In a recent post, I claimed that functionalism about consciousness implies dualism. Since most functionalists think their philosophy is an alternative to dualism, I’d better present an argument.
But before I go further, I’ll link to orthonormal’s series on dissolving the problem of “Mary’s Room”: Seeing Red: Dissolving Mary’s Room and Qualia, A Study of Scarlet: The Conscious Mental Graph, Nature: Red in Truth, and Qualia. Mary’s Room is one of many thought experiments bandied about by philosophers in their attempts to say whether or not colors (and other qualia) are a problem for materialism, and orthonormal presents a computational attempt to get around the problem which is a good representative of the functionalist style of thought. I won’t have anything to say about those articles at this stage (maybe in comments), but they can serve as an example of what I’m talking about.
Now, though it may antagonize some people, I think it is best to start off by stating my position plainly and bluntly, rather than starting with a neutral discussion of what functionalism is and how it works, and then seeking to work my way from there to the unpopular conclusion. I will stick to the example of color to make my points—apologies to blind and colorblind readers.
My fundamental thesis is that color manifestly does exist—there are such things as shades of green, shades of red, etc—and that it manifestly does not exist in any standard sort of physical ontology. In an arrangement of point particles in space, there are no shades of green present. This is obviously true, and it’s equally obvious for more complicated ontologies like fields, geometries, wavefunction multiverses, and so on. It’s even part of the history of physics; even Galileo distinguished between primary qualities like location and shape, and secondary qualities like color. Primary qualities are out there and objectively present in the external world, secondary qualities are only in us, and physics will only concern itself with primary qualities. The ontological world of physical theory is colorless. (We may call light of a certain wavelength green light or red light, but that is because it produces an experience of seeing green or seeing red, not because the light itself is green or red in the original sense of those words.) And what has happened due to the progress of the natural sciences is that we now say that experiences are in brains, and brains are made of atoms, and atoms are described by a physics which does not contain color. So the secondary qualities have vanished entirely from this picture of the world; there is no opportunity for them to exist within us, because we are made of exactly the same stuff as the external world.
Yet the “secondary qualities” are there. They’re all around us, in every experience. It really is this simple: colors exist in reality, they don’t exist in theory, therefore the theory needs to be augmented or it needs to be changed. Dualism is an augmentation. My speculations about quantum monads are supposed to pave the way for a change. But I won’t talk about that option here. Instead, I will try to talk about theories of consciousness which are meant to be compatible with physicalism—functionalism is one such theory.
Such a theory will necessarily present a candidate, however vague, for the physical correlate of an experience of color. One can then say that color exists without having to add anything to physics, because the color just is the proposed physical correlate. This doesn’t work because the situation hasn’t changed. If all you have are point particles whose only property is location, then individual particles do not have the property of being colored, nor do they have that property in conjunction. Identifying a physical correlate simply picks out a particular set of particles and says “there’s your experience of color”. But there’s still nothing there that is green or red. You may accustom yourself to thinking of a particular material event, a particular rearrangement of atoms in space, as being the color, but that’s just the power of habitual association at work. You are introducing into your concept of the event a property that is not inherently present in it.
It may be that one way people manage to avoid noticing this, is by an incomplete chain of thought. I might say: none of the objects in your physical theory are green. The happy materialist might say: but those aren’t the things which are truly green in the sense you care about; the things which are green are parts of experiences, not the external objects. I say: fine. But experiences have to exist, right? And you say that physics is everything. So that must mean that experiences are some sort of physical object, and so it will be just as impossible for them to be truly green, given the ontological primitives we have to work with. But for some reason, this further deduction isn’t made. Instead, it is accepted that objects in physical space aren’t really green, but the objects of experience exist in some other “space”, the space of subjective experience, and… it isn’t explicitly said that objects there can be truly green, but somehow this difference between physical space and subjective space seems to help people be dualists without actually noticing it.
It is true that color exists in this context—a subjective space. Color always exists as part of an “experience”. But physical ontology doesn’t contain subjective space or conscious experience any more than it does contain color. What it can contain, are state machines which are structurally isomorphic to these things. So here we can finally identify how a functionalist theory of consciousness works psychologically: You single out some state machines in your physical description of the brain (like the networks in orthonormal’s sequence of posts); in your imagination, you associate consciousness with certain states of such state machines, on the basis of structural isomorphism; and now you say, conscious states are those physical states. Subjective space is some neural topographic map, the subjectively experienced body is the sensorimotor homunculus, and so forth.
