. The evaluative system fires. A judgment forms. This happens in milliseconds. Before language. Before deliberation. Before anything that could be called reasoning.
Then the reasoning module activates. Not to evaluate the judgment. To justify it.
Haidt called it the rider and the elephant. The elephant goes where it wants. The rider holds the reins and tells a story about choosing the direction.
But this metaphor is too kind to the rider.
The rider does not merely narrate. The rider actively constructs logical-sounding arguments for wherever the elephant has gone. The rider finds evidence. Builds chains of reasoning. Dismisses counter-evidence. Constructs what feels from the inside like a rational analysis. And the rider believes every word it says.
This is not a metaphor for bias. This is the architecture.
The Evidence
Nisbett and Wilson demonstrated in 1977 that people have no direct access to their own cognitive processes. When asked why they chose one thing over another, subjects confabulated. They generated plausible-sounding reasons that had no connection to the actual causes of their behavior.
They did not lie. They believed their explanations. They experienced the explanations as accurate memories of their own reasoning.
The machinery that generates the explanation is not the machinery that generated the decision. They are separate systems. And the explanation system has no read access to the decision system. It is working from the output alone, constructing a backward narrative that sounds like the kind of process that would have produced that output.
This is what logic is for most people, most of the time.
Not a process of discovery. A process of construction. After the fact.
Myside Bias
Here is the test that reveals the architecture most clearly.
Keith Stanovich spent decades studying what he calls myside bias. The finding: intelligence does not protect against biased reasoning about topics where the person has a prior opinion. High-IQ individuals are no better than low-IQ individuals at generating arguments against their own position. They are significantly better at generating arguments for it.
Read that again.
Intelligence makes you better at defending what you already believe. It does not make you better at questioning it.
This is precisely what you would expect from a system designed to argue rather than a system designed to think. A better engine produces better justifications. Not better truth-tracking. A faster press secretary with a larger vocabulary.
MYSIDE BIAS BY INTELLIGENCE
TASK: Generate arguments FOR your position
Low IQ: ████████████
High IQ: ████████████████████████████
TASK: Generate arguments AGAINST your position
Low IQ: ████████
High IQ: █████████
(Stanovich, 2013 — West, Meserve, Stanovich)
Intelligence amplifies justification.
Intelligence does not amplify self-correction.
The smartest person in the room is not the one most likely to be right. The smartest person in the room is the one most likely to construct an airtight case for being wrong.
PART THREE: THE LOGIC OF WHAT PEOPLE SAY
Speech Is Not Reporting
When someone says “I think the economy is going to crash,” they are not reporting the output of an economic model running in their head.
They are doing one of several things, none of which is what it looks like.
They might be affiliating. Signaling membership in a group that holds that view. The content of the statement is secondary to its social function. The sentence is a badge, not a prediction.
They might be positioning. Establishing themselves as someone with insight. Someone who sees what others miss. The statement serves the speaker’s identity more than it serves the listener’s understanding.
They might be testing. Floating a position to see how others react. Calibrating their public stance based on the social response. The logic comes later, after the reaction tells them which direction is safe.
They might be performing reasoning. Constructing in real time a chain of arguments that demonstrates their capacity to think. Not because the thinking was done privately and is now being shared. Because the performance of thinking is itself the point.
In none of these cases is the person reporting a conclusion reached through logical analysis. The logical analysis, if it appears at all, is the decoration. The structure is social.
Arguments Are Not What They Appear
Two people are arguing. Each presents evidence. Each constructs logical chains. Each identifies flaws in the other’s reasoning. Each experiences themselves as pursuing truth while the other is being obstinate.
What is actually happening is a status negotiation.
The person who changes their mind in an argument loses. Not factually. Socially. The act of conceding signals that your previous position was wrong. That your reasoning was flawed. That the other person’s model of the world is superior to yours.
This is why arguments almost never change minds in real time. The social cost of updating your position in front of the person who challenged it is too high. The logic is real. The evidence is real. The willingness to integrate them is gated by a social computation that has nothing to do with truth.
People do change their minds. But they do it later. Alone. In private. When the social cost of updating is zero. They integrate the evidence quietly, and the next time the topic comes up, they hold the new position as if they had always held it.
