24 May 1997 Stan, I think we have a miscommunication, which I may not easily able to remedy. It depends on the difference between a factual claim and a conceptual claim. This is familiar to philosophers, but most scientists most of the time don't have to pay attention to it. I was making a conceptual claim, but you apparently took me to be making a factual claim. I'll give some examples of conceptual claims: some, but not all, have factual implications. 1. When Frege produced his new logic-based analysis of the concepts of arithmetic (embraced and modified by Russell) he was not thereby claiming that any arithmetical proposition anyone had ever previously believed was false, e.g. 3 + 5 = 8. He merely claimed that they had not realised WHAT it was that they believed. I.e. he and Russell claimed that because the concepts had deep hidden complexities, this was a much more complex belief than anyone had realised and that it had deep links with logical notions such as predication, quantification, implication, conjunction, disjunction etc. And those philosophers of mathematics who disagreed with Frege and Russell, claiming that numerical concepts were distinct from logic (e.g. saying they were based on something like perceptual experiences of groups of objects) were not thereby claiming that any different set of mathematical propositions were true. Both sides agreed on the arithmetical FACTS, while disagreeing on the conceptual analysis. Well not quite all: the intuitionists thought that our concepts of number and continuity were so deeply different from what we had previously thought that even some of the things we previously thought true were false, e.g. that the reals are uncountable. But it isn't always the case that offering a new conceptual analysis implies any disagreement with what were previously thought to be facts. 2. Similarly when mathematicians in the last century first came up with a precise analyse of the very old and familiar notion of continuity, they were NOT thereby claiming that previous beliefs were false, e.g. they agreed that some motion was continuous and some not, or that some sets of numbers were continuous and some not, some functions continuous and some not. Rather they were claiming that the previously existing concept of continuity had some deep and previously unnoticed properties. (E.g. allowing a function to be continuous in an interval while nowhere differentiable - a possibility disputed by intuitionists.) (Even now there are subtle problems about what we really mean by continuity: the 19th century conception of a continuum raises some hard problems if you are a constructivist mathematician. I was claiming in my message about reduced precision classical mechanics that there is a concept of a continuum that involves a different topology from the normal concept, and which was sometimes the one implicitly used when classical mechanics was applied.) 3. Likewise, Einstein reflecting on the notion of simultaneity came to the conclusion that although it FELT to most people that the notion of two events anywhere in space being simultaneous was a simple and clear notion, e.g. something they understood from direct experience of simultaneity, in fact it was not at all simple and depended on some previously acknowledged assumptions about the possibility of sending signals between the events, etc. In most everyday contexts his new analysis had no new factual implications, but in special unfamiliar contexts it had profound new implications. 4. The analysis of causation is another example: how to analyse the notion of one thing causing another is still a deep philosophical problem, though in this case it is not clear whether any of our normal beliefs about what causes what will change as a result of finding an acceptable analysis. We may have direct experience of causation: e.g. a desire causing a decision, but it doesn't follow that we really understand what we MEAN by "cause". [end of examples] My (not yet finished) conceptual analysis of what it is to have a visual experience of a red patch likewise does not make any claim that we do not have any such experience. Of course we do. But it doesn't follow that we know how the concepts we use in saying that work. I think there are implicit complexities that we are not aware of but we can come to realise by asking a lot of pointed questions about what sort of additional experiences are potentially involved in seeing a red patch. (Wittgenstein's discussion in part II, section xi of his Philosophical Investigations, where he attempts to analyse various cases of "seeing as" is very relevant to this.) In all these cases, a claim is made that our previous concepts did not work as we thought they did. In some cases this makes a difference to whether old factual beliefs are true, in others not. In all cases the reply: "I know what I mean by number/continuity/simultaneity/cause etc. from direct experience of the phenomenon" is unacceptable. That's exactly what you would expect people to say about concepts that they don't really grasp even though they think they do. Summary: I am saying that our pre-existing concept of "having an experience" and many other concepts related to mental states and processes, do not work the way we intuitively think they do. I am not thereby saying that any of our commonly held BELIEFS about which experiences do and do not occur are false. It may be that some beliefs that some philosophers (and philosophical scientists) hold are false. But it doesn't follow that any *ordinary* *everyday* belief is false, such as that we are conscious of some things and not others, and that sometimes were are unconscious, and therefore conscious of nothing, at other times fully conscious/ The beliefs may be confused, vague, etc. But that's not the same as being false. Of course a new analysis may leave all the intermediate and fuzzy cases as unclear as ever, or it may point to new ways of revising and extending the old concepts so as to deal more satisfactorily with intermediate and fuzzy cases. [AS] > A crucial step, which you may not like, is that we do not know "from > inside" how our concepts work, though we THINK we do. > > So when we say things like: "Well, I KNOW that that's not what I mean by > subjective experience" we may be just wrong, because we don't know what > we mean by these apparently simple and clear but actually deep and > complex concepts. [SK] > I don't think anyone is claiming that we think we know how our concepts > work. Well, either you didn't understand that I was referring to conceptual knowledge as opposed to factual knowledge using the concepts, or you have not talked to the philosophers that I have, who reject the notion that there is any question of their being unclear what they *mean* by saying that a certain kind of robot will or will not have consciousness, or when they say that it is perfectly meaningful to ask whether our colour qualia can flip without the flip being detectable. Surely you have encountered a philosophical discussion where one person says such claims are meaningless or incoherent and the other person says he/she understands their meaning very well and they make perfectly clear sense? Certainly Searle used to argue as if we had a perfectly clear grasp of what it is to experience or understand something etc. (I don't know what he now thinks). Penrose has written like that (e.g. in The Emperor's new mind). And many philosophers think that the one thing that cannot