I am a physicalist, at least in this sense: I believe (though I know I cannot prove) that the mind is a result of purely physical processes occurring within the brain, and, furthermore, that how this works could, at least in principle, be explained within our current understanding of the physical world, requiring physics no more fundamental than is needed for biochemistry.
Nevertheless, I suppose that Mary will learn something new when she sees red, and I do not do so with a shrug - on the contrary, I feel that asserting she would know what it's like prior to seeing red requires assuming the brain has certain capabilities that are implausible and lacking empirical evidence.
Physicalism (or at least my flavor of it) effectively implies that coming to know what it is like to see red requires some physical change in the brain - changes which can be traced back to the stimulation of red receptors in the retina, and while the specifics are probably different in detail from one person to the next, are similar at some appropriate level of functional abstraction. Under the premise that Mary knows every relevant physical fact, we must assume that she knows (e.g. through scanning and modeling) exactly what changes will occur when she sees any particular red-containing scene, but there's no reason to believe she has the ability to bring about those changes through mental effort alone - and, according to the physicalist premise, if the changes do not occur, she will not know what seeing red is like. Consequently, any physicalist who claims that Mary will learn from her studies what seeing red is like, carries the burden of providing some justification for assuming that she can use her knowledge to bring about the requisite neurophysical changes.
I am saying that there is no good evidence, in either personal experience or neuroscience, that we would be able to bring about the necessary changes at the neuron and synapse level through mental effort alone, unaided by tools - but what about the examples you give? Even if Hume was right, interpolating between very similar experiences seems considerably less demanding than synthesizing them from scratch, and it seems (at least to me) unlikely that someone who had only seen blues and greens could come to know what seeing orange and red is like, yet Mary is expected to do even more with less. Similarly, if you do not already know what red is like, it seems unlikely that you can know, except in abstract terms, what the Japanese flag looks like. As for Crane's argument, if you had never experienced pain (and there are some people who do not), I suppose you could, on hearing other peoples' talk about pain, deduce that they have experienced something you have not, but that would not tell you what it is like. Finally, Swamp Mary is not coming to know what seeing red is like through her own mental effort.
How, then, can I dismiss the threat to physicalism that Jackson's thought experiment allegedly poses? Basically, it is for the same reason as I object to the thesis that Mary will know what seeing red is like prior to doing so: no matter what she knows, she cannot bring about the requisite physical changes. We understand well enough how various complex and indisputably physical systems work - the weather and biological reproduction, for example - yet no-one expects that learning this knowledge will create a thunderstorm or get you pregnant. Why should anyone think the mind is any different? I think Churchland answered this question in an unfortunately hard-to-find article with the title "Knowing Qualia: a Reply to Jackson" way back in 1989: the argument equivocates over the meaning of the verb "to know." While we talk of knowing what it is like, it is not knowledge in the sense of having a justified belief in certain propositions (if it were, how come we cannot say what these propositions are? And why would anyone assume that Mary could not learn them from her studies?)
Here's an analogy which might serve as an intuition pump for this view: suppose a safe-cracker has broken into a bank one night, and she has X Rays of the vault, so detailed that she can deduce the combination from them. Nevertheless, she still cannot open it if it contains a well-designed time lock that is preventing anyone from doing so until tomorrow morning. She has all the relevant physical knowledge, but not the means to use it.
of course every physicalist will grant that any transition from ignorance to knowledge will require physical changes. but what *positive* reason is there for a physicalist to believe that there’s only one way to know what it’s like to see red, especially in light of the fact that this is not a general truth about knowing what it’s like , as you grant by agreeing with me and Hume?
I thought about it a bit and I now think that I just have certain assumptions about human brains in particular. If she was an alien I would have no problem thinking that she could know what it's like to see red from just learning about it from books.
I guess it's something like that you need some processing in the visual cortex to trigger the relevant changes in the brain and that can't be achieved just by learning propositional knowledge.
At least if the changes are big enough, like seeing red for the first time, I do of course agree that we can assemble the Japanese flag from other things we've already seen.
Thanks. The main problem with this is that we have no neuroscientific evidence that the human brain actually works in the way you describe. We don’t know which part of the brain, if any, is the “knowing what it’s like” part and why some signals to it but not others would trigger knowledge. This is just the intuition that Mary must learn something new re-stated in a neuro sounding vocabulary, not a report of neuroscientific evidence for the truth of the intuition.
You're right. As a bit of context, before I thought that even other kinds of beings wouldn't be able to what it's like without leaving the room, so I think I'm closer to your view than I was before at least.
Maybe I also misunderstand what "could" means in this context. There is an epistemic possibility that she could know it, it's not that some weird metaphysical stuff is going on, it just depends on contingent facts about the human brain.
If the necessary connections exist in the typical human brain, then she will know it, otherwise she won't. Knowledge is just a brain state but whether or not reading books brings about the necessary changes in brain states in a human is just unknown to us, it's possible in theory though.
If I'm not mistaken, the only aspect of Hume's philosophy that has come up so far is his premise that a person could interpolate what it is like to see a shade of blue they have never seen before, if presented with a sequence of similar colors, ordered by hue, on either side of the missing one. With respect to that premise, I wrote "even if Hume was right [about that]..." I'm not convinced that we have that ability, but I'm willing to accept, at least for the sake of argument, that some people do - but even then, it seems highly implausible to me that this ability can be extrapolated to the point where it defeats Jackson's argument, which would require Mary being able to know what every color experience she could possibly have would be like, not by interpolation between colors she has experienced, but purely by deduction from a collection of propositional knowledge about color vision, no matter how complete that collection is.
I'm not sure how I gave the impression that I think there's only one way to know what it is like to see red, but I'm guessing it was where I wrote "...and while the specifics are probably different in detail from one person to the next, [the physical changes] are similar at some appropriate level of functional abstraction." What I did not make clear here is how loose and how abstract that similarity might be: I would not be surprised if every experience of a red color is as unique as a snowflake, yet I feel that there must be some level of similarity between a given person's red experiences for that individual to have a coherent concept of redness, and also a probably weaker similarity between different individuals, given that we are using the concept of what it is like to see red in our discussion without feeling that the other person is writing incoherent gibberish. I regard this ability to have a dialogue about what it is like to see colors as a positive reason (albeit rather weak) for surmising that there is some commonality to having sensory experiences of a given type. I will also put forward the fact that our neural architecture is at least roughed out genetically as another weak reason for adopting this view.
As far as I can see, the positions I put forward in my previous post are not predicated on there being only one way to know what it's like to see red.
I think you’ve done an excellent job of describing something alleged to be indescribable. upon having any of those experiences, I would recognize them as what you were talking about.
I think this is a response to my 2 + 2 comment above? (If not I apologize - it is a little hard to follow the threading in substack.)
If so, I'm not sure it's correct that you would be able to recognize the type of experience described while having it. Part of the case, I think, would involve positing (and specifying) some sort of cognitively-impenetrable mechanism by which you are prevented from thinking about the meaning of 2 + 2 while having the experience of believing that 2 + 2 = 5 (or a functionally analogous belief about an epistemic relationship of acquaintance...) Presumably that mechanism would also block you from recognizing that that's what you're doing while in that state?
someone couldn’t *know* that 2+2 isn’t 4 but someone could think about 2+2 without also thinking that it sums to 4. a buddy of mine was once hypnotized by a stage hypnotist to forget the number 3. he seemed, when sent back to the table for us to mess with him, fully cognizant of all numerical facts and relations that you’d expect from someone with a degree in electrical engineering, but until the hypnotist snapped him out if it, it was like threeness was totally erased from his mind. i believe this story in part because once while on acid i became fully convinced i knew german and watched a whole german language tv show convinced i understood everything they were saying (in retrospect i obviously did not). i have believed contradictions before. i know what that’s like.
If your buddy was hypnotized again to forget the number 3, presumably under hypnosis he wouldn't think "oh right, I know what it's like to forget 3". Instead, when you were messing with him, I assume his reaction would still just be "huh!?"
Maybe the KWIL distinction goes the other way from how it's usually framed. In this example, a distinction between KWIL as knowing or having a memory *that*, in the past, he was hypnotized to forget 3, vs. *not* knowing / not recognizing that "this is what it's like to forget 3" while actually under hypnosis.
In this view, I guess there would still be an epistemic gap between the two, though the gap would not involve a gain of knowledge like proponents of the KA suggest, but instead a sort of active, introspective suppression* of knowledge under certain circumstances (i.e., not just not thinking about 3, but being prevented, by some mechanism, from thinking about 3 under circumstances where you should be able to.) So as you say, KWIL just is descriptive knowledge, but then the subject's epistemic situation during experience would be something *less than* KWIL.
(I'm not actually suggesting repeating the experiment BTW, just pondering. We all did things in our college / grad school days that makes great "dad lore" - as my son calls it - but which we wouldn't necessarily want to repeat, even for the sake of philosophy or consciousness research :-)
* "Active, introspective suppression" in the above is meant as a hand wavy gesture to try to distinguish this from the personal / sub-personal distinction.
Yes, good. The “it” is this statement (which I personally believe to be false, fwiw): “There exists at least one experience type, for example, visual experiences of red by color-sighted humans, such that while one can come to know what it is like by undergoing the experience itself, and maybe even by being a swamp being or futuristic surgical patient, one cannot come to know what it is like simply by understanding a description, no matter what the description is.” I would be genuinely interested in seeing a physicalist construct an argument that had IT as its conclusion that was based on something other than appeals to its intuitiveness or the alleged lack of arguments for its falsehood. What positive reasons are there, available to a physicalist, for believing IT? I genuinely have no recollection of having yet encountered one.
Here's a possible example of a type of experience that one could not come to know what it is like by way of a description.
Let's say that Orwell's account in "1984" of what Winston Smith experiences after his torture in the Ministry of Love - i.e., a belief that 2 + 2 = 5 - describes a real psychological possibility. That is, after what he has gone through, Winston reaches a state where, in some sense, he actually believes that 2 + 2 = 5.
Now consider the pre-torment Winston. Any description of what it's like to believe that 2 + 2 = 5 would have to include "and you can't think about the meaning of 2 + 2." To follow these instructions, Winston would have to think about what it's like to believe something that requires you not to think about it, which would be impossible. In this analogy, when Mary sees the apple, and it seems to her that she is directly acquainted with the redness of the apple, maybe the "seeming to be acquainted" is - somehow - like "seeming that 2 + 2 = 5" (though hopefully without subjecting Mary to actual torture in Room 101...after all, she's had a hard enough time already.)
This would require developing an account of "seeming to be acquainted" that contradicts "thinking about seeming to be acquainted" in a similar way, but the payoff would be an account of KWIL that does not involve Mary gaining any new knowledge when she sees the apple while also - in a backhand way - validating the intuition that she couldn't have known what it's like from her black and white room. This would also, I think, be distinct from the view that in seeing red Mary gains a new ability / knowledge how. After all, while gaining the ability to ride a bike by practicing riding a bike, there is no problem with also thinking about what you are doing in riding the bike.
