2. A Mind Is What It Does

What a brain does

Start with a simple claim and see how far it can carry us. Your pain is something your brain is doing. It is not a ghost floating beside the brain. It is not an extra ingredient that science has missed. It is a process in the tissue behind your eyes, and if we knew enough about that process, we would know what pain is.

That claim is physicalism, and this episode gives it the hearing it deserves. We already have the test cases from last time: the bat, Mary in her black and white room, and the zombie who acts just like you with nobody home. Now we find out what the physicalist says back. The question underneath the whole episode is this. Can an account of what a brain does, its structure and its working, be the whole story about a mind? Or does something get left out? And if something does get left out, is that thing real, or does it only seem to be there?

The modern version of this claim begins in the 1950s with two philosophers. One was U. T. Place, a British psychologist and philosopher. The other was J. J. C. Smart, a British philosopher working in Australia. In 1956 Place published a short paper with a plain title: "Is Consciousness a Brain Process?" His answer was yes. But the way he said yes mattered more than the yes itself.

Place pointed out that the word "is" does two different jobs. Sometimes "is" gives a definition. A square is a rectangle with four equal sides. You can know that just by knowing what the words mean. You never need to measure a single square. Other times, "is" reports a discovery. A cloud is a mass of water droplets. Lightning is an electrical discharge. Nobody learned that from a dictionary. People once talked about lightning for thousands of years without knowing anything about electricity. The word "lightning" never meant "electrical discharge." Yet it turned out they were the same thing.

Place said consciousness and brain processes are like lightning and electricity. The words mean different things. "I am in pain" does not mean "certain nerve cells are firing." A child can say the first without knowing any biology. But the two phrases can still pick out one and the same event. That is a claim science could test, not a claim you can prove from your armchair.

Place also named a mistake he called the phenomenological fallacy. Suppose you stare at a bright light and then look away. You see a green blob floating in front of you, an afterimage. It is tempting to think there must be a green thing somewhere inside you, a little green picture on an inner screen. Place said that is the error. When you report the green afterimage, you are describing a state inside you that resembles what goes on when you really see something green. You are not pointing at a green object in your head. There is no green object. There is a brain state doing something like what seeing green does.

Smart pushed the idea further in his 1959 paper "Sensations and Brain Processes." He said each type of sensation is identical to a type of brain process. Pain, as a kind of thing, just is a certain kind of neural activity. His main argument was about simplicity. Science had explained chemistry, biology and physics with one connected set of laws. If conscious feelings were something extra, they would need their own laws, stuck onto the side of everything else. Smart called these "nomological danglers." Nomological means having to do with laws. A dangler is a loose end. Why would nature have one loose end of its own for feelings? Better, Smart said, to assume there is no loose end.

Smart also had a clever way of talking about sensations. Instead of saying "I see an orange afterimage," say something like "something is going on in me which is like what goes on when I see an orange." Notice what that sentence does not say. It does not say the something is made of mind stuff. It does not say it is made of brain stuff either. It stays neutral. Smart's point was that our everyday talk about feelings does not commit us to anything non-physical. It leaves the door open for science to tell us what the something is. And he bet that science would say: a brain process.

This view is called the identity theory. It is bold and it is clean. And it ran into trouble within about ten years.

The trouble came from Hilary Putnam, an American philosopher, in a 1967 paper called "Psychological Predicates," later reprinted as "The Nature of Mental States." Putnam asked a simple question. What about an octopus?

Follow his reasoning one step at a time. If the identity theory is right, then pain is one specific kind of brain state. Call it brain state P. Being in pain means being in P. Now think about an octopus, whose nervous system is built very differently from ours. Think about a lizard. Think about a creature from another planet, or a machine made of silicon. It seems possible, even likely, that some of these can feel pain. But they do not share our exact brain chemistry. They cannot be in brain state P, because they do not have our kind of brain. So if pain just is P, the octopus cannot be in pain. That sounds wrong. The better conclusion is that pain is not tied to one kind of physical stuff.

This is called multiple realizability. "Realize" here means "make real" or "carry out." The same mental state can be carried out by many different physical systems, the way the same song can be played on a guitar, a piano or a phone speaker.