come from any new scientific theory is a new analysis of old concepts of mind. I sometimes get the impression Keith holds that view. I don't. [SK] > We just KNOW that we have subjective experiences. Yes indeed. That's a factual claim I don't dispute. My claim is simply that you don't know what it is that you know. That doesn't mean it's false. I.e. what you claim may be true, but what it means may be confused in various ways. (E.g. ask yourself: is it clear what difference it would make if you left out the word "subjective". Would your meaning be any different? Could there be non-subjective experiences? Objective experiences?) It's also not clear what the word "just" adds. If the word "just" is meant to imply that knowing that you have subjective experience is a simple unanalysable fact, then that's a philosophical claim and that could be false. But if you leave out the word "just", and de-emphasise "KNOW", then there's nothing wrong with it. Similarly we (just) know that 3+5 = 8. Frege's claim was NOT that that was incorrect but that we did not know WHAT it was that we knew. We (just) know that near the end of that orchestral performance the cymbals clashed and the trumpet sounded simultaneously (or very nearly simultaneously) whilst the drums and violins did not play simultaneously but in succession. Indeed we do. But Einstein's claim is that we don't know WHAT it is that we know. Likewise I am not disputing that we know we have (subjective) experience. That is a FACT that we are all trying to explain. (Maybe not Dennett: sometimes I read his discussion of the intentional stance as implying that descriptions of mental states are not factual, merely pragmatically useful. I don't agree with that.) I don't deny that we know these facts. But I do deny that we know what it is that we know: the pre-theoretical concepts in all the above claims are just unclear and not properly understood by the people who use them. Human knowledge, even useful knowledge, can make use of unclear, deeply muddled concepts with hidden complexities. The concept of "cause" is one of the main examples. We could not survive without it it. Yet, to my knowledge, nobody has produced a satisfactory analysis of what we mean when we say X caused Y. (One of my former Sussex PhD students, Chris Taylor, produced a PhD thesis arguing that there wasn't one concept but a whole cluster of different concepts suited to slightly different contexts, where the differences can be described only using surprisingly intricate logical formulae. Unfortunately he never published his work, though I think it's available as a technical report from Sussex.) > ...The > experiments show that we have minimal access to what is going on in our > visual world (hardly any of it gets into any memory). But that doesn't in any > way diminish the vivid subjective world that I see. Who says it does? Dennett? Perhaps, but if he does I'd be surprised: he too has vivid subjective experiences. Isn't that the starting point of his discussion? I don't say anything diminishes what's vivid. Why should I? You must have misread me. I am merely trying to analyse what we mean when we say such things as a step towards understanding the relationship between such descriptions and the descriptions of the underlying physical machinery (which I claim is a relationship of implementation = supervenience.) > It is indeed an illusion to > think that I have access to much of it. Much of what? Your subjective experience? Your brain states? > I know Dennett also likes to call it an > illusion. If you would like to use the word illusion like that I can't stop you. This completely misses my point. I am not disputing the claim that we have vivid and clear visual experiences (and sometimes dim or unclear visual experiences, in mist or dim light or in peripheral vision). These are just everyday facts. I am asking what these claims *mean*. If you say you just know what they mean and there's no problem about how we understand them, then you are making the sort of philosophical claim I talk about, and which you say nobody makes. If you merely reassert that they are facts, then you are missing the point for I am not denying that they are facts. > I am confident that before long .... Maybe. I really don't know whether we are close to the answer or a thousand years away. I don't expect to be alive when the answers are clear. > ...we will understand the NCC I also don't know exactly what you mean by "understand the NCC". We may discover lots of correlations between neural phenomena and conscious phenomena without thereby learning very much about WHY those correlations exist. There are already lots of known correlations. But I assume that you are referring to more than the discovery of correlations. You are also talking about understanding WHY it is that when the neural mechanisms are in place and working well, consciousness exists. That's also what I am talking about. But first I want to clarify what the concepts are that are used in the second half of that assertion, whereas I have the impression you think the concepts don't need any clarification, you seem to think it is all perfectly clear and the only thing missing is more facts about what explains the phenomena. > ...not only of our > access consciousness, but also of our pheneomenal consciousness. (a) (b) And here I am not even sure there's any well defined distinction between (a) and (b), though there may appear to be because the concepts we use are muddled. That's not to deny that all the phenomena are real. > When that > happens we should be able to implement it in robots. Before then, I think it > will be difficult for you to write an article on virtual machines that will > convince people that you know how to give subjectivity to robots. Yes of course. I.e. if I have not shown how particular sorts of physical states (including animal brain states) produce mental phenomena then I will not be able to convince people that I can explain how particular sorts of physical states (including robot brain states) can produce mental phenomena. I hope I have not written anything that implies otherwise. I am trying to avoid saying incoherent things. (Some people will not be convinced by *any* argument, for they want to believe that something like their experiences can survive the total destruction of their bodies. Or they deeply dislike the idea that they are "mere" machines, whatever "mere" means. When abandoning a belief is very painful, even very good arguments can fail completely. Argument sometimes has to give way to therapy.) Aaron PS In case it's not clear, I was not agreeing nor disagreeing with the Frege/Russell conceptual claims. I think that our intuitive arithmetical concepts are overloaded: they have both logical and perceptual components and we switch between them often. Likewise our pre-theoretic notions of continuity, which have both a perceptual basis which does not directly imply anything about infinite precision, and additional complications which do. I suspect nobody has properly analysed the relationships. This would require a much deeper study of the growth of mathematical concepts in children, especially children who become mathematicians, than has ever been done. That's another research programme waiting to get off the ground. When we've designed our intelligent infant robots we can also explore how their grasp of arithmetic, continuity, simultaneity, causation, experience, etc. develops. Expect very great surprises.