9 Septebmer 2026

Note

We have now walked through four doors together.

First, I questioned the architecture itself: why should Proposition and Opposition exhaust the possibilities?

Then we gave Israel and Palestine the Backspace Key and asked whether the words in the motion—Israel, never, truly, peace, Palestine—were already carrying conclusions before the debate began.

Next came Popper’s harder lesson: defeating A does not make B true. Winning an argument is not the same as solving the problem.

Now comes Article 5, and perhaps the most uncomfortable question in the series:

What if brilliant people are especially capable of remaining wrong?

Oxford is almost the perfect laboratory. The speakers are articulate, informed, fast, rhetorically formidable. Yet intelligence can become an extraordinary machine for producing arguments for the conclusion one has already been assigned to defend.

That is not rationality.

Rationality begins somewhere more difficult:

What would make me change my mind?

The Khaled Protocol distinguishes intelligence—the capacity to solve problems and construct sophisticated arguments—from rationality: the willingness and institutional capacity to expose explanations to criticism and replace them when better ones emerge.

So this article asks us Kurdish-brade style: after four articles criticizing the two doors, are we prepared to press Backspace on our own door?

Because the real test of rationality is not how brilliantly we defeat the other side.

It is whether criticism can still reach us.

Article 5: Intelligence Is Not Rationality — Why brilliant people can become exceptionally good at defending conclusions they should abandon.

Article 5 Intelligence Is Not Rationality

Why brilliant people can become exceptionally good at defending conclusions they should abandon

The Oxford Union debate on Israel and Palestine contained no shortage of intelligence.

The speakers were articulate.

Historically informed.

Quick with counterexamples.

Skilled at framing.

Capable of anticipating objections.

Able to transform complicated histories into memorable arguments before a live audience.

That is precisely what makes the debate philosophically interesting.

Because intelligence was everywhere.

Rationality was harder to locate.

I do not mean that the speakers were irrational people.

I mean something more structural.

The institution had assigned highly intelligent people to opposite conclusions and then rewarded them for making those conclusions survive.

That is not the same task as discovering whether the conclusions are wrong.

And the distinction matters far beyond Oxford.

What Is Intelligence For?

One of the principles I increasingly use in my own work is:

Intelligence is not rationality.

Intelligence can help someone:

find evidence;

recognize patterns;

remember historical details;

construct analogies;

anticipate criticism;

identify weaknesses in an opponent;

reframe counterexamples;

and produce increasingly sophisticated defenses of an existing position.

All of that may be cognitively impressive.

None of it guarantees that the person will correct an error.

Indeed, intelligence can do the opposite.

A highly intelligent person may be better equipped than an unintelligent one to protect a bad explanation.

Every criticism generates a counterargument.

Every anomaly receives an interpretation.

Every uncomfortable fact is fitted into a larger narrative.

The explanation survives.

Not necessarily because it is good.

Because its defender is good.

That is why rationality requires something intelligence alone cannot supply:

the willingness and institutional capacity to let criticism change the explanation.

Oxford Gives Intelligence a Job

Before the Israel debate began, the speakers already knew what success meant.

Some were speaking for Proposition.

Others for Opposition.

The motion was fixed:

“This House Believes Israel Never Truly Wanted Peace with Palestine.”

The audience would eventually choose between them.

Now imagine the effect of that architecture.

A speaker on Proposition does not enter the room primarily asking:

Is this motion true?

The institutional task is:

Make the strongest case that it is true.

The Opposition receives the inverse task:

Make the strongest case that it is false.

An intelligent person then begins doing what intelligent people do extremely well.

Searching.

Selecting.

Connecting.

Arguing.

Rebutting.

The result can be formidable.

But notice what has happened.

Intelligence has been supplied with an objective function.

The objective is not:

discover the error.

It is:

defend the side.

That is an entirely different optimization problem.

The Debate Tells Us This Explicitly

The transcript occasionally makes the incentive visible.