But if we stick to any standard notion of physical theory, all those brain parts still don’t have any of the properties they need. There’s no color there, there’s no other space there, there’s no observing agent. It’s all just large numbers of atoms in motion. No-one is home and nothing is happening to them.
Clearly it is some sort of progress to have discovered, in one’s physical picture of the world, the possibility of entities which are roughly isomorphic to experiences, colors, etc. But they are still not the same thing. Most of the modern turmoil of ideas about consciousness in philosophy and science is due to this gap—attempts to deny it, attempts to do without noticing it, attempts to force people to notice it. orthonormal’s sequence, for example, seems to be an attempt to exhibit a cognitive model for experiences and behaviors that you would expect if color exists, without having to suppose that color actually exists. If we were talking about a theoretical construct, this would be fine. We are under no obligation to believe that phlogiston exists, only to explain why people once talked about it.
But to extend this attitude to something that most of us are directly experiencing in almost every waking moment, is … how can I put this? It’s really something. I’d call it an act of intellectual desperation, except that people don’t seem to feel desperate when they do it. They are just patiently explaining, recapitulating and elaborating, some “aha” moment they had back in their past, when functionalism made sense to them. My thesis is certainly that this sense of insight, of having dissolved the problem, is an illusion. The genuineness of the isomorphism between conscious state and coarse-grained physical state, and the work of several generations of materialist thinkers to develop ways of speaking which smoothly promote this isomorphism to an identity, combine to provide the sense that no problem remains to be solved. But all you have to do is attend for a moment to experience itself, and then to compare that to the picture of billions of colorless atoms in intricate motion through space, to realize that this is still dualism.
I promised not to promote the monads, but I will say this. The way to avoid dualism is to first understand consciousness as it is in itself, without the presupposition of materialism. Observe the structure of its states and the dynamics of its passage. That is what phenomenology is about. Then, sketch out an ontology of what you have observed. It doesn’t have to contain everything in infinite detail, it can overlook some features. But I would say that at a minimum it needs to contain the triad of subject-object-aspect (which appears under various names in the history of philosophy). There are objects of awareness, they are being experienced within a common subjective space, and they are experienced in a certain aspect. Any theory of reality, whether or not it is materialist, must contain such an entity in order to be true.
The basic entity here is the experiencing subject. Conscious states are its states. And now we can begin to tackle the ontological status of state machines, as a candidate for the ontological category to which conscious beings belong.
State machines are abstracted descriptions. We say there’s a thing, it has a set of possible states; here are the allowed transitions between them, and the conditions under which those transitions occur. Specify all that and we have specified a state machine. We don’t care about why those are the states or why the transitions occur; those are irrelevant details.
A very simple state machine might be denoted by the state transition network “1<->2”. There’s a state labeled 1 and another state labeled 2. If the machine is in state 1, it proceeds to state 2, and the reverse is also true. This state machine is realized wherever you have something that oscillates between two states without stopping in either. First the earth is close to the sun, then it is far from the sun, then it is close again… The Earth in its orbit instantiates the state machine “1<->2″. I get involved with Less Wrong, then I quit for a while, then I come back… My Internet habits also instantiate the state machine “1<->2”.
A computer program is exactly like this, a state machine of great complexity (and usually its state transition rules contain some dependence on external conditions, like user input) which has been physically instantiated for use. But one cannot claim that its states have any intrinsic meaning, any more than I can claim that the state 1 in the oscillating state machine is intrinsically about the earth being close to the sun. This is not true, even if I write down the state transition network in the form “CloseToTheSun<->FarFromTheSun”.
This is another ontological deficiency of functionalism. Mental states have meanings, thoughts are always about something, and what they are about is not the result of convention or of the needs of external users. This is yet another clue that the ontological status of conscious states is special, that their “substance” matters to what they are. Of course, this is a challenge to the philosophy which says that a detailed enough simulation of a brain will create a conscious person, regardless of the computational substrate. The only reason people believe this, is because they believe the brain itself is not a special substrate. But this is a judgment made on the basis of science that is still at a highly incomplete stage, and certainly I expect science to tell us something different by the time it’s finished with the brain. The ontological problems of functionalism provide a strong apriori reason for this expectation.