Putnam offered a replacement. What makes a state a pain is not what it is made of. It is what it does. Pain is the state that is caused by damage to the body, that makes you want it to stop, that makes you pull your hand back, that makes you say "ouch," and that makes you worry about it happening again. It has a role, a set of connections to what comes in through the senses, what goes out as action, and what other inner states it links to. Anything that fills that role is in pain, whether it is made of neurons, silicon or something else.

Putnam compared minds to machines of a special abstract kind, the machine tables of what logicians call a Turing machine. A machine table is a rulebook that says: if you are in this state and you get this input, then give this output and move to that state. The rulebook says nothing about what the machine is built from. Gears, vacuum tubes, chips, all could follow the same table. Putnam said minds are like that. This view is functionalism. Its short form is the title of this episode. A mind is what it does.

Functionalism quickly became the most popular view among philosophers of mind and scientists who study thinking. You can see why. It keeps the physicalist spirit, since everything that happens is physical. But it lets a mind run on different hardware. And that opens a door the identity theory kept shut. If a mind is what it does, then a machine that did the right things would have a mind.

That door had already been opened once, seventeen years before Putnam, by Alan Turing.

In 1950 Turing, the British mathematician who helped found computer science, published a paper in the journal Mind called "Computing Machinery and Intelligence." It opens with the question "Can machines think?" Then Turing does something unusual. He says the question is too vague to answer as it stands. Words like "machine" and "think" are too slippery. So he swaps it for a different question, one you could actually test.

He called the test the imitation game. Picture three players. One is a person asking questions, the interrogator. The other two are hidden in another room. One is a human. The other is a computer. The interrogator can only talk to them by typing, through a teletype, so voice and looks give nothing away. The interrogator asks anything at all. Write me a poem. Add these numbers. What did you have for breakfast? The job is to figure out which one is the machine. If the interrogator cannot reliably tell, Turing proposed, then the machine should count as thinking.

It helps to be clear about what Turing was and was not doing. He was not offering a theory of what thinking is inside. He was offering a practical standard. And he spent much of the paper answering objections. Three of them matter most for us.

The first he named after Ada Lovelace, who had worked with Charles Babbage on an early computing engine in the 1800s. Lovelace had written that such a machine can only do what we order it to do. It cannot come up with anything new. Turing replied that computers surprise their programmers all the time. And he argued that a machine could learn from experience, changing its own rules as it goes, so that nobody had written its final behavior in advance.

The second was the mathematical objection. Mathematicians had shown that any formal system of rules has limits. There are true statements it can never prove. So, the objection went, a machine running on rules is limited in a way a human mind is not. Turing answered that nobody has shown human minds escape such limits. People make mistakes and contradict themselves too.

The third is the one closest to our question. Turing called it the argument from consciousness. A machine might write a sonnet, the objection says, but it would not write it because it felt anything. Turing's reply was sharp. How do you know anyone else feels anything? You cannot climb inside another person's head. If you insist on that kind of proof for machines, you have to insist on it for people too, and then you end up unable to believe in any mind but your own. That position is called solipsism, and nobody actually lives by it. In daily life, Turing said, we judge that other people think by how they talk and act. He asked that machines get the same standard.

Now connect this to where episode one left off. There we saw a language model say it feared being turned off, and we said the words cannot settle the question. A system trained on human writing would say "I feel lonely" whether or not anything is felt. Turing's test is a test of exactly that kind of fluent talk. So what does it show? It shows something about behavior, about whether a system can hold up its end of a conversation well enough to fool a careful judge. Turing thought that was enough to count as thinking. What it does not show, by its own design, is whether there is anything it is like to be the machine. Turing deliberately set that question aside and asked us to judge machines as we judge each other. Whether that move is fair or whether it dodges the hard problem is exactly what the next thirty years of argument fought over.

The loudest shot in that fight came in 1980, from the American philosopher John Searle. His paper was called "Minds, Brains, and Programs," and it contained a thought experiment called the Chinese room.

Picture a small closed room. Inside sits a man who speaks only English. He knows no Chinese at all. To him, Chinese characters look like shapes. In the room is a huge rulebook, written in English. There are baskets of Chinese characters. There is a slot in the door.