This seems very odd: the post of yours that I am replying to here, posted at 14:56Z on Tuesday, came after my last post prior to this one, posted at 11:50Z on Tuesday [1], and appears to be replying to it, yet, at least in my browsers, it is presented as a reply to a post I made on Saturday. Furthermore, the post that I am replying to here seems to consist primarily of what looks like a quote, except that the text in this 'quote' does not seem to be found in any of my responses to your article, and nor does it appear to be from any other response to your article, either. If the post I am replying to here *is* intended to be a response to my most recent post prior to this one, and is quoting something that has been written in response to your article, could you provide me with a link to the post being quoted, so that I can read it in context before replying?
ok, so you grant that it’s not necessary to have had an experience of red in order to come to know what it’s like to see red? Yes? Regarding Hume, you agree with me and Hume that we can know what it’s like to see the “missing shade” of blue in advance of actually seeing that shade of blue. One way to put the key points here is by reference to this statement, S: “For all x, if one knows what it is like to experience x, then on must either have already experienced x or one is currently experiencing x.” I think S is false. I guess you do too, but I’m not sure exactly which cases you agree with me are counterexamples to S. I think the japanese flag and hume’s blue are obvious counterexamples. Another counterexample is a perfect forgery of the Mona Lisa. I further think the physicalist should say red is a counterexample. You seem at least sometimes to disagree about that last one. But I don’t know. You tell me.
Sure - I will go through my positions one case at a time:
1) I think Swamp Mary is possible, even though (like Botzmann Brains) beyond-astronomically unlikely. She provides no support for the premise that Mary will KWIL prior to her release, however, as she has not arrived at the state of knowing what it is like to see red as a consequence of studying propositional knowledge.
2) With regard to Crane's claim (at least as you have made use of it here), I assume that if you have experienced pain - *and can remember something about what it is like* - then you will generally be able to tell when you are not in pain (though people who have profound dementia may lack the self-awareness to recognize whether or not they are in pain or hungry.) As this depends on having had the experience, however, I do not see how it might provide any support for the premise that Mary will KWIL, prior to her release, as a consequence of reading anything.
3) The Japanese flag case appears to me to be very much like Crane's argument, in that it depends on a prior experience of the right type: I suppose that a person who has not experienced redness will learn something about what looking at the flag is like on first seeing it, no matter how thoroughly it has been described to them beforehand.
4) Hume's color-interpolation hypothesis seems plausible to me, and though I'm not entirely convinced that it is correct, I have said that I will assume it is for the purposes of this discussion. Even then, it provides no support to the premise that Mary will KWIL prior to her release, as it depends on knowing what the bordering colors are like. As far as I can see, there's no reason to think that, just because Mary could do this once she has seen these colors, she could arrive at a state of knowing what those colors are like solely as a consequence of studying propositional knowledge.
5) I agree that S is false, but I do not see how its falsity provides any affirmative support for the notion that Mary could come to know what it is like to see red solely as a consequence of studying propositional knowledge of any sort. My own objection to the knowledge argument, like several others, neither depends on nor implies proposition S.
Ok, great. I just needed the part in 5 where you said that S is false. I didn’t learn anything I didn’t already know about you from the rest. Now I’ll repeat my question I earlier asked you, and didn’t notice any answer, but will this time avoid the befuddling and distracting phrase “only one way”. WHY must Mary experience red in order to know what it is like to see red? What set of premises, consistent with physicalism, collectively entail that Mary must experience red in order to know what it is like to see red?
On rereading your article, I think I now see where you are going with this line of questioning: the knowledge argument never explicitly rules out the possibility that Mary could come to KWIL by some means other than experiencing colors. Instead, it tacitly assumes that proposition S is true (without this tacit premise - which is not the only one here - or something equivalent, the premise that Mary learns something on her release lacks justification.)
In this view, your objection (if I'm understanding it correctly) is quite similar to Daniel Dennett's first response to the knowledge argument, which was to say that we cannot imagine, let alone deduce, what knowing all the relevant scientific knowledge would entail, and so the knowledge argument's proponents are merely relying on their probably-biased intuitions when they assume Mary will learn something about colors when she first experiences them (this is the argument in which he introduces the blue banana trick.) The philosopher and knowledge-argument proponent Torin Alter concedes that this is a legitimate objection to it.
The problem for Dennett was that those who were on the fence generally found proposition S to be very plausible, and I think you will find the same. In your case, all but one of your four arguments against S depend on bootstrapping from some initial KWIL, and the fourth (Swamp Mary) requires an extremely improbable event, while the knowledge argument supposes that Mary will always (or at least almost always) learn something about seeing colors on first experiencing them.
Dennett later turned to the argument found in 'What RoboMary Knows', in which (at least in my reading of it) he argues that the widespread acceptance of S (at least under the 'rules' of what Mary can and cannot do in isolation, which is a topic in its own right [1]) is merely on account of physical limitations, imposed by our brain's architecture, on what Mary can do with her knowledge of physics.
Note: I wrote this some hours before my other reply to the parent post.
So you needed to hear that S is false? Well, I am happy to have been able to do that for you, and maybe I will have further opportunities to say things you need to hear!
My response to your question "why must Mary experience red in order to know what it is like to see red?" is that she doesn't (an answer that should come as no surprise, given that I have just said S is false.) For example, given so-far hypothetical tools for selecting and modifying the physical state of individual neurons and synapses without doing any damage, and the knowledge to know what needs to be changed, I suppose that one would be able to create false memories in Mary's brain of seeing things in color, and from then on she would know what seeing those colors is like to the same extent as she would if she had actually had the experiences that the fabricated memories seem to be of.
On the other hand, I reject, as an alternative to experience, the notion that Mary could come to know what it is like to see red solely by learning a body of propositional knowledge.
I can only vaguely guess where you are going with this line of inquiry, and I am eager to see where you take it next.
Since you seem to think it’s important, i’ll focus on this. I am talking about language because someone else made a claim about different senses of “knowledge” . If people want to make claims about senses of words, they should appreciate that they are talking about language, and further, that how one determines whether a word has multiple senses has been a subject of intense and serious investigation for many years by linguists and philosophers of language. It’s not something serious people just pull out of their …hats. If my dialectical opponents don’t want to talk about language, then they shouldn’t have brought it up in the first place. if the language-phobe wants to conduct the investigation of the mary argument solely with minimal explicit mention of language and logical entailments of words and phrases, they would still at least need non-laughable operational definitions of terms like “knowing what it’s like to see red” and “seeing red”. I won’t say that it is impossible to do that, but no one yet has. And I venture that it would be exceedingly difficult, as is evidenced by the many years of research that has gone into investigating the closely related “Molyneux’s question” scientifically without anyone turning up any settled consensus on the answer. (Besides looking up “Molyneux’s question,” if you’re interested also look up “Molyneux’s babies”.)
Thanks - given that you can speak authoritatively on the matter, it is helpful for me to have your confirmation that no-one has yet given operational definitions of terms like "knowing what it is like to see red", let alone one which would justify seeing the knowledge argument as defeating physicalism.
I think Churchland achieves that, but that's where the tautology of needing the brain-state of KWIL comes in.
Knowing in the usual sense: having the desired physical state, the KWIL-state.
Knowing in the pre-release sense: having good knowledge *about* the KWIL-state, in the form of circuit diagrams and neural-net concepts and the physics of the retina, and so on.
But one of them is *about* the substrate of the other. They have entirely different contents, and entirely different representational relationships to the out-of-head environmental property of redness.
Does one ultimately imply the facts entailed by the other? Yes, of course, if physicalism is true. But KWIL barely counts as factual; it cannot be expressed in propositional terms.
It's a bit like comparing GPT4's internal representation of "redness" with a Word document listing the relevant neural weights. One entails the other. (And of course I am not suggesting an imagistic internal representation, or anything like a redness quale.)
If we had some deficient version of GPT4 that had been specially trained with a redacted corpus, and it had never encountered redness and its semantic associations in text, we would not expect that feeding the Word document to the deficient GPT4 would ever substitute for the missing training. GPT4 doesn't have the right sort of self-manipulation to update itself like that, for fairly obvious reasons, but humans also lack this sort of access to their own substrate.
Obviously, much more discussion is required to flesh out what types of knowledge we expect to be inter-convertible in humans, but almost everything we can say about these two knowledge states suggest that they have a very different set of epistemic properties. That alone is enough to consider them distinct types of knowledge, and their fundamental neuroanatomical separation reflects that the difference between their epistemic properties is best explained in physical terms, without reference to the ontological relationship between their contents.
And of course, this is all only an assertion if the full discussion is not attached.
But each comment has to stop somewhere, with important things still left unsaid.
A difference in what they’re about doesn’t entail a difference in sense of ‘know’. I can know what time it is and know where I put my lunchbox. It doesn’t follow, or even suggest, that I’m knowing in two different senses of “know.”
Regarding physicalism, my take has long been that the setup to the thought experiment assumes its purported conclusion, non-physicalism, by implying that Mary can have all the physical information and yet still subsequently learn something new. If we remove that assumption, then it becomes an assertion that Mary can have all the information but learn new information after seeing red, a contradiction.
The question of whether it's possible for her to KWIL with all the possible third party information is a more interesting question. Swamp Mary is a nice move! It seems similar to Daniel Dennett's RoboMary putting her brain in the same state as if she'd seen red. In both cases, KWIL is achieved without the experience of seeing being in the causal history. Nice!
I also think it's worth noting that even a Mary only with contemporary knowledge could guess a lot about what she'd see. She'd know that reds blend into oranges, which blend into yellows, greens, cyans, blues, etc. She'd expect red to jump out more than green or blue. And she'd be expecting it as a commonality for roses, blood, and ripe strawberries and a strong distinction from the sky, grass, or ripe bananas.
I think most of our intuition about how she'd react seems more based on how we'd react. But without her knowledge, out intuition doesn't seem useful here.
Thank you! My hunch about where people get their broken intuition about needing to see red is that they over-generalize a folky ocular epistemology: things that are farther away are harder to see and thus know about, so if they are brought so close as to be all the way “in here” [points to mind with finger] then that’s the exemplar of ultimate knowledge: having the thing known all up in it (the knower).
Interesting post, but I'm one of those physicalists who disagrees with you.
I think we need to distinguish between knowing what redness looks like in the usual way, and knowing what it looks like in the bland scientific way that Mary can achieve. Both of these concepts can point at the same underlying ontological truth, but transitioning between them is a physical process that can indeed be impossible.
I agree that swamp-Mary breaks the link between seeing and KWIL, but she is also astronomically implausible. We could break that link in several other ways: pre-release Mary could be handy with a nano-scalpel, and just give herself KWIL for redness. Mary could be a mutant that is capable of cognitive transitions denied to normal folk. She could have a neural prosthesis that achieves the KWIL state without employing photons of the requisite wavelength.
What we can't do is break the link between KWIL and achieving the right brain structure for KWIL.
Physicalism is not at all bothered by the fact that brain-states can be out of reach during access to accurate descriptions of those brain-states.
I'll be posting on this topic myself soon.
But I agree with the idea that physicalists should love Mary. In the right context, this thought experiment is strongly supportive of physicalism and actually casts anti-physicalists as fairly gullible.
assuming physicalism in the first place, It’s an uninformative tautology to say that you can’t KWIL unless you have the right brainstate for KWIL. But what reason is there for the physicalist to think that book learning can’t get you into that brain state?
I think it is not so much a tautology as a shift in focus from the contentious to the obvious. It is the shift that is important, not the fact that it lands on a truism. It is also a truism that swamp-Mary has KWIL, for instance, but it seems to make a point worth making, because it breaks the link between seeing and KWIL.
Say we have concept 1 (C1) and concept 2 (C2), and some dispute about whether C1 implies C2. Do we need to refer to the truism that brain-state of C2 is necessary for C2? Usually not.