Ghada Karmi begins by saying that the subject is not really a debate but a statement of fact. Israel’s lack of interest in peace, in her framing, is already established; the remaining task is essentially to explain why.

On the other side, Einat Wilf explicitly says that she intends to address the proposition as framed and prove the counter-case.

These are very different substantive positions.

Structurally, however, they reveal the same thing.

The conclusion has been assigned.

Reasoning follows.

That does not make either speaker dishonest.

It makes them debaters.

And that is precisely the problem I am interested in.

Advocacy Intelligence

We need a name for this.

Call it:

advocacy intelligence.

Advocacy intelligence asks:

How can I make my position stronger?

What evidence helps me?

How can I explain away evidence that hurts me?

What inconsistency can I expose in the other side?

What emotionally powerful example makes my interpretation memorable?

What concession can I safely make without surrendering the conclusion?

These are real intellectual skills.

Lawyers need some of them.

Politicians use them.

Campaigners depend on them.

Academics are not immune to them.

Social media rewards them constantly.

But advocacy intelligence can operate perfectly well without error correction.

Its success criterion is survival of the position.

Rationality Has a Different Objective

A rational person asks something far more dangerous:

What would make us abandon this explanation?

That question changes everything.

Now evidence is not ammunition.

It is potentially destructive.

An objection is not merely something to defeat.

It is an opportunity to discover that one’s explanation fails.

A counterexample is not an inconvenience.

It may be progress.

The rational person needs what Karl Popper built into his conception of knowledge:

conjecture
→ criticism
→ error discovery
→ revised problem
→ new conjecture.

The objective is not protection.

It is correction.

And this requires something psychologically difficult:

the possibility that the best outcome of an argument is to lose an idea.

Losing an Argument Can Be Winning

This is one of the strangest inversions produced by Popperian epistemology.

Imagine that I begin an argument believing A.

You criticize A.

I cannot answer your criticism.

I realize that the explanation I defended was mistaken.

I abandon it.

From the perspective of competitive debate:

I lost.

From the perspective of knowledge growth:

I succeeded.

An error that I carried into the conversation no longer controls my thinking.

That is progress.

The more important the error, the greater the achievement.

Yet our public culture routinely treats this moment as weakness.

Politicians are attacked for changing position.

Debaters are expected to defend their side.

Commentators are rewarded for consistency.

Online communities preserve screenshots of previous statements as evidence of hypocrisy.

We have built institutions in which being corrected can impose a higher status cost than remaining wrong.

That is a remarkable design failure.

Backspace Is Expensive

This is why I use the Backspace Key as a metaphor.

A rational intellectual system must make it possible to delete.

Not people.

Errors.

Assumptions.

Premises.

Failed explanations.

Bad institutional arrangements.

Sometimes entire questions.

But Backspace has a cost when identity becomes attached to the answer.

Once I become:

Proposition

or:

Opposition

criticism acquires a new meaning.

It no longer threatens only the conjecture.

It threatens my performance.

My side.

My reputation.

Perhaps my audience.

Perhaps my political identity.

Perhaps even my moral standing.

The error has acquired protection.

And intelligence can now be recruited to defend it.

The Stronger the Defender, the More Dangerous the Error

This leads to an uncomfortable conjecture:

Under the wrong institutional incentives, greater intelligence can increase the durability of error.

An unintelligent defense may collapse quickly.

A brilliant defense can survive for years.

This is one reason bad ideas can flourish in sophisticated civilizations.

They are not always protected by stupidity.

Sometimes they are protected by extraordinary intelligence.

Lawyers rationalize them.

Scholars theorize them.

Officials bureaucratize them.

Journalists narrate them.

Technologists scale them.

Politicians weaponize them.

The relevant question is therefore never merely:

How intelligent are the people running the system?

The better question is:

Where can their explanations be corrected?

Oxford as a Laboratory

The Oxford debate supplies miniature examples of this.

The Proposition presented settlement expansion, occupation, statements by Israeli leaders, military conduct and historical arguments about Zionism as evidence supporting the motion.