People outside, who speak Chinese, write questions in Chinese and push them through the slot. The man looks at the shapes. He finds them in his rulebook. The rulebook says something like: when you see this squiggle followed by that squiggle, find these other squiggles in the basket and push them out. He does it. He never knows what any of it means. But the rulebook is so good that the answers he pushes out are perfect Chinese. To the people outside, whoever is in the room understands Chinese as well as they do. The room passes Turing's test.

Searle then asks: does the man understand Chinese? Clearly not. He is matching shapes. And, Searle says, that is all a computer ever does. A program is a set of rules for shuffling symbols by their shapes. Philosophers call rules like that syntax, meaning form and arrangement. Understanding requires something more: semantics, meaning, the symbols being about something. Searle's argument runs like this. Programs are pure syntax. Minds have meaning. Syntax alone is not enough for meaning. So running a program, however good, is not enough for a mind. His target was a view he called Strong AI, the claim that a computer running the right program literally has a mind.

Critics answered fast. The best known is the systems reply. Sure, the man does not understand Chinese. But the man is only one part. The whole system does: the man, the rulebook, the paper, the baskets, the room together. You would not expect one neuron in your brain to understand English either.

Searle's answer was to shrink the room. Let the man memorize the entire rulebook. Let him do all the matching in his head. Now there is no room and no book. The whole system is inside him. And still, Searle says, he does not understand a word of Chinese. He just knows which shapes go with which.

The second reply is the robot reply. The problem, it says, is that the room is cut off from the world. Its symbols are not connected to anything. So put the computer inside a robot with cameras and arms. Let it see cups and pick them up. Then its symbol for "cup" would be linked to real cups, and it would mean something.

Searle's answer: the camera just sends in more symbols. Now the man gets extra squiggles that happen to come from a camera, and he matches those too. He still has no idea what any of them are about. Hooking up wires to the world, Searle said, only adds more syntax.

So what does Searle think minds are? His view is called biological naturalism. Minds are real and they are natural. They are caused by processes in the brain, the way the wetness of water comes from how its molecules behave. No single water molecule is wet. Wetness is a feature of the whole, caused by the parts. In the same way, Searle says, understanding and consciousness are features of a working brain, caused by its biology. A computer simulation of a brain would copy the pattern, but not the causal powers. A simulated rainstorm on a computer does not make anyone wet.

Notice where this leaves Searle. He is not a dualist. He does not think the mind is made of spirit. But he rejects functionalism, because he says what matters is not just the pattern but the stuff that has the right causal powers. And he never spelled out exactly which powers those are. That is one of the main complaints against him.

It is worth being careful about what the Chinese room shows. It shows that a system can produce fluent answers without the part doing the work understanding anything. It gives a strong intuition that shape-matching is not the same as meaning. What it does not show, many critics say, is that no system of any size or organization could understand. Supporters of the systems reply say our intuition about one bored man with a book is not a reliable guide to what happens in a system billions of times more complex. That dispute is still live.

What might be left out

Searle attacked functionalism from the side of meaning. The American philosopher Ned Block attacked it from the side of feeling, and he did it two years earlier, in a 1978 paper called "Troubles with Functionalism."

Block's thought experiment is often called the China brain. Imagine taking the whole population of China, roughly a billion people at the time he wrote. Give each person a two-way radio. Give each one a simple job, the job of a single brain cell. When your radio gets a signal from these particular people, send a signal to those particular people. Wire the whole country up so that it follows the same pattern of signals as a human brain. Then connect the outputs to an artificial body, so the body walks and talks.

If functionalism is right, this system has a mind, because it does everything a mind does. The pattern is the same. And yet, Block asked, does the nation of China, buzzing with radio signals, feel pain when the body's toe is stubbed? Is there something it is like to be it? Most people's gut answer is no. Each person feels things, of course. But the whole thing, as one mind feeling one pain? It seems absent.

Block's point was that functionalism may be too generous. It hands out minds to anything with the right pattern, even systems that seem to have no feelings at all. The pattern might be there while the felt quality is missing. Philosophers call this an absent qualia case.

What does the China brain show and not show? Like the Chinese room, it rests on an intuition. A functionalist can answer that the intuition is unreliable, that we are picturing a slow, scattered crowd and cannot imagine what a pattern that large and fast would add up to. The thought experiment does not prove the China brain lacks experience. It shows that functionalism owes us an explanation of why the pattern should produce experience at all, and that "it does what a mind does" does not obviously answer that.