C1 could be the concept of "2+2" and C2 could be the concept "4". Nearly all discussions about whether "2+2" implies "4" rightly interact with this at the semantic level, ignoring the actual brain-states C1 and C2. There could be a post-torture state in which Winston Smith from 1984 follows a different transition, but this is usually not relevant. There could be a genetically engineered species that was otherwise intelligent but could never achieve this transition because evil experimenters had wired their brain strangely for shits and giggles. They might think "4" was of divine importance.
But what if C1 is full knowledge of redness at the physical level, and C2 is the redness KWIL? Here, the focus should not be on the semantic truth of whether the content of one brain-state entails the content of the other; the primary focus must be on the actual physical transition (which in turn leans on recalling the truism).
The C1 to C2 transition is ultimately not dependent on the truth of what those concepts represent, and barriers of this nature abound in the brain's examination of its own cognitive substrate.
Sure, we need an actual physical reason that C1 does not lead to C2 in Mary's case, but the reason is there in the neuroanatomy. Patricia Churchland, for instance, has a Youtube vid on the relevant neuroanatomy. I will post on this issue soon with a more extensive consideration of the actual physical barrier involved.
EDIT: I didn't knowingly use the same mathematical analogy as in the other thread, but I saw it when replying to your other reply, so it is funny that the threads converged like this.
You won’t be the first to suggest that neuroscience can shore this up, but I haven’t seen anyone, including Pat, make good on the suggestion. I look forward to your attempt.
It might be a couple of weeks. I have actually written a hundred-page version of my ideas on this issue, but that's too long to post.
I have my personal terminology for the spectrum of opinions on this. It starts with "Gap Denialism", which might sound pejorative but is not intended that way, and it ranges through various forms of "Gap Compatabilism" through to full-blown "Strong Hardism".
You seem to be a Gap Denialist, or maybe a Gap Not-yet-convincedalist. (The latter is not really on my list.)
The logic of the conclusion seems to require Mary can know ALL of the physical facts about seeing red, at least to reach the contradiction if Mary learns a new fact by seeing red, because she already knew ALL the facts about seeing read.
But I suspect that this "knowing all of the physical facts about red" contradicts physicalism itself, at least in the physical universe we live in, because all the facts about seeing read doesn't sound like it could be a finite number of facts.
Facts that are physical, in this physical world, each require a minimum amount of mass and energy to instantiate, according to physicists as I understand it. So it seems that an infinite number of facts can never be accessible to a being, in a finite period time, because they will will be spread over an infinitely large region. Any overly high concentration of facts will collapse into a black whole, in fact, like any other constructions of matter and energy.
So, at least in this universe, I think this refutes the logic of the argument that derives a contradiction if Mary learns a new fact by seeing red, because she already knew ALL the facts about seeing read.
I don't actually think this is just an optional quirk of this particular physical universe. On the contrary, I suspect imagining a world in which infinitely many facts can be "known" physically to a physical being of finite extent may be the same as negating the difference between physical and non-physical facts. That is, I think such a world would not really meet our intuitive concept of physical, as well as being very unlike a physicists concept of physical (I assume that is what counts as the non-intuitive concept of "physical".)
For example, once you can have an infinite amount of information in a finite space, you can have, for example, life forms evolving at all possible scales. There is no bottom limit to the intricacy of structures of information, like DNA and cell mechanism. As beings in that world, we would immediately be eaten from the inside by much smaller, more information dense new life that evolves faster than we do, and that new life would already be eaten alive by new life smaller than it that evolves faster than it does, and so on... beautiful, but maybe hard for anything to reach the philosophizing stage, too.
Assuming Mary knows ALL the physical facts is a simplifying assumption so that the argument can be stated in a concise way. Mary only needs to know as many physical facts as we would plausibly expect to be relevant to the best physcalistic scientific theory of an empirical phenomenon. No other scientific theory has ever required a scientist to have godlike knowledge. There are facts about the numbers of planets outside of our lightcone that couldn’t possibly be relevant to understanding human color vision. Mary knows all the physical facts that could plausibly be relevant to a scientific understanding of human consciousness. Assuming that knowing what it’s like to see red is relevant to understanding human consciousness, and that all the plausibly relevant physical facts are knowable by book learning, is a physicalist theory of consciousness necessarily incomplete?
In the argument, not specifying fully "all" the physical facts removes the logical force of contradiction from Mary gaining a new physical fact, so it can't be claimed to prove something, I guess.
I find it interesting that the logical contradiction can be considered to support either physicalism or anti-physicalism, depending on which term you decide the contradiction applies to.
I.e. if you decide it supports that super-Mary can never know the experience of red until she experiences it, that might (maybe) make you conclude for non-physicalism.
Or if you decide it supports that super-Mary must have already known the experience of red, then you might conclude for physicalism.
Removing the force of "all", then you move more into the discussion of the plausibility or content of KWIL, for example @disagreeableme's considerations of what in reality the physical facts of the experience of seeing red might be and how they could be really represented or learned by Mary. A super-Mary who can change her own neural network at will would be able to "learn" the experience of red this way, without requiring access to a physically impossible "all of the physical facts", so this Mary seems physically possibly, just implausible, which is fine for a thought experiment.
i gave you a recipe for rewriting the argument to yield a slightly different but still valid argument, but you have to apply it to more than just the first premise.
For me the knowledge argument kind of misses the point, the thing I actually wonder about is why there is subjective experience in the first place and how it works.
I’m going to go from the theme that The Good Determinist provided earlier — I think I’d need to be the bug or whatever that experiences ultraviolet vision in order to effectively “know what it’s like”. And like him I don’t think this means that Jackson’s original thinking on the matter mandates the existence of anything non-physical. But Tom only provided an intuition to this effect. I’ll provide a theory.
My theory is that by conflating “ontology” (or what exists), with “epistemology” (or what’s understood), the thought experiment beguiles us to make a category error. Notice that if I were some sort of god with a perfect grasp of the systemic causality under which I exist, then I’d inherently know what science doesn’t know, which is to say the mechanics behind seeing red, dragonfly vision, and so on. But this epistemology in itself would always remain just that, which is to say understandings of how such dynamics work. No epistemology should ever exist as actually “seeing red” or anything else ontological, and regardless of the effectiveness of that epistemology. I can go the other way with this reasoning as well. People who actually do see red can obviously still be clueless about what causes them to do so.
It’s interesting to me that philosophers have fully separate classifications for studying “what exists” versus “understandings of what exists”, and yet have had a difficult time grasping the category error in this thought experiment.
yes, begging the question. no, not useful. it’s a leftover from an ancient and very naive epistemology whereby to know stuff you have to have it in front of your face so you can see it really well. the past several hundred years of progress in math and science have blown that epistemology way out of the water. As descartes pointed out you can perfectly comprehend the difference between a circle, a 1000-sided polygon, and a 999-sided polygon even though you could neither see nor visually imagine the difference
Either (1) it’s a general rule that knowing what it’s like to have an experience requires you to have that experience or (2) it’s not a general rule. If (1), then you have to deny, in the face of arguments and intuitions to the contrary, cases like Hume’s blue and the Japanese flag. If (2), then you incur the obligation of explaining what generally distinguishes the cases that conform to the rule from the ones that seem to be exceptions to it. Why should we credit your hunch about ultraviolet? How do you know that it’s true? It’s a claim about what will or won’t happen in the future, and those are generally hard to be certain about unless they’re just grounded in some definition (like “no triangle will ever have four sides”). But there’s no plausible definition that anyone’s floated that’s going to help you here. So do you pick (1)? Or (2)?
You can put me down for (1) Pete. Yes I can imagine a Japanese flag if someone describes it, but that shouldn’t be the same as actually seeing one by which I would KWIL. And I can certainly imagine the missing blue if I’m presented with a spectrum of blues, but seeing the blue should be different. What’s the difference? The ontological consciousness of “seeing” versus “conceptualizing”. (It should be easier to demonstrate this once science grasps the causality of consciousness, though as a strong naturalist I at least presume such a difference.) Someone who merely knows the mechanics behind how their brain theoretically could create the experience of red, should not thus experience red. This is not to say that Mary should be said to “learn something new” with such an experience — Jackson used a sleight of hand there (and I think later recanted whether or not he grasped the associated category error). Instead here Mary should be said to “experience something that she never had before”. This point is demonstrated again when people who know nothing about the brain can still experience red and thus KWIL during those experiences. (Later they may have memories of knowing what it’s like, or MoKWIL. 🤪 ) Epistemology should not be confused with ontology, though Jackson beguiles us into this category error, and not entirely unlike the one of positing which color triangles happen to be.
I don’t see why imagining would be relevant for whether you KWIL. Before you imagine it, you need to have prior knowledge of what to imagine. And that is knowledge that you retain between imaginings, and even while you’re unconscious. And knowing what’s it’s like is something you use to correctly judge “i am not currently having a red experience “
That’s a great point. Yes in practice I use my memory for the concept of red, which is to say an epistemological understanding of it even when I’m not currently experiencing an ontological red. But I should be able to adjust my argument. Notice that if I had never had an ontological experience of red then I’d never have gained a memory of it that I might use epistemologically for such a general understanding. So I still consider there to be a category error in positing an experience by means of an understanding of what creates an experience.
Or to put my position here more bluntly, let’s say that I’m held captive and tortured for the rest of my life. To say the very least, that would fucking suck! Let’s also say that Jack is with us, the pain scientist who knows all there is to know about how the brain creates pain, as well as everything that these sadistic bastards do to inflict it upon me, though he has a genetic defect such that he can only ever feel neutral to amazingly good and so has never felt anything that even approaches “pain”. I can see how one might argue that he could think about the difference between feeling neutral versus wonderful and so posit pain as an opposite that thus feels “bad” rather than the good that he knows so well, but unless he were to actually feel pain (like Mary who ultimately sees red), I don’t think he’d be able to grasp my actual misery. In any case it seems to me that knowing everything about the neurology that creates my horrible existence, wouldn’t even provide the backwards conceptual association between pleasure versus not. The neurology behind what I experience should not give him any sense of my pain because epistemological understandings should not create ontological experiences.
i’m not quite following the argument here but suspect it’s trading on an ambiguity concerning “memory of”. Do I have a memory of Columbus arriving in North America in 1492. In one sense of “memory of” I clearly do not, since I wasn’t born until 1969. In another sense I obviously do, since I learned it a long time ago, and still remember it. A crucial question is which kind of memory is involved in knowing what x is like. There are lots and lots of cases in which it’s quite clear that you know what x is like even though you have no history of contact with x. I know what it would be like to look at a perfect forgery of the mona lisa, although I haven’t seen one yet. SOOOO, what reason is there for a physicalist to believe the experience of red must be different in this regard? I know lots of people find it to be intuitive that this must be true. But what argument is there for believing that intuition? I haven’t encountered one yet that isn’t based on some fallacy like begging the question or equivocation.
Yes it should be important to refine the concept of memory here since everything we can say one “knows” will in some sense exist in the form of a memory. For the moment I consider it useful to define this broadly as “past consciousness that in some sense remains for present use”. And as you’ve mentioned, you’ve never had a consciousness of Columbus actually arriving in America, but rather various memories which suggest that he once arrived. Maybe it was a hoax? Regardless it’s the various suggestions you’ve encountered over the years that have given you such a memory rather than such an event itself. That would be the ontology which provided your epistemology. This doesn’t suggest to me that if I knew the brain mechanisms which create an experience of red (epistemology) then that would also give me the experience of seeing red (ontology). I’d think you’d need a positive argument that one must lead to the other to build a good case against mine, though in the end I can’t fathom a natural possibility.