The argument did not fail. It worked. But the timeline of the update was controlled by social machinery, not logical machinery. And the person will never acknowledge the argument as the cause. That would be conceding retroactively. The social system prevents it.
WHAT AN ARGUMENT LOOKS LIKE FROM OUTSIDE:
Person A: [evidence] → [logic] → [conclusion]
Person B: [counter-evidence] → [counter-logic] → [counter-conclusion]
Resolution: The better argument wins.
WHAT AN ARGUMENT ACTUALLY IS:
Person A: [identity + status] → [searches for supporting evidence]
→ [deploys logic as weapon] → [conclusion unchanged]
Person B: [identity + status] → [searches for counter-evidence]
→ [deploys logic as counter-weapon] → [conclusion unchanged]
Resolution: Both leave believing they won.
Actual update: Happens 3 days later, alone, unattributed.
The Logic of Persuasion
Since logic does not change minds in real time, what does?
Not better arguments. Not more evidence. Not clearer reasoning.
What changes minds is new group membership. New identity. New incentive structures. New social environments where the desired belief is the default rather than the minority position.
A person surrounded by people who believe X will find themselves believing X. Not because they were persuaded by arguments. Because the social computation shifted. The belief that was costly to hold became costless. The belief that was costly to abandon became costless to release.
Then the logic arrives. The reasons appear. The person constructs a narrative in which they carefully evaluated the evidence and came to a new conclusion.
The narrative is sincere. And it is fabricated.
The logic of persuasion is not logical. It is environmental.
PART FOUR: THE LOGIC OF WHAT PEOPLE DO
Actions Have Their Own Language
A manager says: “My team’s development is my top priority.”
The manager spends no time on development conversations. Cancels one-on-ones. Gives feedback only when performance threatens their own metrics.
Is the manager lying?
No.
The manager means it. At the level of self-concept, team development is genuinely valued. At the level of behavior, it is not. These two systems operate independently and contradict each other routinely.
When they contradict, the behavior is the truth. Always. Without exception.
Not because behavior is more honest. Because behavior is computed by a system that integrates actual costs, actual incentives, actual energy availability, and actual priorities. The verbal system computes what the person wants to believe about themselves. These are different computations with different inputs and they produce different outputs.
The Revealed Preference Machine
Economists have a concept called revealed preference. What you choose tells the truth about what you value. Not what you say you value. What you trade time, money, and energy for.
This is not a moral judgment. It is an observation about architecture.
The system that generates behavior has access to the full cost-benefit computation. It knows what everything actually costs in effort, time, risk, and social capital. It integrates these costs automatically and produces the action that optimizes across all of them.
The system that generates speech has access to the self-concept. It knows what the person wants to be true. It generates language consistent with that self-concept.
When a person says they value health but does not exercise, the speech system is reporting the self-concept (I am a health-conscious person) and the behavior system is reporting the actual computation (the cost of exercise exceeds its perceived benefit given current energy, competing demands, and the fact that the consequences of not exercising are not yet visible).
Both systems are running logic. Real logic. With real inputs and real computations.
They just have different inputs.
TWO LOGICAL SYSTEMS, ONE PERSON
┌──────────────────────────────────────────────────┐
│ SPEECH LOGIC (the verbal system) │
│ │
│ Inputs: │
│ - Self-concept (who I believe I am) │
│ - Social context (who is listening) │
│ - Desired impression (how I want to appear) │
│ │
│ Output: What the person says │
│ │
│ Optimizes for: Identity coherence │
│ │
└──────────────────────────────────────────────────┘
┌──────────────────────────────────────────────────┐
│ ACTION LOGIC (the behavioral system) │
│ │
│ Inputs: │
│ - Actual energy available │
│ - Actual costs (time, effort, risk) │
│ - Actual incentive structure │
│ - Habit architecture │
│ - Immediate vs delayed consequences │
│ │
│ Output: What the person does │
│ │
│ Optimizes for: Cost minimization │
│ │
└──────────────────────────────────────────────────┘
When outputs conflict, action logic is always
the more accurate report of the real computation.