What is more challenging is to form a conception of the elementary parts and relations that could form the basis of an alternative ontology. But we have to do this, and the impetus has to come from a phenomenological ontology of consciousness that is as precise as possible. Fortunately, a great start was made on this about 100 years ago, in the heyday of phenomenology as a philosophical movement.
A conscious mind is a state machine, in the sense that it has states and transitions between them. The states also have structure, because conscious experiences do have parts. But the ontological ties that combine those parts into the whole are poorly apprehended by our current concepts. When we try to reduce them to nothing but causal coupling or to the proximity in space of presumed physical correlates of those parts, we are, I believe, getting it wrong. Clearly cause and effect operates in the realm of consciousness, but it will take great care to state precisely and correctly the nature of the things which are interacting and the ways in which they do so. Consider the ability to tell apart different shades of color. It’s not just that the colors are there; we know that they are there, and we are able to tell them apart. This implies a certain amount of causal structure. But the perilous step is to focus only on that causal structure, detach it from considerations of how things appear to be in themselves, and instead say “state machine, neurons doing computations, details interesting but not crucial to my understanding of reality”. Somehow, in trying to understand conscious cognition, we must remain in touch with the ontology of consciousness as partially revealed in consciousness itself. The things which do the conscious computing must be things with the properties that we see in front of us, the properties of the objects of experience, such as color.
You know, color—authentic original color—has been banished from physical ontology for so long, that it sounds a little mad to say that there might be a physical entity which is actually green. But there has to be such an entity, whether or not you call it physical. Such an entity will always be embedded in a larger conscious experience, and that conscious experience will be embedded in a conscious being, like you. So we have plenty of clues to the true ontology; the clues are right in front of us; we’re subjectively made of these clues. And we will not truly figure things out, unless we remain insistent that these inconvenient realities are in fact real.
Does functionalism imply dualism?
This post follows on from Personal research update, and is followed by State your physical explanation of experienced color.
In a recent post, I claimed that functionalism about consciousness implies dualism. Since most functionalists think their philosophy is an alternative to dualism, I’d better present an argument.
But before I go further, I’ll link to orthonormal’s series on dissolving the problem of “Mary’s Room”: Seeing Red: Dissolving Mary’s Room and Qualia, A Study of Scarlet: The Conscious Mental Graph, Nature: Red in Truth, and Qualia. Mary’s Room is one of many thought experiments bandied about by philosophers in their attempts to say whether or not colors (and other qualia) are a problem for materialism, and orthonormal presents a computational attempt to get around the problem which is a good representative of the functionalist style of thought. I won’t have anything to say about those articles at this stage (maybe in comments), but they can serve as an example of what I’m talking about.
Now, though it may antagonize some people, I think it is best to start off by stating my position plainly and bluntly, rather than starting with a neutral discussion of what functionalism is and how it works, and then seeking to work my way from there to the unpopular conclusion. I will stick to the example of color to make my points—apologies to blind and colorblind readers.
My fundamental thesis is that color manifestly does exist—there are such things as shades of green, shades of red, etc—and that it manifestly does not exist in any standard sort of physical ontology. In an arrangement of point particles in space, there are no shades of green present. This is obviously true, and it’s equally obvious for more complicated ontologies like fields, geometries, wavefunction multiverses, and so on. It’s even part of the history of physics; even Galileo distinguished between primary qualities like location and shape, and secondary qualities like color. Primary qualities are out there and objectively present in the external world, secondary qualities are only in us, and physics will only concern itself with primary qualities. The ontological world of physical theory is colorless. (We may call light of a certain wavelength green light or red light, but that is because it produces an experience of seeing green or seeing red, not because the light itself is green or red in the original sense of those words.) And what has happened due to the progress of the natural sciences is that we now say that experiences are in brains, and brains are made of atoms, and atoms are described by a physics which does not contain color. So the secondary qualities have vanished entirely from this picture of the world; there is no opportunity for them to exist within us, because we are made of exactly the same stuff as the external world.