On the Mona Lisa, it seems to me that unless I’m looking directly at the original, then any version that I do see will be a copy. But reasonable copies ought to give me a good sense of the original. In the examples of an amazing color scientists who never sees color, as well as an amazing pain scientist who never feels anything except good, I don’t see how they’d get a reasonable sense of the red or the pain that they never actually experience, that is until they do.
Perhaps you’ll say that I’m begging the question by defining knowledge of what exists differently from experience of what exists? But did I arbitrarily define these terms this way, or are they generally useful to define this way?
A possible response to Swamp Mary (similar to the point by Bill Robinson) is that her KWIL to experience red is still only latent, just like Mary’s in her black and white room. She only would KWIL if she activated that memory by recalling the apple in her mind’s eye. Or to put the concern another way, what if Mary (and therefore Swamp Mary) had aphantasia? Would aphantasic Swamp Mary really KWIL to see red (before her next encounter with a red thing)? At her next encounter, her reaction might be “oh right, *now* I remember WIL to see red”, but is that really any different from Mary’s supposed “so *that’s* WIL to see red” reaction when she emerges from the room?
What reason is there for thinking “latent” knowledge isn’t knowledge? I know all sorts of things over long stretches that are punctuated by episodes of me being unconscious. Having knowledge is like having a memory; whereas the direct analog to deploying that knowledge is recall, remembering, or recognizing. Imagery seems a red herring when discussing knowledge here, as one would need to know what to imagine before they imagined it.
Thanks! I guess my point - not particularly earth-shaking - just comes down to Swamp Mary not being a knockout punch for those that hold to a distinction between Russellian knowledge by acquaintance and knowledge by description. I'm not actually one of them, but just playing devil's advocate. Also I enjoyed giving Mary aphantasia just for fun :-)
what it’s a knockout punch for isn’t Russellian knowledge by acquaintance (KA), but instead the related view that KA is part of the analysis of knowing what it’s like. Swamp Mary is one of several counterexamples to such analyses.
They both were locked in the room for all those years and both acquired the same BLAD. They’re released at the same time and see the same apple. Then they’re told to close their eyes and describe what they’ve seen.
Mary, as mentioned, has aphantasia. Harry, on the other hand, has hyperphantasia. Let’s stipulate that they have the same knowledge by description (KD) of the apple - that is, Mary can describe the apple and its properties (as well as all the physical facts about what happened in her brain, etc when she saw the apple) in the same detail as Harry when asked. One might be inclined to say that, in addition to their shared KD, Harry also has (or seems to have) KA of the apple (well I might be inclined to say that… not sure if you would?) Am I understanding you right that, for purposes of KWIL, when they close their eyes there isn’t an essential difference between them? (Or please set me straight)
Thank you again - I’m a fan of your work and grateful for the dialog!
Thanks - you're a rock star! (And I mean a cool rock star, like Tom Petty, not an annoying one like Chris Martin. There are other philosophers of mind who remind me of Chris Martin, but I won't name any names.)
Nice discussion, but I’m not satisfied. The main problem is that you have Mary knowing red as a highly complex, brain event property, but there is nothing about red in ordinary experience that remotely suggests any such complexity. So, Mary may know a lot about red, but there is still something she doesn’t know, namely how red appears to those who have ordinary experiences caused by ripe strawberries, bullfighters capes, etc.; (or, for those who like the WIL formulation, she doesn’t know WIL for trichromats outside the room to see a ripe strawberry).
Swamp Mary still has duplicates of the neural effects that seeing something red caused in Mary. If (post-release) Mary can generate a lively memory image of a ripe strawberry, so can Swamp Mary; which does not support physicalism unless consideration of Mary alone does. If (post-release) Mary cannot generate such an image, she may still have a capacity to classify a new experience correctly as ‘red’. If so, Swamp Mary will also have that capacity. – My conclusion is that Swamp Mary can’t really add anything to what can or can’t be shown by considering Mary herself.
I just don't grant anti-physicalists the initial presumption that anything about "ordinary experience" lends itself towards the view that there's anything "it's like" (whatever that means) to see red in the first place. That is, I don't just deny that people have qualia, I deny that the notion that we have qualia/phenomenal states is part of ordinary experience. I think the belief that we do was invented and perpetuated by philosophers.
As such, in the case of Mary, I think the thought experiment will typically appeal to people influenced by certain philosophical traditions, and that its appeal is an artifact of training and enculturation, not some shared feature of our psychology.
big agree although my take on “is like” is slightly different. knowing what x is like is one and the same as knowing something or other about x. So, if I know that ruffles have ridges, then I know what ruffles are like. If I know that water is wet, then i know what water is like. This works exactly the same when you plug “seeing red” or “experiencing red” in for x. So, if I know that Russel saw red on a Tuesday, or that seeing red makes your c-fibers fire, I know what red is like. Somewhere around here my opponent will say something like “no, stop, you’re not using those words in our special, proprietary, Nagel-blessed way” at which point I automatically win.
thanks, Bill! However, lots of ordinary (non-future-science) experience reveals that red is complex. all reds are complexes of hue, saturation, and brightness. red resembles orange more than it resembles green, and such relational facts about red are implausible to suggest are merely accidentental. red is a secondary color in cmyk mixture schemes (and is mixable from magenta and yellow). shades of red seems darker than equiluminent shades of yellow. future science is unnecessary for buoying up the premise about knowing phenomenal red by description
Yes, red has _some_ complexity -- hue, saturation, and brightness. (The other things you mention are its relations, not its internal complexity. E.g., I am brother of John Robinson, but that, _pace Leibniz_ doesn't make me more complex than I would be without him.) However, the kind of complexity Mary must be aware of is myriad neural firing rates, and rates of change of those rates, and so forth. Red does not have _that_ kind (or, degree) of complexity (otherwise, Ancient and Medieval philosophers would have written about it). --- As to knowing by description, I think congenitally blind people often know a great deal _about_ what red is, but like pre-release Mary, they don't know what red _is_. Yes, physicalists can consistently claim to know a great deal _about_ red, but they do not have the actual property, red, in their inventory of what there is.
I get that that’s an antiphyscalist view of what red is, but it can’t be an input to a non-question-begging argument against physicalism. Anyway, I don’t know what the internal/external distinction would mean wrt color aside from essential/accidental. And more to the point, if one’s descriptions nail down a color’s essential properties, one cannot be accused of leaving out something essential .
I don’t think it’s question-begging to take it as a datum that when we are, in ordinary circumstances, having an experience caused by, e.g., looking at ripe strawberries, there is nothing in our consciousness that is anywhere near as complex as what would have to be in pre-release Mary’s thoughts of neural event structures, if those structures were to have sufficient complexity to make them correlates of, say, red in particular and not just, say, some sensory input or other. I agree that internal/external for a property like red probably is the same as essential/accidental (provided that causes are taken as contingently related to their effects). But “if one’s descriptions nail down a color’s essential properties” is a big if. It would be question-begging against my view to assume that that conditional is satisfied, because on my view, what pre-release Mary knows about is, e.g., the causes of color experiences and not the essences of colors at all.
The claims that are striking me as question begging against the physicalist are, first and foremost your earlier “[physicalists] do not have the actual property, red, in their inventory of what there is” but possibly I’m misunderstanding whether that’s supposed to be part of your argument or just a statement of your view. I didn’t think the complexity claims were specifically question-begging, but we still disagree about them (and it’s an interesting disagreement). How many properties, essential or otherwise, does someone need to know about x to know what x is like? It’s implausible that one needs to need to know all of the properties. I know what it’s like to taste Cabernet, although a sommelier probably knows way more about what it’s like. Probably if someone knew zero essential properties of x and only accidental properties, it might be a stretch to say that they know what x is like. So back to pre-release Mary. Without violating the rules of the thought experiment, I can allow her all sorts experiences, as long as none of them are what both sides of the debate would call “red experiences”. So Mary can know all sorts of stuff “by acquaintance” via seeing other hues, etc. She also knows ALL sorts of cool correlational stuff about which neural modulations go along with which experiences, just none that acquaint her with so-called “red experience”. Both sides of the debate are obligated to describe this in neutral non-question-begging ways. Just like the physicalist cannot assert at this point that these correlations are actually identities, nor can the anti physicalist assert that these are mere, non-identity, correlations. The main residual disagreement between the two of us seems to be whether the complexity pre-release Mary accesses is complexity enough for knowing what it’s like to have a red experience. Your saying she has nowhere near enough needs to be bolstered by something non-question-begging. Why does it fall short? Simply because it’s not acquaintance? Or is there more that can be said?
“Physicalists do not have the actual property, red, in their inventory of what there is” is a statement of my view, and absolutely not a premise.
I’d like to hear more about just how you disagree with the claim that ordinary experiences of, e.g. red do not have any hint of the degree of complexity that would have to be in the content of pre-release Mary’s thoughts about the neural activation events that are the red experiences that she knows others have on occasions of their being exposed to, e.g., ripe strawberries.
I don’t think sommeliers know any more than I do about what it’s like for me to experience sipping cabernet. (I do think their training may well have caused their experience while sipping cabernet to be different from mine – although not entirely dissimilar.)
I think the original experiment had pre-release Mary in a room with only black, white and shades of gray. I agree that one might get to know what it would be like to experience Hume’s missing shade of blue, in the conditions he imagined; but I don’t think pre-release Mary comes even close to having that kind of background.
I think it’s a phenomenological datum that when you look at a ripe strawberry, you don’t thereby have anything in your consciousness that has remotely the degree of complexity that would have to be in pre-release Mary’s thoughts, if those thoughts were adequate to be specific to the brain events that people (or just one person whose brain scans have been intensively studied) have had when looking at ripe strawberries, bullfighters’ capes, and so forth. I think that’s obvious, but there is also an argument that our Ancient and Medieval forebears were certainly smart, and none of them gave any hint that they found experiences to have the kind of complexity that would have to be in pre-release Mary’s thoughts.
I agree that statements like ‘red is more like orange than it is like blue’ are necessary truths, but I don’t think that knowing a humongous number of such statements would bring anyone closer to knowing what any of the mentioned colors is (or, WIL to experience them). I think congenitally blind people could know a huge number of such truths and not know what red is.
Typically, when I think about what it means to "know what it's like," to have some experience, I interpret that as meaning something like, "I can simulate the experience in my head (e.g., I can see the color red in my mind's eye) in a way that feels similar enough to having the real experience," so I can point to that and say, "That's what it's like." The Mary's Room argument only works by conflating this sense of KWIL to a sense that's more related to knowing some specific proposition about the experience of seeing red. The latter interpretation seems to be the one you have, and on this interpretation, physicalism does imply that Mary should be able to KWIL to see red even before seeing it herself, but, as you explained in the article, this is no problem for physicalists. There's no reason to think she couldn't KWIL, unless you already reject physicalism, and on physicalism, it seem perfectly reasonable how she would come to know.
Under the former interpretation, it's not true that learning all the physical facts about color perception would let her KWIL to see the color red, since, after all, there's no necessary connection between knowing some information and being able to perform a mental simulation that generates a particular experience at will. Even if the information you know is about how that mental simulation works neurologically, that doesn't magically grant you the ability to just make it happen. But of course, all this mean that, under this interpretation, physicalism doesn't imply that Mary would KWIL to see the color red before she actually sees it.
So we have two interpretations. On one of them, the first premise of the Mary's Room argument, that Mary learns some new fact when seeing red for the first time that she couldn't have learned by reading about it, is false, while the second premise that physicalism implies that she should be able to learn WIL to see red just by reading about it is true at least in principle. On the second interpretation, the first premise is true, but the second premise is false. On no interpretation are both premises of the argument true, so Mary's Room fails to disprove physicalism.