Reading People Through Action Logic
Once you understand that actions and words run on separate systems, you gain a specific capability.
You can read what a person actually values by ignoring what they say and watching what they do. Not as a moral exercise. Not to catch them in hypocrisy. As a measurement technique.
Where does their time go? Not where they say it goes. Where does it actually go? Calendar, not stated priorities.
What do they do under pressure? Not what they say they would do. What do they actually do when the cost of the right action rises? The gap between the two is the gap between their self-concept and their actual computation.
What do they sacrifice? Sacrifice is the purest signal. A person who gives up something real (time, money, comfort, status) for a stated value has a behavioral system that actually computes that value as worth the cost. A person who states the value but never sacrifices for it has a speech system running ahead of a behavioral system that has not been updated.
This is not cynicism. Most people are not lying. They are running two honest systems that happen to produce different outputs. The verbal system is honestly reporting the self-concept. The behavioral system is honestly computing the costs.
The confusion arises because most people believe they are one system. They experience their stated reasons as the actual reasons. They feel the logic that produces their words and mistake it for the logic that produces their actions.
PART FIVE: THE EVOLUTIONARY FUNCTION
Why We Reason
Hugo Mercier and Dan Sperber proposed in 2011 that reasoning did not evolve for individual cognition at all. It evolved for social argumentation.
The argumentative theory of reasoning.
The claim: the human capacity for reasoning is a specialization for producing and evaluating arguments in social contexts. Not for thinking clearly in private. Not for solving abstract problems. Not for finding truth in isolation.
For winning arguments.
This would explain everything that otherwise looks like a design flaw.
It explains why we are better at finding flaws in others’ reasoning than in our own. That is exactly what you would expect from a system designed for argumentation. In a debate, the adaptive value is in dismantling the opponent’s case, not in stress-testing your own.
It explains why reasoning is biased toward our own positions. In an argumentative context, generating reasons for your conclusion is the task. Generating reasons against it is helping the opposition.
It explains why reasoning works best in groups with disagreement. When two biased reasoners argue opposing positions, the flaws in each position get exposed by the other. The group reaches better conclusions than either individual would alone. Not because either individual was reasoning well. Because the argumentative structure created error-correction through opposition.
It explains why solitary reasoning so often leads to poor conclusions. Without an opponent to find the flaws, the justification machine runs unchecked. The press secretary writes the entire narrative with no journalist asking questions.
The Confirmation Engine
Confirmation bias is not a bug in the logic system. It is the logic system.
When a person encounters evidence that confirms their existing belief, the brain processes it with minimal scrutiny. It flows through. It feels right. The prior is reinforced.
When a person encounters evidence that contradicts their existing belief, the brain activates effortful processing. It searches for flaws. It questions methodology. It generates alternative explanations. It finds reasons to dismiss.
This asymmetry is not an error in rational processing. It is the architecture working as designed. A system built for social argumentation needs to rapidly accumulate evidence for its position and rapidly dismantle evidence against it.
The problem is that we live in a world where we need to think clearly about complex problems that have no social structure. Climate data. Medical treatments. Financial decisions. Policy tradeoffs. In these domains, the argumentative architecture works against us. It locks us into positions. It makes the strength of our conviction inversely correlated with the quality of our reasoning about the evidence.
The better you are at arguing, the worse you are at updating.
THE CONFIRMATION ARCHITECTURE
Confirming evidence:
┌──────────────────────────────────────┐
│ Low scrutiny │
│ Fast processing │
│ Minimal counter-argument generation │
│ "Makes sense." │
└──────────────────────────────────────┘
│
▼
Prior strengthened
Disconfirming evidence:
┌──────────────────────────────────────┐
│ High scrutiny │
│ Slow processing │
│ Active counter-argument generation │
│ "Something's wrong with this." │
└──────────────────────────────────────┘
│
▼
Prior maintained
PART SIX: THE CATEGORIES
A Taxonomy of Logic as It Actually Operates
There is not one kind of logic. There is not even one kind of thing that people call logic. The word covers at least five distinct operations, each with its own inputs, its own outputs, and its own function.