The Opposition presented partition, peace negotiations, territorial proposals, Israeli withdrawals, Palestinian violence and Israeli peace movements as evidence against it.

Now notice what each side’s intelligence must do when confronted with inconvenient evidence.

If Israel negotiated seriously, Proposition must explain why the negotiation does not defeat the motion.

If Israel simultaneously expanded settlements, Opposition must explain why that does not undermine its peace narrative.

If Palestinians accepted compromise at some historical moment, Opposition must integrate it.

If Palestinian organizations pursued violence or rejected proposals, Proposition must integrate that.

An intelligent advocate can usually do this.

The narrative becomes more elaborate.

Exceptions become qualifications.

Counterexamples become contextualized.

Every new fact produces another layer of interpretation.

At some point an explanation can become almost impossible to kill.

That is not necessarily evidence of depth.

It may be evidence that intelligence has become a defense system.

Deutsch’s Hard-to-Vary Test

David Deutsch gives us a way to resist this.

A good explanation should be hard to vary while retaining its explanatory power.

Its components should be there for reasons connected to the problem.

A bad explanation is often flexible in precisely the wrong way.

Something happens?

It can absorb it.

The opposite happens?

It can absorb that too.

A peace proposal proves sincerity.

Unless one already believes it was tactical.

A failed negotiation proves rejectionism.

Unless one already believes the offer was designed to fail.

Violence proves existential hostility.

Unless one already believes violence is the predictable product of occupation.

Withdrawal proves willingness to compromise.

Unless one already believes withdrawal was strategic repositioning.

An explanation capable of digesting every possible observation without risking itself has stopped functioning as a serious conjecture.

Intelligence makes this danger worse because intelligent defenders are extraordinarily good at variation.

They can continually modify the surrounding story while preserving the protected center.

Rationality Means Letting the Center Move

This is where rationality differs.

The rational person does not ask only:

Can I save the theory?

They ask:

Should this theory still be saved?

That is Backspace.

And it applies not merely to peripheral details.

It must be able to reach the center.

Perhaps the problem is not:

Israel never wanted peace.

Perhaps it is not:

Palestinians repeatedly rejected peace.

Perhaps both formulations are too coarse.

Perhaps different actors wanted incompatible things at different times.

Perhaps some genuinely wanted settlements while genuinely wanting peace as they defined it.

Perhaps some Palestinians genuinely wanted statehood through negotiation while others pursued incompatible aims.

Perhaps violence changed what Israelis considered possible.

Perhaps settlement changed what Palestinians considered credible.

Perhaps repeated failures altered both political systems.

The rational mind must permit those possibilities to emerge even if they destroy the argument with which it began.

Bold Conjecture Is Not Bold Advocacy

This is why I distinguish Bold Conjecture from ideological confidence.

A bold conjecture is not:

I will defend an extreme claim without retreat.

That is stubbornness.

A genuinely bold conjecture says:

Here is my best explanation. Now try to destroy it.

The boldness lies partly in exposure.

The conjecture risks death.

And its author must be prepared to help kill it.

This is the opposite of much contemporary debate.

We frequently treat the person who never retreats as strong.

Popperian strength looks different.

It is the capacity to say:

That criticism reaches the foundation of my explanation. I need another explanation.

That is not weakness.

It is rational courage.

Brett Hall and the Preservation Problem

Brett Hall’s principle is:

Do not destroy the means of error correction.

Apply that to debate.

An institution preserves error correction when participants can:

criticize one another;

change their own minds;

revise their positions;

introduce new explanations;

acknowledge successful criticism;

and leave the discussion with fewer protected premises than they brought into it.

Oxford strongly protects the first.

Competitive binary structure can make the rest difficult.

The speakers may be free to speak.

But are they institutionally rewarded for correction?

That is the harder test.

Freedom of speech is indispensable.

It is not sufficient.

A room can contain unlimited speech while every participant remains trapped inside a conclusion.