I am a physicalist and a physician, but I don't agree at all. I love Mary too, but because it illustrates the following:
1. differing physical causes (coloured light, monochromatic light) lead to different physical effects (brain states). It really needn't be more complicated than that. Changing the stimuli changes the downward effects, if there is any sensitivity to the change in stimuli. It's incredible to me how frequently this is missed.
2. Mary is an expert on ordinary color vision in ordinary humans. She hasn't a clue about her personal brain state - a brain that has been deprived of colour stimuli entirely. She has a very different brain than you or me. If she were a super scientist specialising in her own brain, then she could predict her brains future physical states. Which is obviously not the same as being in that state - but still, she could, in principle, predict what she would say if she hadn't made the prediction (since predicting is a physical process this affects the future physical states).
3. Mary certainly doesn't see the world in black-and-white. She sees no lack of colour in her room, a lack which is crucial for the B&W "feel". And if she can distinguish colours properly when let out (questionable), then she certainly won't experience the redness of red that you or I are familiar with. Every experience is created in light of previous experience.
1 is a red herring. I don’t think any party to the debate is denying or missing 1. Seems instead like a totally obvious point that has no bearing on the discussion.
Thank you for the article and for your response to my somewhat sloppy comment, you deserved better.
3. Thanks, I'm glad to hear! Have you or anyone else pushed this point? (I presume I'm not the first). I just posted "Mary in the Coloured Room" (https://markslight.substack.com/p/mary-in-the-coloured-room) which doesn't quite push this angle exactly, but a related one. I'm soon going to post "Mary Returns", where she several decades after release, with no memory of experiencing any lack of colour in the room, returns to it. Check it out if you're interested.
2. Fair enough. Of course she can have a clue! My bad. But as far as I can see the point I was trying to make stands: there's a crucial difference between a super scientist specialised in the olfactory and gustatory neurophysiology of sommeliers or a super scientist specialised in the olfactory and gustatory neurophysiology of people who av never ever smelled or tasted wine. And if you're a super scientist specialised in sommelier neurophysiology, you're an expert at predicting sommelier reactions and responses to various wines. There's simply no way you can accurately predict your own response to a particular wine, from your immense knowledge of sommelier neurophysiology. Or what am I missing?
1. Okay. Maybe it's a red herring among physicalists (I'm in no position to say), but is it really, in general? There must be something obvious I'm missing here. Knowledge is physical, and aquiring knowledge is a physical process brought about by physical inputs. I don't see how it could ever be a threat to physicalism that input A (black letters on a white background) does not put Mary's brain in the same state as input B (being released). To me, the word "knowledge" here is a red herring, just confusing matters. For Mary to be able to acquire the same brain state as in post-release, she would either have to perform brain surgery on herself to put her brain in exactly that state, or she would have some strange neurophysiology that does the equivalent - in which case she would be very non-human and she might not end up KWIL to see red for ordinary humans. As a mere physician not well versed in philosophy, I am of course humble to the fact that I could very well be misunderstanding you. To me, at least, non-physicalists seems to miss this distinction all the time.
Re: 2 and 1. I can put myself in the right brain state to know all sorts of things about Jupiter without flying all the way to Jupiter or even looking at Jupiter through a telescope. What reason is there, besides a hunch, for a physicalist to believe knowing what it’s like to see red is relevantly different? I’ve given arguments for why the hunch is false. All the arguments I’ve seen so far for believing the hunch is true are long-winded restatements of the hunch, not positive evidence for the truth of the hunch. Stay tuned for another post where I spell this out even further.
3: Thank you very much! I'm afraid I can't access the papers. I will try from work.
2 and 1 - Thank you for your response! I have read your arguments in this post and done my best to understand them. I must say that it seems to me that if you agree with 3, then that is not consistent with your objections to 2 and 1.
But first: I'm humble to my limitations due to having never taken a single philosophy course so far (I'm starting in May). But my perhaps narrow physician-physicalist perspective is that you in this post and responses to me are ignoring or not properly addressing important physical and biological objections to your view. As long as Mary is even remotely human, which is reasonable to presume, black text on white paper cannot put her brain in the same state as building a repertoire of "automatic" visual colour data processing that you and I have trained over many years. This is not just a hunch. It's a reasonable conclusion based on what we know about physical, biological, cognitive systems.
Thus, Mary cannot KWIL for ordinary humans to see red by any other means than building that repertoire. Similarly, you cannot train a co-joined LLM+visual processing AI to recognise and report on colours merely by training it with text token data. The LLM may report that it can imagine seeing something red, but it's not going to be able to draw or recognise something red reliably. That requires experience (training). Ultimately I think these are physical/biological/computational issues, not philosophical ones. Although what counts as "human" and "KWIL" are of course philosophical issues.
Re: 3 and applying it to 2,1:
As I see it, everything is conceptual and relational, including colour experience and KWIL to see a red apple. KWIL red apple requires not only to be acquainted with red stimuli and apples, but also with grey, blue, green and yellow stimuli, and other fruits and objects. It requires the associations to fire, blood, blushing, ripe apples, and it requires the "unlikeness" to green grass, the blue sky, bananas, and the unlikeness to the absence of any visual stimuli. Likewise, imagining a shade of blue one has never seen requires acquaintance with not only other blue shades, but with all the other colours, and with with the concept of shades, to begin with. We have evolved and trained the concept of shade discrimination and we cannot just imagine it without ever having seen any shades.
For these reasons, when Mary is captive, she does not know what black and white looks like to ordinary people. And if someone gave her a red apple in her room, she would still not know what a red apple looks like to us - as her brain would not "automatically" see the likeness with ripe raspberries, blood, and the unlikeness to green grass. It wouldn't just come to her without the appropriate past experience. The same goes when she is released. She learns many new things, but even if she could distinguish red from green, she will not learn what redness looks like to ordinary people (which is, of course, not identical between any two persons). Redness cannot be separated into how it looks on one hand and how we react to how it looks on the other. (If you and I have different favourite colours, that is enough to determine that we do not see colours the same way).
KWIL to see a Japanese flag without having seen one requires familiarity with concepts of roundness, rectangles, white, red, all the other colours which are absent, flags and countries and how they relate to each other. And of course, reporting a successful imagining of flag does ensure the successful imagining of a flag.
You sure can learn a whole bunch of things about Jupiter without even seeing an image of Jupiter. But I think your hunch-objection goes both ways. Even if you then go to Jupiter, and report "yes, this experience is exactly as I imagined it", if brain scans tell a different story, I will trust the brain scan more. That goes for my own self-reports too, of course. But I think you cannot separate out KWIL to see redness from remembering redness.
There's a slightly autistic but I suspect important distinction to be made between KWIL to imagine the japanese flag and KWIL to have seen it. I think we lump these together, for good cognitive/functional reasons, but they're not the same thing. Not sure.
I take it that it follows from your position that if Mary was congenitally blind she could still KWIL to see a red apple?
Looking forward to your next post on the topic! And I just saw you're into meditation too, gonna read that post! thanks.
yes. there is no knowledge that isn’t objective knowledge. in other words, there is no knowledge that requires some particular kind of sensory experience to be had by the knower of that knowledge .
Okay, I see. I find no other way of interpreting you than that you claim than this: I can, in principle, be born into a completely dark, silent and odorless room, kept alive through intravenous nutrition, yet I can by reading Braille somehow train my visual, auditory, gustatory and olfactory cortex so as to be able to distinguish all sorts of shapes, colours, movement, tastes, smells, voices, etc. When I am released I am not at all surprised by anything (provided I have prepared properly), and can navigate the world like anyone else, communicate, recognise colours, flavours, smells, etc, recognise celebrity faces and voices, and so forth. Correct?
In my view, this where I think philosophy that is not rooted in the reality of biology and neuroscience risks being very misleading. Philosophy, knowledge, cognition, all of that is always physical and biological. I would think you agree with that, and I see you're a cognitive scientist too! And as someone who hasn't studied philosophy, I get that I'm probably missing something. But I'm sincerely very puzzled and I fail to see you addressing this! Anyway, thanks for trying to get me to understand and for your engagement. It's quite unique. Namaste! (first time I use that word)
Where was this when I was asking you philosophy questions on twitter? Very well written, also wasn’t aware of your Swamp Mary paper, shall give it a read. (Maestro btw)
First, just wanted to point out that doing the Mary experiment is completely do-able today, with a slight variation. If the only light available to Mary is via sodium lamps everything she sees will be shades of orange, including her blood, etc. There was a museum-type place in San Francisco (forget the name, Exploratorium?) that had such a room. On one wall there was what appeared to be a black & white (well, orange) photo of the city. They provided flashlights for visitors, and if you shined one on the wall you’d see it was in full color.
Second, what would happen if the first colored thing Mary sees is an apple which someone has dyed blue? (Think Dennett came up with that.)
We’ve long been able to upload images to ChatGPT; no need relevant to the current question to bolt on an additional “vision system”. It can answer all sorts of questions for you about how the pictured apple seems, what it’s like, etc. I have no idea, though, what would motivate anyone saying that it (or anyone) has got anything direct, infallible, or ineffable.
Uploading an image is not the same thing as experiencing blue right now. But notice the case where the system experiences something, but doesn’t know what. If the system was capable of learning, it might come to call that new experience “red”. But before that learning, how is that experience effable? How is it not direct? How is it not infallible (barring intervention, which would make it hallucination)? BTW, direct, ineffable, and infallible necessarily refer to the system’s perspective, right?
There is some evidence that indicates that indeed people blind from birth and even prenatal more mature fetuses might have visual experiences of color blobs that could resemble those we directly experience (though presumably fetuses may not be entirely deprived of seeing light in certain circumstances, I'd argue). However, even with Mary's precise physicalist knowledge of the brain I cannot see how she would correlate sensory perceptions she does have with any of those blobs as to say which is which, unless there are *very* specific brain functions that correlate phenomenally and spectrally with colors that exclude errors such as color-blindness. In other words Mary would have to know a lot about not just "normal" visual experience that correlates to brain function with respect to colors, but also to know that she is not having an erroneous experience of a kind of color blindness.
"...a complete physical description of seeing red should let someone KWIL it...knowing what it’s like isn’t chained to experiencing it."
Many animals, e.g., butterflies, fish, birds, can see in the ultraviolet spectrum. Can we know (in principle) what it's like for them? On physicalism, there's nothing categorically private about experiences - they're public objects in principle amenable to complete physical description - so on your view I guess the answer would be yes. But it also seems like one would have to instantiate that description to know what seeing ultraviolet is like. Is that necessarily to smuggle in a non-physical ghost?
Can you say some more about “instantiating a description” means? Suppose I know that Jupiter is a gas giant, and thus that to describe Jupiter as a gas giant is to utter a truth. Which of us, me or Jupiter, instantiates the description “is a gas giant”? If I had to guess, it’s Jupiter doing the relevant instantiating here. Is that correct?
Yup. For me to know what seeing ultraviolet is like I'd have to be a system in which the physical description obtained. At least that's my intuition. Reading pages and pages of equations and flow charts and seeing slides and videos of the relevant neural states (e.g., the total conceptual/propositional/quantified description of the NCC of a butterfly seeing in the ultraviolet spectrum) doesn't seem to make me KWIL to see in the ultraviolet spectrum.
missing shade of blue; japanese flag; swamp mary. additionally , reasons to think the sensory idea of visible ultraviolet is a complex and thus describable independently of any assumption of physicalism (e.g. being a color, it would be a complex of hue, saturation, and brightness). and more!