Formal logic. The kind taught in philosophy departments. Syllogisms. If P then Q. Modus ponens, modus tollens, truth tables. This is the logic of mathematics and proof. Humans can learn it. Humans almost never use it spontaneously. It requires effortful override of every natural reasoning tendency. When people say “be logical,” they are pointing at this. Almost no one is actually doing this. The few who do it professionally (mathematicians, logicians, programmers) will confirm that it feels nothing like what regular people call “thinking logically.”
Motivated logic. The kind that runs by default. Conclusion first, evidence arranged to fit. Operates continuously, in every domain, in every person. The architecture described throughout this document. When someone says “I’ve thought about it carefully,” this is what happened. Not always. But far more often than anyone wants to admit.
Social logic. The kind deployed in conversations. Arguments, debates, negotiations. Its function is not truth-finding but position-defense and status-negotiation. The logical structures are real. The syllogisms are valid. The evidence is genuine. But the purpose of deploying them is social, and the selection of which evidence and which arguments to deploy is controlled by the social computation, not by a truth-seeking one.
Action logic. The kind that governs behavior. Operates on costs, incentives, energy, and habit rather than on propositions and evidence. Does not use language. Does not produce arguments. Produces behavior. Is the most honest form of logic a person runs because it cannot self-deceive. You can tell yourself you value something. You cannot trick your behavioral system into paying a cost it has computed as not worth paying.
Emotional logic. The kind that operates beneath all the others. Not irrational. Not illogical. Running a different logic. The logic of survival, attachment, threat, and reward. When someone has a “gut feeling” they cannot explain, this is the emotional computation producing an output without the verbal system having access to the inputs. The feeling is the conclusion of a process that ran too fast and too deep for the narrative system to follow.
THE FIVE LOGICS
┌────────────────────┬──────────────────┬──────────────────┐
│ TYPE │ FUNCTION │ AWARENESS │
├────────────────────┼──────────────────┼──────────────────┤
│ Formal │ Proof │ Fully conscious │
│ Motivated │ Justification │ Feels conscious │
│ Social │ Persuasion │ Partially aware │
│ Action │ Optimization │ Unconscious │
│ Emotional │ Survival │ Pre-conscious │
└────────────────────┴──────────────────┴──────────────────┘
Most people think they run Formal.
Most people actually run Motivated.
In conversation, they run Social.
Their body runs Action.
Beneath everything runs Emotional.
Only Formal can be checked against truth.
And Formal is the one that almost never runs.
PART SEVEN: WHAT TO DO WITH THIS
Nothing
This is not a self-improvement document.
There is no technique that makes you logical. No practice that overrides the architecture. No awareness that prevents the motivated reasoning from running. Knowing about the press secretary does not fire the press secretary. It adds a second press secretary who explains why you are now immune to press secretaries.
The architecture is the architecture.
But there is one thing that shifts.
When you stop believing you are logical, you stop defending your positions as if they were you. The identity attachment loosens. Not because you decided to be open-minded. Because you recognized that the thing you were defending was never the output of logic in the first place. It was a position you arrived at through processes you have no access to, decorated with reasons you generated after the fact.
This does not make you right more often. It makes you wrong less expensively. The cost of being wrong drops when you stop experiencing your conclusions as extensions of your identity. Updating becomes cheaper. Not easy. Cheaper.
And there is one observation that functions like a tool.
When you want to know what someone actually believes, actually values, actually computes as important, ignore everything they say. Watch what they do. Watch what they sacrifice. Watch where their time goes when no one is watching.
The logic of action cannot lie.
Everything else can.
The Silence After
Logic will not save you. It never did.
What saves you, in the rare moments anything does, is the willingness to notice that you are not thinking. You are justifying. You are not reasoning. You are defending. You are not searching for truth. You are constructing a case for what you already believe.
That noticing lasts about three seconds before the machinery recaptures it and turns it into another justification.
But in those three seconds, something real happens. The conclusion loosens its grip. The evidence rearranges itself, briefly, into something that looks less like a fortress and more like a landscape. Multiple paths become visible. Not because you applied logic. Because you suspended it.
The suspension is the closest a human gets to clear sight.
It is not logic. It is the absence of logic.
And it is the only honest place to think from.