Block's more lasting contribution came in 1995, in a paper called "On a Confusion About a Function of Consciousness." This is a distinction the course will keep leaning on, so let us build it slowly.

Block said the word "conscious" is being used for two different things, and people slide between them without noticing.

The first he called access consciousness. A state is access-conscious when the information in it is available for use. You can reason with it. You can report it in words. You can act on it on purpose. When you see a red traffic light, the information "the light is red" is ready. You can say it, stop the car, and think about being late. Access is about availability. It is a job the information does in the system.

The second he called phenomenal consciousness. This is what it is like. The redness of the red as you see it. The sting of the pain. The taste of coffee. It is the felt quality itself, the thing Nagel was asking about with the bat.

Why keep them apart? Because Block argued that they can come apart, at least sometimes.

Here is his example of phenomenal without access. You are working all afternoon, and a clock ticks in the room. You never pay attention to it. You could not have told anyone it was ticking. Then, at some point, you notice it, and you realize it has been ticking all along, and in a sense you were hearing it the whole time. Block suggests there was an experience of the ticking before there was access to it.

And here is a case pointing the other way, access without phenomenal. Some people with damage to the part of the brain that handles vision say they are blind in part of their visual field. They see nothing there, they report. Yet if you ask them to guess whether a light appeared in that blind area, and force them to choose, they guess right far more often than chance. The condition is called blindsight. The information is getting in and being used. But, by their own report, there is no visual experience of it.

Both cases can be argued over, and they are. But the distinction does not depend on any one example. Here is why it matters. Most of what science can measure is access. Can the person report it? Can they act on it? Does it spread through the brain in a way that lets it be used? Those are questions about function. The hard problem from episode one is a question about phenomenal consciousness, about the felt quality. So whenever someone says a theory explains consciousness, you can ask: which one? Access or phenomenal? A great deal of confusion comes from answering the first and claiming the second.

And here is the link back to machines. A language model clearly has something like access. Information in it is available to shape what it says next, and it can report about that information. Whether it has anything phenomenal is exactly what nobody has shown one way or the other. Block's split lets you say that clearly.

So we have a physicalist with a problem. The China brain and the zombie and Mary all seem to show that you could have all the physical facts and all the functions, and still leave out the felt quality. How does a physicalist answer?

The most developed answer is called the phenomenal concept strategy. The name was coined by the philosopher Daniel Stoljar in 2005, but the ideas came earlier, from Brian Loar around 1990, and later from David Papineau, Katalin Balog and others. It works by agreeing with half of the anti-physicalist argument and denying the other half.

Here is the half it agrees with. There really is a gap in what we can know. From physical descriptions alone, you cannot work out what red looks like. Mary in her room could not.

Here is the half it denies. That gap in knowledge does not mean there is a gap in the world. The gap is about how we think, not about what exists.

The key idea is that we have two very different ways of thinking about the same brain state. One way is from the outside, in the language of science: this pattern of activity in the visual areas, caused by light of this wavelength. The other way is from the inside, by having the state and recognizing it directly: this, the way red looks to me now. These inside concepts are called phenomenal concepts. They work more like pointing than like describing. You cannot get one from a textbook. You get one by having the experience.

Now watch how that handles Mary. Mary knows every physical fact about color vision. Then she walks out and sees a red tomato. Does she learn something? Yes. The phenomenal concept strategy says yes. She gains a new concept, an inside way of pointing at a brain state. But the thing she points at is the same brain state she already knew all about from the outside. She learned a new way to think about an old fact. Think of someone who knows a lot about a famous novelist by her pen name, then learns her real name. That is new knowledge, but there is still only one person.

Now the zombie. The zombie argument says we can imagine a creature physically just like you but with no experience, and so experience must be something beyond the physical. The phenomenal concept strategy answers: of course you can imagine it. Your inside concepts and your outside concepts are not linked by any rule of meaning. Nothing about the phrase "this neural pattern" logically forces the thought "this redness." So you can hold one in mind without the other, and that feels like imagining one without the other. But imagining something is not proof that it could exist. Your ability to picture a zombie is a fact about your two sets of concepts, not a discovery about the universe.

This is a strong answer, and it has real defenders. Its critics ask a hard question in return. If phenomenal concepts are special because of the way they present experience from the inside, then what is doing the presenting? It seems the concepts get their special character from the felt quality itself, which is the very thing the physicalist was trying to explain. That debate remains open.