So I could know what ultraviolet is like for the butterfly from knowing the exhaustive description of the quantitative parameters involved in its hue, saturation, and brightness, even if I were blind. I don't need to have had any visual experience to know what such experiences are like on your view since KWIL is just to know all the physical facts. This suggests that a system that doesn't have experiences could know what it's like to have them.
Firstly, I *do* agree that that Mary's Room doesn't at all disprove physicalism. (FWIW, I have a physicalist metaphysics.) And I do agree that Swamp Mary would know the experience of seeing red because it would be in her memory as obtained from OG Mary.
Where we disagree is about objective versus subjective knowledge. They involve different brain states, and I see no way to accomplish the same brain states involved with subjective experience through objective learning. The only way around it is for Mary to use her expert knowledge to construct a machine capable of generating those brain states. Essentially, to create a VR experience, but that is indistinguishable from having the experience.
Why does Mary have to have subjective experience brain states in order to know what it would be like to have those brain states? It’s the same question we’ve been circling around, now with the phrase “brain states” inserted in it.
Because the only way to have those brain states is through experiencing the phenomena associated with them. You can't induce the retina, visual cortex, and brain activity of seeing red by thinking about it objectively.
FWIW, the subjective/objective difference doesn't AT ALL deny physicalism as far as I'm concerned. It's just that brains are unique in our reality in having both an outside view (the object view) and an inside view (the subjective view). Everything else we study, we only study from the outside.
The infamous Hard Problem is the question of how the hell something can even have subjective experience, and at this point we just don't know enough, but there's no reason to believe the answer goes beyond physicalism. It's just that brains are the most mind-bendingly beyond-complicated systems we've ever encountered.
I am a physicalist, at least in this sense: I believe (though I know I cannot prove) that the mind is a result of purely physical processes occurring within the brain, and, furthermore, that how this works could, at least in principle, be explained within our current understanding of the physical world, requiring physics no more fundamental than is needed for biochemistry.
Nevertheless, I suppose that Mary will learn something new when she sees red, and I do not do so with a shrug - on the contrary, I feel that asserting she would know what it's like prior to seeing red requires assuming the brain has certain capabilities that are implausible and lacking empirical evidence.
Physicalism (or at least my flavor of it) effectively implies that coming to know what it is like to see red requires some physical change in the brain - changes which can be traced back to the stimulation of red receptors in the retina, and while the specifics are probably different in detail from one person to the next, are similar at some appropriate level of functional abstraction. Under the premise that Mary knows every relevant physical fact, we must assume that she knows (e.g. through scanning and modeling) exactly what changes will occur when she sees any particular red-containing scene, but there's no reason to believe she has the ability to bring about those changes through mental effort alone - and, according to the physicalist premise, if the changes do not occur, she will not know what seeing red is like. Consequently, any physicalist who claims that Mary will learn from her studies what seeing red is like, carries the burden of providing some justification for assuming that she can use her knowledge to bring about the requisite neurophysical changes.
I am saying that there is no good evidence, in either personal experience or neuroscience, that we would be able to bring about the necessary changes at the neuron and synapse level through mental effort alone, unaided by tools - but what about the examples you give? Even if Hume was right, interpolating between very similar experiences seems considerably less demanding than synthesizing them from scratch, and it seems (at least to me) unlikely that someone who had only seen blues and greens could come to know what seeing orange and red is like, yet Mary is expected to do even more with less. Similarly, if you do not already know what red is like, it seems unlikely that you can know, except in abstract terms, what the Japanese flag looks like. As for Crane's argument, if you had never experienced pain (and there are some people who do not), I suppose you could, on hearing other peoples' talk about pain, deduce that they have experienced something you have not, but that would not tell you what it is like. Finally, Swamp Mary is not coming to know what seeing red is like through her own mental effort.
How, then, can I dismiss the threat to physicalism that Jackson's thought experiment allegedly poses? Basically, it is for the same reason as I object to the thesis that Mary will know what seeing red is like prior to doing so: no matter what she knows, she cannot bring about the requisite physical changes. We understand well enough how various complex and indisputably physical systems work - the weather and biological reproduction, for example - yet no-one expects that learning this knowledge will create a thunderstorm or get you pregnant. Why should anyone think the mind is any different? I think Churchland answered this question in an unfortunately hard-to-find article with the title "Knowing Qualia: a Reply to Jackson" way back in 1989: the argument equivocates over the meaning of the verb "to know." While we talk of knowing what it is like, it is not knowledge in the sense of having a justified belief in certain propositions (if it were, how come we cannot say what these propositions are? And why would anyone assume that Mary could not learn them from her studies?)
Here's an analogy which might serve as an intuition pump for this view: suppose a safe-cracker has broken into a bank one night, and she has X Rays of the vault, so detailed that she can deduce the combination from them. Nevertheless, she still cannot open it if it contains a well-designed time lock that is preventing anyone from doing so until tomorrow morning. She has all the relevant physical knowledge, but not the means to use it.
of course every physicalist will grant that any transition from ignorance to knowledge will require physical changes. but what *positive* reason is there for a physicalist to believe that there’s only one way to know what it’s like to see red, especially in light of the fact that this is not a general truth about knowing what it’s like , as you grant by agreeing with me and Hume?
I thought about it a bit and I now think that I just have certain assumptions about human brains in particular. If she was an alien I would have no problem thinking that she could know what it's like to see red from just learning about it from books.
What are the human-brain assumptions you’re making?
I guess it's something like that you need some processing in the visual cortex to trigger the relevant changes in the brain and that can't be achieved just by learning propositional knowledge.
At least if the changes are big enough, like seeing red for the first time, I do of course agree that we can assemble the Japanese flag from other things we've already seen.
Thanks. The main problem with this is that we have no neuroscientific evidence that the human brain actually works in the way you describe. We don’t know which part of the brain, if any, is the “knowing what it’s like” part and why some signals to it but not others would trigger knowledge. This is just the intuition that Mary must learn something new re-stated in a neuro sounding vocabulary, not a report of neuroscientific evidence for the truth of the intuition.
You're right. As a bit of context, before I thought that even other kinds of beings wouldn't be able to what it's like without leaving the room, so I think I'm closer to your view than I was before at least.
Maybe I also misunderstand what "could" means in this context. There is an epistemic possibility that she could know it, it's not that some weird metaphysical stuff is going on, it just depends on contingent facts about the human brain.
If the necessary connections exist in the typical human brain, then she will know it, otherwise she won't. Knowledge is just a brain state but whether or not reading books brings about the necessary changes in brain states in a human is just unknown to us, it's possible in theory though.
If I'm not mistaken, the only aspect of Hume's philosophy that has come up so far is his premise that a person could interpolate what it is like to see a shade of blue they have never seen before, if presented with a sequence of similar colors, ordered by hue, on either side of the missing one. With respect to that premise, I wrote "even if Hume was right [about that]..." I'm not convinced that we have that ability, but I'm willing to accept, at least for the sake of argument, that some people do - but even then, it seems highly implausible to me that this ability can be extrapolated to the point where it defeats Jackson's argument, which would require Mary being able to know what every color experience she could possibly have would be like, not by interpolation between colors she has experienced, but purely by deduction from a collection of propositional knowledge about color vision, no matter how complete that collection is.
I'm not sure how I gave the impression that I think there's only one way to know what it is like to see red, but I'm guessing it was where I wrote "...and while the specifics are probably different in detail from one person to the next, [the physical changes] are similar at some appropriate level of functional abstraction." What I did not make clear here is how loose and how abstract that similarity might be: I would not be surprised if every experience of a red color is as unique as a snowflake, yet I feel that there must be some level of similarity between a given person's red experiences for that individual to have a coherent concept of redness, and also a probably weaker similarity between different individuals, given that we are using the concept of what it is like to see red in our discussion without feeling that the other person is writing incoherent gibberish. I regard this ability to have a dialogue about what it is like to see colors as a positive reason (albeit rather weak) for surmising that there is some commonality to having sensory experiences of a given type. I will also put forward the fact that our neural architecture is at least roughed out genetically as another weak reason for adopting this view.
As far as I can see, the positions I put forward in my previous post are not predicated on there being only one way to know what it's like to see red.
I think you’ve done an excellent job of describing something alleged to be indescribable. upon having any of those experiences, I would recognize them as what you were talking about.
I think this is a response to my 2 + 2 comment above? (If not I apologize - it is a little hard to follow the threading in substack.)
If so, I'm not sure it's correct that you would be able to recognize the type of experience described while having it. Part of the case, I think, would involve positing (and specifying) some sort of cognitively-impenetrable mechanism by which you are prevented from thinking about the meaning of 2 + 2 while having the experience of believing that 2 + 2 = 5 (or a functionally analogous belief about an epistemic relationship of acquaintance...) Presumably that mechanism would also block you from recognizing that that's what you're doing while in that state?
someone couldn’t *know* that 2+2 isn’t 4 but someone could think about 2+2 without also thinking that it sums to 4. a buddy of mine was once hypnotized by a stage hypnotist to forget the number 3. he seemed, when sent back to the table for us to mess with him, fully cognizant of all numerical facts and relations that you’d expect from someone with a degree in electrical engineering, but until the hypnotist snapped him out if it, it was like threeness was totally erased from his mind. i believe this story in part because once while on acid i became fully convinced i knew german and watched a whole german language tv show convinced i understood everything they were saying (in retrospect i obviously did not). i have believed contradictions before. i know what that’s like.
Wunderbar! :-)
Still, I wonder...
If your buddy was hypnotized again to forget the number 3, presumably under hypnosis he wouldn't think "oh right, I know what it's like to forget 3". Instead, when you were messing with him, I assume his reaction would still just be "huh!?"
Maybe the KWIL distinction goes the other way from how it's usually framed. In this example, a distinction between KWIL as knowing or having a memory *that*, in the past, he was hypnotized to forget 3, vs. *not* knowing / not recognizing that "this is what it's like to forget 3" while actually under hypnosis.
In this view, I guess there would still be an epistemic gap between the two, though the gap would not involve a gain of knowledge like proponents of the KA suggest, but instead a sort of active, introspective suppression* of knowledge under certain circumstances (i.e., not just not thinking about 3, but being prevented, by some mechanism, from thinking about 3 under circumstances where you should be able to.) So as you say, KWIL just is descriptive knowledge, but then the subject's epistemic situation during experience would be something *less than* KWIL.
(I'm not actually suggesting repeating the experiment BTW, just pondering. We all did things in our college / grad school days that makes great "dad lore" - as my son calls it - but which we wouldn't necessarily want to repeat, even for the sake of philosophy or consciousness research :-)
* "Active, introspective suppression" in the above is meant as a hand wavy gesture to try to distinguish this from the personal / sub-personal distinction.
Yes, good. The “it” is this statement (which I personally believe to be false, fwiw): “There exists at least one experience type, for example, visual experiences of red by color-sighted humans, such that while one can come to know what it is like by undergoing the experience itself, and maybe even by being a swamp being or futuristic surgical patient, one cannot come to know what it is like simply by understanding a description, no matter what the description is.” I would be genuinely interested in seeing a physicalist construct an argument that had IT as its conclusion that was based on something other than appeals to its intuitiveness or the alleged lack of arguments for its falsehood. What positive reasons are there, available to a physicalist, for believing IT? I genuinely have no recollection of having yet encountered one.
Here's a possible example of a type of experience that one could not come to know what it is like by way of a description.