The Most Rational Sentence in the Room

Imagine a future Oxford debate in which a speaker stands halfway through and says:

“I came here assigned to defend Proposition. An argument from Opposition has convinced me that part of my case is wrong. I still reject their overall explanation. I now think both sides have missed a third possibility.”

What should happen?

In the traditional architecture, this creates an awkward problem.

Has the speaker failed Proposition?

In an error-correcting institution, the audience should recognize something else:

Knowledge may just have grown.

The speaker has used criticism rather than survived it.

That should be a mark of success.

Perhaps the most rational sentence anyone can utter in a debate is:

I was wrong about that.

The fact that it is so rare should concern us.

The Intelligence Trap in Politics

The problem extends far beyond Oxford.

Political parties work the same way.

A person joins a party.

Now intelligence is recruited to defend the party’s package.

The politician does not independently rediscover a position on taxation, immigration, climate, education, foreign policy and constitutional reform every morning.

Positions arrive bundled.

One’s side generates explanations.

The other side generates objections.

Each camp becomes increasingly sophisticated at demonstrating that the other is wrong.

Social media intensifies this.

An argument is not evaluated only for explanatory quality.

People ask:

Who said it?

Which side benefits?

Who will share it?

Who will attack me if I concede it?

The question:

Is this true?

is crowded out by:

Whose side does this help?

That is advocacy intelligence at civilizational scale.

Intelligence Without Backspace

Now we can formulate the danger more precisely.

Intelligence without Backspace is the capacity to solve problems while protecting the premises that generated them.

It can be astonishingly effective.

A state can become better at administering a failed policy.

A military can become better at fighting an unwinnable war.

A bureaucracy can become better at implementing the wrong rule.

An ideology can become better at explaining away its failures.

A debater can become better at defending a proposition that criticism should have killed.

This is why improvement in performance is not necessarily improvement in knowledge.

A system can become increasingly competent while moving further from correction.

What Should a Rational Debate Reward?

If the purpose is knowledge growth, the incentive structure must change.

A rational debate should reward at least five things.

First: successful criticism.

Not rhetorical humiliation.

Identification of an actual error.

Second: successful correction.

A participant who revises an explanation after criticism should gain intellectual credit, not lose it.

Third: problem reformulation.

If the debate reveals that the original motion was badly framed, changing the question should be permitted.

Fourth: alternative creation.

A new conjecture generated during the discussion should have institutional standing even if it belongs to neither side.

Fifth: preservation of Backspace.

No conclusion should become immune simply because its side won.

That would transform the meaning of debate.

The objective would cease to be:

Who defended their answer best?

It would become:

What became more corrigible because we argued?

Intelligence Is Valuable

None of this is an argument against intelligence.

Quite the opposite.

Intelligence is one of the great resources of human civilization.

But a powerful resource requires the right architecture.

Nuclear energy can illuminate cities or destroy them.

Computational power can solve problems or scale surveillance.

Institutional capacity can protect freedom or perfect repression.

And intelligence can create explanations or construct fortresses around errors.

The question is always:

What problem is the capability being used to solve?

Oxford’s brilliant advocates demonstrate how much intellectual power human beings possess.

The next step is to design institutions that make that power serve correction.

The Rationality Test

So I propose a simple test.

At the beginning of a debate, ask every participant:

What evidence or criticism would make you change your mind?

At the end, ask:

Did anything tonight do so?

If the answer is always no, something has gone wrong.

Perhaps every speaker happened to enter with the correct explanation.

Possible.

But improbable.

A more disturbing possibility is that the system is better at generating defenses than corrections.

That is precisely what an error-correcting civilization must learn to detect.

The deepest intellectual danger is not stupidity.

Stupidity is often visible.

It collapses under pressure.

The more dangerous condition is an error surrounded by enough intelligence to protect itself indefinitely.

That is why intelligence is not rationality.

Intelligence gives us the power to answer.

Rationality preserves the possibility that the answer itself is wrong.

The point of criticism is not to discover which intelligent person can defend an error longest. It is to make even the most intelligent person capable of pressing Backspace.