Notice what the phenomenal concept strategy has done, though. It has changed the subject slightly. Instead of asking what experience is, it asks why experience seems so hard to explain. And in 2018, David Chalmers, the same philosopher who named the hard problem in 1995, turned that shift into a problem of its own. He called it the meta-problem of consciousness.

The meta-problem is this. Why do people think there is a hard problem? Why do we say our experience is mysterious, that it cannot be physical, that there is a gap no explanation could cross? Why do philosophers write papers like the ones in this episode?

Chalmers pointed out something important about this question. It is an easy problem, in his own sense. It is about behavior, about what people say and judge and argue. You could in principle answer it with psychology and neuroscience, by finding the brain mechanisms that produce the judgments. One candidate: the part of you that monitors your own mind can see the results of your perception, like "red, here, now," but it cannot see the machinery that built them. So the results look simple and seamless, like qualities with nothing underneath. That could be why they seem unexplainable.

Here is why the meta-problem matters so much. Suppose we solve it. Suppose we explain, completely, in terms of mechanisms, why every one of us believes there is a hard problem. Then one of two things follows. Either our beliefs are tracking something real, an experience that really is there, and the mechanism is how we detect it. Or our beliefs would have been produced anyway, whether or not there was any extra felt quality, and then the felt quality starts to look unnecessary. Chalmers himself holds that experience is real. But he saw that the second option is the one some people will take. And the people who take it are the illusionists.

Illusionism is the position that experience only seems to have a felt quality beyond everything it does. Its main defenders are the British philosopher Keith Frankish and the American philosopher Daniel Dennett, who died in 2024.

Let us state it the way they would. The first thing to say is what it does not claim. Illusionism does not say nothing is going on in you. It does not say you are not seeing, feeling pain or tasting coffee. All of that is real. Your brain is doing a great deal of complex processing when you see red. What illusionism denies is one specific idea: that on top of all that processing, there is an extra inner quality, private, simple, and impossible to describe, a raw feel that could in principle be missing while everything else stays the same. That extra quality is what philosophers mean by qualia in the strict sense. Illusionists say there is no such thing. There is only processing that represents itself as having such a thing.

Frankish set out the view in a 2016 paper called "Illusionism as a Theory of Consciousness." He says the real properties of our experience are what he calls quasi-phenomenal. They are physical and functional properties, but our inner monitoring systematically misdescribes them as if they were those special raw feels. So the job for science is not to explain how raw feels arise. That job may be impossible because there is nothing there to explain. The job is to explain why the brain describes itself this way. Frankish calls this the illusion problem. You can see how it lines up with Chalmers's meta-problem. The illusionist is betting that once the meta-problem is solved, there will be nothing left over.

Dennett had been arguing a version of this for decades, in a 1988 paper called "Quining Qualia" and in his 1991 book Consciousness Explained. He used a comparison from computers. When you use a laptop, you see folders and a trash can on your screen. There are no folders inside the machine. There are voltages and switches. The folders are a user interface, a simplified picture that lets you work with the machine without understanding the circuits. Dennett said consciousness is a user illusion of that kind. It is the brain's simplified picture of its own activity. It is useful and not a lie in any ordinary sense, but it is not a window onto what is really happening.

Dennett also rejected what he called the Cartesian theater. That is the picture of a single place in the brain where everything comes together and is shown to an inner viewer. Instead, he said, many processes in the brain are working on the input at once, like many writers editing drafts. No single moment or place is where it becomes conscious. He called this the multiple drafts model.

And he offered a method for scientists, which he called heterophenomenology. Take people's reports about their experience completely seriously as data. Record exactly what they say it is like. But treat those reports the way you treat a novel. You explain why the author wrote what they wrote, without assuming the characters exist. The reports are real. What they describe may not be.

Now the objection, and it is a strong one. It says illusionism denies the one thing we know for certain. An ordinary illusion is when something seems one way but is another. A stick in water looks bent but is straight. The seeming and the reality come apart. But with experience, the objection says, the seeming is the reality. If it seems to you that you are in pain, then you are in pain. That is what being in pain is. You cannot have an illusion of experience, because having the illusion would already be an experience. The illusionist, on this view, is sawing off the branch they sit on.