Yet the “secondary qualities” are there. They’re all around us, in every experience. It really is this simple: colors exist in reality, they don’t exist in theory, therefore the theory needs to be augmented or it needs to be changed. Dualism is an augmentation. My speculations about quantum monads are supposed to pave the way for a change. But I won’t talk about that option here. Instead, I will try to talk about theories of consciousness which are meant to be compatible with physicalism—functionalism is one such theory.
Such a theory will necessarily present a candidate, however vague, for the physical correlate of an experience of color. One can then say that color exists without having to add anything to physics, because the color just is the proposed physical correlate. This doesn’t work because the situation hasn’t changed. If all you have are point particles whose only property is location, then individual particles do not have the property of being colored, nor do they have that property in conjunction. Identifying a physical correlate simply picks out a particular set of particles and says “there’s your experience of color”. But there’s still nothing there that is green or red. You may accustom yourself to thinking of a particular material event, a particular rearrangement of atoms in space, as being the color, but that’s just the power of habitual association at work. You are introducing into your concept of the event a property that is not inherently present in it.
It may be that one way people manage to avoid noticing this, is by an incomplete chain of thought. I might say: none of the objects in your physical theory are green. The happy materialist might say: but those aren’t the things which are truly green in the sense you care about; the things which are green are parts of experiences, not the external objects. I say: fine. But experiences have to exist, right? And you say that physics is everything. So that must mean that experiences are some sort of physical object, and so it will be just as impossible for them to be truly green, given the ontological primitives we have to work with. But for some reason, this further deduction isn’t made. Instead, it is accepted that objects in physical space aren’t really green, but the objects of experience exist in some other “space”, the space of subjective experience, and… it isn’t explicitly said that objects there can be truly green, but somehow this difference between physical space and subjective space seems to help people be dualists without actually noticing it.
It is true that color exists in this context—a subjective space. Color always exists as part of an “experience”. But physical ontology doesn’t contain subjective space or conscious experience any more than it does contain color. What it can contain, are state machines which are structurally isomorphic to these things. So here we can finally identify how a functionalist theory of consciousness works psychologically: You single out some state machines in your physical description of the brain (like the networks in orthonormal’s sequence of posts); in your imagination, you associate consciousness with certain states of such state machines, on the basis of structural isomorphism; and now you say, conscious states are those physical states. Subjective space is some neural topographic map, the subjectively experienced body is the sensorimotor homunculus, and so forth.
But if we stick to any standard notion of physical theory, all those brain parts still don’t have any of the properties they need. There’s no color there, there’s no other space there, there’s no observing agent. It’s all just large numbers of atoms in motion. No-one is home and nothing is happening to them.
Clearly it is some sort of progress to have discovered, in one’s physical picture of the world, the possibility of entities which are roughly isomorphic to experiences, colors, etc. But they are still not the same thing. Most of the modern turmoil of ideas about consciousness in philosophy and science is due to this gap—attempts to deny it, attempts to do without noticing it, attempts to force people to notice it. orthonormal’s sequence, for example, seems to be an attempt to exhibit a cognitive model for experiences and behaviors that you would expect if color exists, without having to suppose that color actually exists. If we were talking about a theoretical construct, this would be fine. We are under no obligation to believe that phlogiston exists, only to explain why people once talked about it.
But to extend this attitude to something that most of us are directly experiencing in almost every waking moment, is … how can I put this? It’s really something. I’d call it an act of intellectual desperation, except that people don’t seem to feel desperate when they do it. They are just patiently explaining, recapitulating and elaborating, some “aha” moment they had back in their past, when functionalism made sense to them. My thesis is certainly that this sense of insight, of having dissolved the problem, is an illusion. The genuineness of the isomorphism between conscious state and coarse-grained physical state, and the work of several generations of materialist thinkers to develop ways of speaking which smoothly promote this isomorphism to an identity, combine to provide the sense that no problem remains to be solved. But all you have to do is attend for a moment to experience itself, and then to compare that to the picture of billions of colorless atoms in intricate motion through space, to realize that this is still dualism.