Let's say that Orwell's account in "1984" of what Winston Smith experiences after his torture in the Ministry of Love - i.e., a belief that 2 + 2 = 5 - describes a real psychological possibility. That is, after what he has gone through, Winston reaches a state where, in some sense, he actually believes that 2 + 2 = 5.
Now consider the pre-torment Winston. Any description of what it's like to believe that 2 + 2 = 5 would have to include "and you can't think about the meaning of 2 + 2." To follow these instructions, Winston would have to think about what it's like to believe something that requires you not to think about it, which would be impossible. In this analogy, when Mary sees the apple, and it seems to her that she is directly acquainted with the redness of the apple, maybe the "seeming to be acquainted" is - somehow - like "seeming that 2 + 2 = 5" (though hopefully without subjecting Mary to actual torture in Room 101...after all, she's had a hard enough time already.)
This would require developing an account of "seeming to be acquainted" that contradicts "thinking about seeming to be acquainted" in a similar way, but the payoff would be an account of KWIL that does not involve Mary gaining any new knowledge when she sees the apple while also - in a backhand way - validating the intuition that she couldn't have known what it's like from her black and white room. This would also, I think, be distinct from the view that in seeing red Mary gains a new ability / knowledge how. After all, while gaining the ability to ride a bike by practicing riding a bike, there is no problem with also thinking about what you are doing in riding the bike.
This seems very odd: the post of yours that I am replying to here, posted at 14:56Z on Tuesday, came after my last post prior to this one, posted at 11:50Z on Tuesday [1], and appears to be replying to it, yet, at least in my browsers, it is presented as a reply to a post I made on Saturday. Furthermore, the post that I am replying to here seems to consist primarily of what looks like a quote, except that the text in this 'quote' does not seem to be found in any of my responses to your article, and nor does it appear to be from any other response to your article, either. If the post I am replying to here *is* intended to be a response to my most recent post prior to this one, and is quoting something that has been written in response to your article, could you provide me with a link to the post being quoted, so that I can read it in context before replying?
[1] https://petemandik.substack.com/p/hail-mary/comment/99614879
it’s not a quote. i apologize for using quote marks for something other than direct quotation.
ok, so you grant that it’s not necessary to have had an experience of red in order to come to know what it’s like to see red? Yes? Regarding Hume, you agree with me and Hume that we can know what it’s like to see the “missing shade” of blue in advance of actually seeing that shade of blue. One way to put the key points here is by reference to this statement, S: “For all x, if one knows what it is like to experience x, then on must either have already experienced x or one is currently experiencing x.” I think S is false. I guess you do too, but I’m not sure exactly which cases you agree with me are counterexamples to S. I think the japanese flag and hume’s blue are obvious counterexamples. Another counterexample is a perfect forgery of the Mona Lisa. I further think the physicalist should say red is a counterexample. You seem at least sometimes to disagree about that last one. But I don’t know. You tell me.
Sure - I will go through my positions one case at a time:
1) I think Swamp Mary is possible, even though (like Botzmann Brains) beyond-astronomically unlikely. She provides no support for the premise that Mary will KWIL prior to her release, however, as she has not arrived at the state of knowing what it is like to see red as a consequence of studying propositional knowledge.
2) With regard to Crane's claim (at least as you have made use of it here), I assume that if you have experienced pain - *and can remember something about what it is like* - then you will generally be able to tell when you are not in pain (though people who have profound dementia may lack the self-awareness to recognize whether or not they are in pain or hungry.) As this depends on having had the experience, however, I do not see how it might provide any support for the premise that Mary will KWIL, prior to her release, as a consequence of reading anything.
3) The Japanese flag case appears to me to be very much like Crane's argument, in that it depends on a prior experience of the right type: I suppose that a person who has not experienced redness will learn something about what looking at the flag is like on first seeing it, no matter how thoroughly it has been described to them beforehand.
4) Hume's color-interpolation hypothesis seems plausible to me, and though I'm not entirely convinced that it is correct, I have said that I will assume it is for the purposes of this discussion. Even then, it provides no support to the premise that Mary will KWIL prior to her release, as it depends on knowing what the bordering colors are like. As far as I can see, there's no reason to think that, just because Mary could do this once she has seen these colors, she could arrive at a state of knowing what those colors are like solely as a consequence of studying propositional knowledge.
5) I agree that S is false, but I do not see how its falsity provides any affirmative support for the notion that Mary could come to know what it is like to see red solely as a consequence of studying propositional knowledge of any sort. My own objection to the knowledge argument, like several others, neither depends on nor implies proposition S.
Ok, great. I just needed the part in 5 where you said that S is false. I didn’t learn anything I didn’t already know about you from the rest. Now I’ll repeat my question I earlier asked you, and didn’t notice any answer, but will this time avoid the befuddling and distracting phrase “only one way”. WHY must Mary experience red in order to know what it is like to see red? What set of premises, consistent with physicalism, collectively entail that Mary must experience red in order to know what it is like to see red?
On rereading your article, I think I now see where you are going with this line of questioning: the knowledge argument never explicitly rules out the possibility that Mary could come to KWIL by some means other than experiencing colors. Instead, it tacitly assumes that proposition S is true (without this tacit premise - which is not the only one here - or something equivalent, the premise that Mary learns something on her release lacks justification.)
In this view, your objection (if I'm understanding it correctly) is quite similar to Daniel Dennett's first response to the knowledge argument, which was to say that we cannot imagine, let alone deduce, what knowing all the relevant scientific knowledge would entail, and so the knowledge argument's proponents are merely relying on their probably-biased intuitions when they assume Mary will learn something about colors when she first experiences them (this is the argument in which he introduces the blue banana trick.) The philosopher and knowledge-argument proponent Torin Alter concedes that this is a legitimate objection to it.
The problem for Dennett was that those who were on the fence generally found proposition S to be very plausible, and I think you will find the same. In your case, all but one of your four arguments against S depend on bootstrapping from some initial KWIL, and the fourth (Swamp Mary) requires an extremely improbable event, while the knowledge argument supposes that Mary will always (or at least almost always) learn something about seeing colors on first experiencing them.
Dennett later turned to the argument found in 'What RoboMary Knows', in which (at least in my reading of it) he argues that the widespread acceptance of S (at least under the 'rules' of what Mary can and cannot do in isolation, which is a topic in its own right [1]) is merely on account of physical limitations, imposed by our brain's architecture, on what Mary can do with her knowledge of physics.
[1] Amy Kind, What Counts as Cheating? Deducibility, Imagination, and the Mary Case https://link.springer.com/article/10.1007/s11406-024-00717-5
Note: I wrote this some hours before my other reply to the parent post.
So you needed to hear that S is false? Well, I am happy to have been able to do that for you, and maybe I will have further opportunities to say things you need to hear!
My response to your question "why must Mary experience red in order to know what it is like to see red?" is that she doesn't (an answer that should come as no surprise, given that I have just said S is false.) For example, given so-far hypothetical tools for selecting and modifying the physical state of individual neurons and synapses without doing any damage, and the knowledge to know what needs to be changed, I suppose that one would be able to create false memories in Mary's brain of seeing things in color, and from then on she would know what seeing those colors is like to the same extent as she would if she had actually had the experiences that the fabricated memories seem to be of.
On the other hand, I reject, as an alternative to experience, the notion that Mary could come to know what it is like to see red solely by learning a body of propositional knowledge.
I can only vaguely guess where you are going with this line of inquiry, and I am eager to see where you take it next.
Since you seem to think it’s important, i’ll focus on this. I am talking about language because someone else made a claim about different senses of “knowledge” . If people want to make claims about senses of words, they should appreciate that they are talking about language, and further, that how one determines whether a word has multiple senses has been a subject of intense and serious investigation for many years by linguists and philosophers of language. It’s not something serious people just pull out of their …hats. If my dialectical opponents don’t want to talk about language, then they shouldn’t have brought it up in the first place. if the language-phobe wants to conduct the investigation of the mary argument solely with minimal explicit mention of language and logical entailments of words and phrases, they would still at least need non-laughable operational definitions of terms like “knowing what it’s like to see red” and “seeing red”. I won’t say that it is impossible to do that, but no one yet has. And I venture that it would be exceedingly difficult, as is evidenced by the many years of research that has gone into investigating the closely related “Molyneux’s question” scientifically without anyone turning up any settled consensus on the answer. (Besides looking up “Molyneux’s question,” if you’re interested also look up “Molyneux’s babies”.)
Thanks - given that you can speak authoritatively on the matter, it is helpful for me to have your confirmation that no-one has yet given operational definitions of terms like "knowing what it is like to see red", let alone one which would justify seeing the knowledge argument as defeating physicalism.
Fully agree. Churchland nailed it way back then. I don't think that equivocation on "know" is the only flaw in the argument, but it is enough.
it would need to be shown that it’s equivocal, not merely suggested.
I think Churchland achieves that, but that's where the tautology of needing the brain-state of KWIL comes in.
Knowing in the usual sense: having the desired physical state, the KWIL-state.
Knowing in the pre-release sense: having good knowledge *about* the KWIL-state, in the form of circuit diagrams and neural-net concepts and the physics of the retina, and so on.
That’s just asserting that there are two different senses of “knowledge”; it does nothing to establish the truth of the assertion.
But one of them is *about* the substrate of the other. They have entirely different contents, and entirely different representational relationships to the out-of-head environmental property of redness.
Does one ultimately imply the facts entailed by the other? Yes, of course, if physicalism is true. But KWIL barely counts as factual; it cannot be expressed in propositional terms.
It's a bit like comparing GPT4's internal representation of "redness" with a Word document listing the relevant neural weights. One entails the other. (And of course I am not suggesting an imagistic internal representation, or anything like a redness quale.)
If we had some deficient version of GPT4 that had been specially trained with a redacted corpus, and it had never encountered redness and its semantic associations in text, we would not expect that feeding the Word document to the deficient GPT4 would ever substitute for the missing training. GPT4 doesn't have the right sort of self-manipulation to update itself like that, for fairly obvious reasons, but humans also lack this sort of access to their own substrate.
Obviously, much more discussion is required to flesh out what types of knowledge we expect to be inter-convertible in humans, but almost everything we can say about these two knowledge states suggest that they have a very different set of epistemic properties. That alone is enough to consider them distinct types of knowledge, and their fundamental neuroanatomical separation reflects that the difference between their epistemic properties is best explained in physical terms, without reference to the ontological relationship between their contents.
And of course, this is all only an assertion if the full discussion is not attached.
But each comment has to stop somewhere, with important things still left unsaid.
A difference in what they’re about doesn’t entail a difference in sense of ‘know’. I can know what time it is and know where I put my lunchbox. It doesn’t follow, or even suggest, that I’m knowing in two different senses of “know.”
Regarding physicalism, my take has long been that the setup to the thought experiment assumes its purported conclusion, non-physicalism, by implying that Mary can have all the physical information and yet still subsequently learn something new. If we remove that assumption, then it becomes an assertion that Mary can have all the information but learn new information after seeing red, a contradiction.
The question of whether it's possible for her to KWIL with all the possible third party information is a more interesting question. Swamp Mary is a nice move! It seems similar to Daniel Dennett's RoboMary putting her brain in the same state as if she'd seen red. In both cases, KWIL is achieved without the experience of seeing being in the causal history. Nice!
I also think it's worth noting that even a Mary only with contemporary knowledge could guess a lot about what she'd see. She'd know that reds blend into oranges, which blend into yellows, greens, cyans, blues, etc. She'd expect red to jump out more than green or blue. And she'd be expecting it as a commonality for roses, blood, and ripe strawberries and a strong distinction from the sky, grass, or ripe bananas.