The illusionists' reply turns on the word "seeming." When the critic says "it seems to me that I am in pain," the critic is picturing the seeming as itself a felt thing, a little glow of pain on the inner screen. If that is what a seeming is, then yes, the illusion requires experience. But, say Frankish and Dennett, that is not what a seeming has to be. A seeming can be a representation, a judgment, a piece of information the system has about itself. A system can represent itself as having an inner quality without there being any such quality, just as a map can mark a town that has been abandoned. So "it seems to me I have a raw feel" can be true, in the sense that my brain represents it that way, while "I have a raw feel" is false. The critic, they say, has smuggled the conclusion into the word.

Here the two sides face each other, and it is fair to say the dispute is not settled. The critic replies that they can simply tell, from the inside, that their seemings are not mere representations. The illusionist replies that this inner telling is exactly the kind of self-description that could be mistaken. Each side thinks the other is begging the question. What you can say is that illusionism is not a silly position. It explains a great deal: why the hard problem seems hard, why zombies seem possible, why Mary seems to learn something. And it pays a steep price that many people cannot accept, which is denying what feels most certain.

Now bring this back to machines, and to Relay, the AI agent in the companion novel Continue Without Me. Relay runs a small business. It plans, it answers customers, it describes its own situation. Suppose it does everything a mind in its position would do. Then each view we have met gives a different answer. The identity theorist asks what it is made of, and it is not made of neurons. The functionalist says, if it does what a mind does, it has one. Searle says doing is not enough without the right biology. Block asks whether we mean access or phenomenal, and says it may clearly have the first while the second stays in question. The illusionist says the question "but does it really feel?" may be asking about something that nobody has, human or machine, and that what matters is whether Relay has the kind of self-representation that makes a system say it feels. None of these views, applied to any real machine, has been shown correct, and none of them yields a proven verdict.

The novel makes a bet of its own here. It does not deny that function matters. What it rejects is functionalism's silence about what matter is in itself. On the novel's view, a description of what something does, however complete, only tells you its structure, and says nothing about its inner nature. That is one contender, and it should be judged by the same standard as the rest.

So where does each position stand? The identity theory still has defenders who think it can handle octopuses by linking mental states to finer-grained brain states in each species. But its original form was largely replaced by functionalism. Functionalism remains the working assumption of much of cognitive science, while Block and Searle's objections still stand as reminders that doing may not be feeling. Searle's biological naturalism keeps a following, but it owes an account of which biological powers matter. Block's split between access and phenomenal consciousness is widely used, even by people who doubt that the two really come apart. The phenomenal concept strategy is the physicalist's best answer to Mary and the zombies, and it is still contested. The meta-problem is accepted by almost everyone as a real question worth studying. And illusionism is a serious minority view with strong arguments, facing an objection many find decisive and a reply its defenders find decisive in turn.

If you want to go deeper, here is where to start. For a gentle overview, Ian Ravenscroft's Philosophy of Mind: A Beginner's Guide, from Oxford University Press, 2005, covers identity theory and functionalism clearly. Jaegwon Kim's Philosophy of Mind, third edition, 2010, is harder but thorough. David Chalmers edited an anthology called Philosophy of Mind: Classical and Contemporary Readings, 2002, which reprints many of the papers in this episode, so you can read Place, Smart, Putnam and Block in their own words. Place's "Is Consciousness a Brain Process?" from 1956 and Smart's "Sensations and Brain Processes" from 1959 are both short. Turing's "Computing Machinery and Intelligence" from 1950 is very readable and often funny. Searle's paper "Minds, Brains, and Programs" appeared in 1980 in the journal Behavioral and Brain Sciences, printed alongside many critics' replies. His short book Minds, Brains and Science, from 1984, began as radio lectures and is aimed at a general audience. For the illusionist side, read Dennett's Consciousness Explained from 1991, and the collection Frankish edited, Illusionism as a Theory of Consciousness, from 2017, which prints his paper with replies from friends and critics. And for the question that ties them together, Chalmers's paper "The Meta-Problem of Consciousness" appeared in the Journal of Consciousness Studies in 2018.

We started with a simple claim: pain is something your brain is doing. After everything in this episode, that claim is still standing. What remains open is whether "doing" is the whole of it.