Searle's room has emptied, but the question is still standing

John Searle died on 17 September 2025. He was ninety-three.

The philosophical world remembers him, for the most part, by a single thought experiment: the Chinese Room.

But to commemorate Searle as "the philosopher who said computers cannot really think" is to reduce one of the most influential philosophers of mind and language of the second half of the twentieth century to a slogan.

For Searle's real question had nothing to do with computers.

The real question was this: where does the meaning of a thing come from?

And it took him from language to mind, from mind to society, from society to money, to the state, to the university, and at last to reality itself.

Language first: what does a sentence do?#

The first great period of Searle's career is the philosophy of language. At Oxford he systematised the road opened by J. L. Austin, and in 1969, with Speech Acts, he set up a framework of his own.

The basic thesis: language does not merely describe the world; it does work in it.

"I swear." "I declare." "I promise." "I call this session to order."

These are not descriptions that could be true or false. They bring something about.

Searle classified such acts and listed the conditions each of them has to satisfy in order to succeed. For a promise to be valid: the speaker must genuinely intend to perform some future action, the hearer must want it, and the action must not be one that was going to happen anyway.

This is a technical contribution to the philosophy of language. But it carries a seed within it: institutional reality is established by acts of language.

A court judgment, a marriage, a contract, a declaration of war — all of them exist by way of speech.

Then the Chinese Room#

In 1980 Searle published a paper that would change the history of the argument over artificial intelligence.

The thought experiment runs as follows.

In a room there is a person who knows no Chinese. In his hands is a thick book of rules written in English. Sheets of paper bearing Chinese characters come in under the door. The person consults the instructions in the book: "if a symbol of this shape comes in, send out a symbol of that shape." And he passes the correct answer back out.

Seen from outside, the room appears to know Chinese. The person inside does not know a word of it.

Searle's conclusion: to process symbols according to rules is not to understand what they mean.

In the technical formulation: syntax is not sufficient for semantics.

Why the argument keeps coming back in the face of today's large language models is plain enough. The systems are far more intricate now: they build context, draw on earlier conversation, reach for tools, assess their own output, make plans.

Does any of this dispose of Searle's objection?

No. But it makes the argument harder. For the book of rules inside the room is no longer a single book; it is a structure of billions of parameters that nobody could read through one by one.

"Searle was against artificial intelligence" — the commonest mistake#

Searle was not arguing that computers are of no use.

His distinction was clear: weak artificial intelligence and strong artificial intelligence.

Weak artificial intelligence: computers can model mental processes, and they are extraordinary instruments for studying them. This Searle accepted.

Strong artificial intelligence: running the right programme endows the computer with a mind in the literal sense of the word. This was what Searle objected to.

The distinction now stands in the textbooks as one of the founding historical divisions in the philosophy of artificial intelligence.

Biological naturalism: a third way#

To understand where Searle stood, one has to look at what he rejected.

It was not dualism. He was not saying that the soul is a separate substance.

Nor was it reductive materialism. He found the reduction of consciousness to the firing of neurons inadequate.

He gave his own position the name biological naturalism. The formula is simple and provocative:

The brain produces consciousness — just as the stomach produces digestion.

Consciousness is a real biological phenomenon. It arises causally out of brain processes. But it has a first-person ontology: to be in pain is really to be in pain, and you cannot convert that into a third-person description.

The position was criticised from both sides. Materialists found him a closet dualist; dualists accused him of not going far enough. Searle gave both the same answer: the trouble is in our concepts; the categories have been set up wrongly.

One wing of the argument over the hard problem, which we set out this week in our file on Chalmers, comes straight out of this. Chalmers takes functional organisation to be decisive; Searle takes biology to be. The position Anil Seth defends today has its ancestry here as well.

The revolution nobody discusses: what is money?#

Searle's most surprising turn came in 1995, with The Construction of Social Reality.

The questions he asked look simple and are very hard to answer.

Why is a piece of paper money? How does a person become president? How does a building count as a university? How do two people become married?

Searle's answer rests on two concepts.

Collective intentionality. People can form intentions not only in the form "I want" but in the form "we are doing". This cannot be reduced to the sum of individual intentions.

Status function. An object is assigned a function that does not derive from its physical properties. The formula runs thus:

"X counts as Y in context C."

This piece of paper (X) counts as money (Y) in the context of the Republic of Turkey (C).

And status functions cannot exist without language. For an object to bear a function that has nothing to do with its physical properties, that function must be capable of being represented. An animal can use an object as a tool; but it cannot say "this is my property".

Three separate books of Searle's thus become a single chain: language → mind → institution.

And a middle course is established: social reality is neither pure imagination nor pure physics. Money, marriage, the state and the law are real — but their reality rests on collective acceptances that have to be kept up.

Withdraw the acceptance and the institution withdraws too. This is the analytic philosopher's version of the question we asked this week in our files on Cicero and on Coke and Cromwell: what is it that holds institutions up?

A contested legacy#

To commemorate Searle is not only to commemorate his ideas.

In 2016 he was accused of sexual harassment and retaliation. The investigation found that he had violated the university's harassment and retaliation policies, and in 2019 he was stripped of his emeritus status.

That is not a footnote; it is an institutional decision, and serious journalism cannot pass over it.

The voices raised around Berkeley after his death were divided too. A former doctoral student described Searle as having moved "in his own orbit" within the department, said colleagues had felt they had no say over his conduct, and found his philosophical manner "quite arrogant" besides. Another figure, who had worked as his research assistant for many years and who runs the Berkeley Social Ontology Group, remembered him as an extraordinary teacher — a man who could explain complicated concepts in ordinary language, and who was easy to approach — and held that the allegations had not been fully represented.

The relation between a thinker's work and a thinker's life is itself one of philosophy's own subjects. There is no easy formula here. But this much can be said: arguments are assessed independently of the character of whoever advances them, and institutional findings are recorded separately, and plainly.

What is left?#

Searle's legacy is at work today on three fronts at once.

Language. What is a system doing when it says "I'm sorry"? On speech act theory, an apology is valid only if the speaker is in certain mental states. Does a language model meet that condition?

Mind. How are we to determine whether a system is conscious? The proposal of a behavioural inference principle we took up this week is a move in precisely the direction Searle rejected.

Institution. Can legal personality, property or responsibility be attributed to an artificial intelligence? Searle's formula for status functions applies directly: X counts as Y in context C — and all that is needed for it is collective acceptance.

The third of these is the least discussed and perhaps the most practical. For legal systems can attribute a status to an artificial intelligence without the question of whether it is "really" a person having been settled at all.

That was exactly what Searle taught: statuses are not discovered; they are conferred.

The last question#

The door of the Chinese Room has closed.

But the question in front of the door is still standing — and the most exasperating thing about Searle was that the question cut both ways.

That a system gives the right answer does not prove that it understands.

Then how are we to prove that it understands?

---

In Turkish: books including Speech Acts, The Rediscovery of the Mind, Mind, Language and Society, The Construction of Social Reality and Consciousness and Language have been translated.