I promised not to promote the monads, but I will say this. The way to avoid dualism is to first understand consciousness as it is in itself, without the presupposition of materialism. Observe the structure of its states and the dynamics of its passage. That is what phenomenology is about. Then, sketch out an ontology of what you have observed. It doesn’t have to contain everything in infinite detail, it can overlook some features. But I would say that at a minimum it needs to contain the triad of subject-object-aspect (which appears under various names in the history of philosophy). There are objects of awareness, they are being experienced within a common subjective space, and they are experienced in a certain aspect. Any theory of reality, whether or not it is materialist, must contain such an entity in order to be true.
The basic entity here is the experiencing subject. Conscious states are its states. And now we can begin to tackle the ontological status of state machines, as a candidate for the ontological category to which conscious beings belong.
State machines are abstracted descriptions. We say there’s a thing, it has a set of possible states; here are the allowed transitions between them, and the conditions under which those transitions occur. Specify all that and we have specified a state machine. We don’t care about why those are the states or why the transitions occur; those are irrelevant details.
A very simple state machine might be denoted by the state transition network “1<->2”. There’s a state labeled 1 and another state labeled 2. If the machine is in state 1, it proceeds to state 2, and the reverse is also true. This state machine is realized wherever you have something that oscillates between two states without stopping in either. First the earth is close to the sun, then it is far from the sun, then it is close again… The Earth in its orbit instantiates the state machine “1<->2″. I get involved with Less Wrong, then I quit for a while, then I come back… My Internet habits also instantiate the state machine “1<->2”.
A computer program is exactly like this, a state machine of great complexity (and usually its state transition rules contain some dependence on external conditions, like user input) which has been physically instantiated for use. But one cannot claim that its states have any intrinsic meaning, any more than I can claim that the state 1 in the oscillating state machine is intrinsically about the earth being close to the sun. This is not true, even if I write down the state transition network in the form “CloseToTheSun<->FarFromTheSun”.
This is another ontological deficiency of functionalism. Mental states have meanings, thoughts are always about something, and what they are about is not the result of convention or of the needs of external users. This is yet another clue that the ontological status of conscious states is special, that their “substance” matters to what they are. Of course, this is a challenge to the philosophy which says that a detailed enough simulation of a brain will create a conscious person, regardless of the computational substrate. The only reason people believe this, is because they believe the brain itself is not a special substrate. But this is a judgment made on the basis of science that is still at a highly incomplete stage, and certainly I expect science to tell us something different by the time it’s finished with the brain. The ontological problems of functionalism provide a strong apriori reason for this expectation.
What is more challenging is to form a conception of the elementary parts and relations that could form the basis of an alternative ontology. But we have to do this, and the impetus has to come from a phenomenological ontology of consciousness that is as precise as possible. Fortunately, a great start was made on this about 100 years ago, in the heyday of phenomenology as a philosophical movement.
A conscious mind is a state machine, in the sense that it has states and transitions between them. The states also have structure, because conscious experiences do have parts. But the ontological ties that combine those parts into the whole are poorly apprehended by our current concepts. When we try to reduce them to nothing but causal coupling or to the proximity in space of presumed physical correlates of those parts, we are, I believe, getting it wrong. Clearly cause and effect operates in the realm of consciousness, but it will take great care to state precisely and correctly the nature of the things which are interacting and the ways in which they do so. Consider the ability to tell apart different shades of color. It’s not just that the colors are there; we know that they are there, and we are able to tell them apart. This implies a certain amount of causal structure. But the perilous step is to focus only on that causal structure, detach it from considerations of how things appear to be in themselves, and instead say “state machine, neurons doing computations, details interesting but not crucial to my understanding of reality”. Somehow, in trying to understand conscious cognition, we must remain in touch with the ontology of consciousness as partially revealed in consciousness itself. The things which do the conscious computing must be things with the properties that we see in front of us, the properties of the objects of experience, such as color.
You know, color—authentic original color—has been banished from physical ontology for so long, that it sounds a little mad to say that there might be a physical entity which is actually green. But there has to be such an entity, whether or not you call it physical. Such an entity will always be embedded in a larger conscious experience, and that conscious experience will be embedded in a conscious being, like you. So we have plenty of clues to the true ontology; the clues are right in front of us; we’re subjectively made of these clues. And we will not truly figure things out, unless we remain insistent that these inconvenient realities are in fact real.