I think most of our intuition about how she'd react seems more based on how we'd react. But without her knowledge, out intuition doesn't seem useful here.
Thank you! My hunch about where people get their broken intuition about needing to see red is that they over-generalize a folky ocular epistemology: things that are farther away are harder to see and thus know about, so if they are brought so close as to be all the way “in here” [points to mind with finger] then that’s the exemplar of ultimate knowledge: having the thing known all up in it (the knower).
Interesting post, but I'm one of those physicalists who disagrees with you.
I think we need to distinguish between knowing what redness looks like in the usual way, and knowing what it looks like in the bland scientific way that Mary can achieve. Both of these concepts can point at the same underlying ontological truth, but transitioning between them is a physical process that can indeed be impossible.
I agree that swamp-Mary breaks the link between seeing and KWIL, but she is also astronomically implausible. We could break that link in several other ways: pre-release Mary could be handy with a nano-scalpel, and just give herself KWIL for redness. Mary could be a mutant that is capable of cognitive transitions denied to normal folk. She could have a neural prosthesis that achieves the KWIL state without employing photons of the requisite wavelength.
What we can't do is break the link between KWIL and achieving the right brain structure for KWIL.
Physicalism is not at all bothered by the fact that brain-states can be out of reach during access to accurate descriptions of those brain-states.
I'll be posting on this topic myself soon.
But I agree with the idea that physicalists should love Mary. In the right context, this thought experiment is strongly supportive of physicalism and actually casts anti-physicalists as fairly gullible.
assuming physicalism in the first place, It’s an uninformative tautology to say that you can’t KWIL unless you have the right brainstate for KWIL. But what reason is there for the physicalist to think that book learning can’t get you into that brain state?
I think it is not so much a tautology as a shift in focus from the contentious to the obvious. It is the shift that is important, not the fact that it lands on a truism. It is also a truism that swamp-Mary has KWIL, for instance, but it seems to make a point worth making, because it breaks the link between seeing and KWIL.
Say we have concept 1 (C1) and concept 2 (C2), and some dispute about whether C1 implies C2. Do we need to refer to the truism that brain-state of C2 is necessary for C2? Usually not.
C1 could be the concept of "2+2" and C2 could be the concept "4". Nearly all discussions about whether "2+2" implies "4" rightly interact with this at the semantic level, ignoring the actual brain-states C1 and C2. There could be a post-torture state in which Winston Smith from 1984 follows a different transition, but this is usually not relevant. There could be a genetically engineered species that was otherwise intelligent but could never achieve this transition because evil experimenters had wired their brain strangely for shits and giggles. They might think "4" was of divine importance.
But what if C1 is full knowledge of redness at the physical level, and C2 is the redness KWIL? Here, the focus should not be on the semantic truth of whether the content of one brain-state entails the content of the other; the primary focus must be on the actual physical transition (which in turn leans on recalling the truism).
The C1 to C2 transition is ultimately not dependent on the truth of what those concepts represent, and barriers of this nature abound in the brain's examination of its own cognitive substrate.
Sure, we need an actual physical reason that C1 does not lead to C2 in Mary's case, but the reason is there in the neuroanatomy. Patricia Churchland, for instance, has a Youtube vid on the relevant neuroanatomy. I will post on this issue soon with a more extensive consideration of the actual physical barrier involved.
EDIT: I didn't knowingly use the same mathematical analogy as in the other thread, but I saw it when replying to your other reply, so it is funny that the threads converged like this.
You won’t be the first to suggest that neuroscience can shore this up, but I haven’t seen anyone, including Pat, make good on the suggestion. I look forward to your attempt.
It might be a couple of weeks. I have actually written a hundred-page version of my ideas on this issue, but that's too long to post.
I have my personal terminology for the spectrum of opinions on this. It starts with "Gap Denialism", which might sound pejorative but is not intended that way, and it ranges through various forms of "Gap Compatabilism" through to full-blown "Strong Hardism".
You seem to be a Gap Denialist, or maybe a Gap Not-yet-convincedalist. (The latter is not really on my list.)
I deny all the gaps.
So I see! We should resume discussion after my own post.
The logic of the conclusion seems to require Mary can know ALL of the physical facts about seeing red, at least to reach the contradiction if Mary learns a new fact by seeing red, because she already knew ALL the facts about seeing read.
But I suspect that this "knowing all of the physical facts about red" contradicts physicalism itself, at least in the physical universe we live in, because all the facts about seeing read doesn't sound like it could be a finite number of facts.
Facts that are physical, in this physical world, each require a minimum amount of mass and energy to instantiate, according to physicists as I understand it. So it seems that an infinite number of facts can never be accessible to a being, in a finite period time, because they will will be spread over an infinitely large region. Any overly high concentration of facts will collapse into a black whole, in fact, like any other constructions of matter and energy.
So, at least in this universe, I think this refutes the logic of the argument that derives a contradiction if Mary learns a new fact by seeing red, because she already knew ALL the facts about seeing read.
I don't actually think this is just an optional quirk of this particular physical universe. On the contrary, I suspect imagining a world in which infinitely many facts can be "known" physically to a physical being of finite extent may be the same as negating the difference between physical and non-physical facts. That is, I think such a world would not really meet our intuitive concept of physical, as well as being very unlike a physicists concept of physical (I assume that is what counts as the non-intuitive concept of "physical".)
For example, once you can have an infinite amount of information in a finite space, you can have, for example, life forms evolving at all possible scales. There is no bottom limit to the intricacy of structures of information, like DNA and cell mechanism. As beings in that world, we would immediately be eaten from the inside by much smaller, more information dense new life that evolves faster than we do, and that new life would already be eaten alive by new life smaller than it that evolves faster than it does, and so on... beautiful, but maybe hard for anything to reach the philosophizing stage, too.
Assuming Mary knows ALL the physical facts is a simplifying assumption so that the argument can be stated in a concise way. Mary only needs to know as many physical facts as we would plausibly expect to be relevant to the best physcalistic scientific theory of an empirical phenomenon. No other scientific theory has ever required a scientist to have godlike knowledge. There are facts about the numbers of planets outside of our lightcone that couldn’t possibly be relevant to understanding human color vision. Mary knows all the physical facts that could plausibly be relevant to a scientific understanding of human consciousness. Assuming that knowing what it’s like to see red is relevant to understanding human consciousness, and that all the plausibly relevant physical facts are knowable by book learning, is a physicalist theory of consciousness necessarily incomplete?
In the argument, not specifying fully "all" the physical facts removes the logical force of contradiction from Mary gaining a new physical fact, so it can't be claimed to prove something, I guess.
I find it interesting that the logical contradiction can be considered to support either physicalism or anti-physicalism, depending on which term you decide the contradiction applies to.
I.e. if you decide it supports that super-Mary can never know the experience of red until she experiences it, that might (maybe) make you conclude for non-physicalism.
Or if you decide it supports that super-Mary must have already known the experience of red, then you might conclude for physicalism.
Removing the force of "all", then you move more into the discussion of the plausibility or content of KWIL, for example @disagreeableme's considerations of what in reality the physical facts of the experience of seeing red might be and how they could be really represented or learned by Mary. A super-Mary who can change her own neural network at will would be able to "learn" the experience of red this way, without requiring access to a physically impossible "all of the physical facts", so this Mary seems physically possibly, just implausible, which is fine for a thought experiment.
i gave you a recipe for rewriting the argument to yield a slightly different but still valid argument, but you have to apply it to more than just the first premise.
For me the knowledge argument kind of misses the point, the thing I actually wonder about is why there is subjective experience in the first place and how it works.
I’m going to go from the theme that The Good Determinist provided earlier — I think I’d need to be the bug or whatever that experiences ultraviolet vision in order to effectively “know what it’s like”. And like him I don’t think this means that Jackson’s original thinking on the matter mandates the existence of anything non-physical. But Tom only provided an intuition to this effect. I’ll provide a theory.
My theory is that by conflating “ontology” (or what exists), with “epistemology” (or what’s understood), the thought experiment beguiles us to make a category error. Notice that if I were some sort of god with a perfect grasp of the systemic causality under which I exist, then I’d inherently know what science doesn’t know, which is to say the mechanics behind seeing red, dragonfly vision, and so on. But this epistemology in itself would always remain just that, which is to say understandings of how such dynamics work. No epistemology should ever exist as actually “seeing red” or anything else ontological, and regardless of the effectiveness of that epistemology. I can go the other way with this reasoning as well. People who actually do see red can obviously still be clueless about what causes them to do so.
It’s interesting to me that philosophers have fully separate classifications for studying “what exists” versus “understandings of what exists”, and yet have had a difficult time grasping the category error in this thought experiment.
yes, begging the question. no, not useful. it’s a leftover from an ancient and very naive epistemology whereby to know stuff you have to have it in front of your face so you can see it really well. the past several hundred years of progress in math and science have blown that epistemology way out of the water. As descartes pointed out you can perfectly comprehend the difference between a circle, a 1000-sided polygon, and a 999-sided polygon even though you could neither see nor visually imagine the difference
Either (1) it’s a general rule that knowing what it’s like to have an experience requires you to have that experience or (2) it’s not a general rule. If (1), then you have to deny, in the face of arguments and intuitions to the contrary, cases like Hume’s blue and the Japanese flag. If (2), then you incur the obligation of explaining what generally distinguishes the cases that conform to the rule from the ones that seem to be exceptions to it. Why should we credit your hunch about ultraviolet? How do you know that it’s true? It’s a claim about what will or won’t happen in the future, and those are generally hard to be certain about unless they’re just grounded in some definition (like “no triangle will ever have four sides”). But there’s no plausible definition that anyone’s floated that’s going to help you here. So do you pick (1)? Or (2)?
You can put me down for (1) Pete. Yes I can imagine a Japanese flag if someone describes it, but that shouldn’t be the same as actually seeing one by which I would KWIL. And I can certainly imagine the missing blue if I’m presented with a spectrum of blues, but seeing the blue should be different. What’s the difference? The ontological consciousness of “seeing” versus “conceptualizing”. (It should be easier to demonstrate this once science grasps the causality of consciousness, though as a strong naturalist I at least presume such a difference.) Someone who merely knows the mechanics behind how their brain theoretically could create the experience of red, should not thus experience red. This is not to say that Mary should be said to “learn something new” with such an experience — Jackson used a sleight of hand there (and I think later recanted whether or not he grasped the associated category error). Instead here Mary should be said to “experience something that she never had before”. This point is demonstrated again when people who know nothing about the brain can still experience red and thus KWIL during those experiences. (Later they may have memories of knowing what it’s like, or MoKWIL. 🤪 ) Epistemology should not be confused with ontology, though Jackson beguiles us into this category error, and not entirely unlike the one of positing which color triangles happen to be.
I don’t see why imagining would be relevant for whether you KWIL. Before you imagine it, you need to have prior knowledge of what to imagine. And that is knowledge that you retain between imaginings, and even while you’re unconscious. And knowing what’s it’s like is something you use to correctly judge “i am not currently having a red experience “
That’s a great point. Yes in practice I use my memory for the concept of red, which is to say an epistemological understanding of it even when I’m not currently experiencing an ontological red. But I should be able to adjust my argument. Notice that if I had never had an ontological experience of red then I’d never have gained a memory of it that I might use epistemologically for such a general understanding. So I still consider there to be a category error in positing an experience by means of an understanding of what creates an experience.
Or to put my position here more bluntly, let’s say that I’m held captive and tortured for the rest of my life. To say the very least, that would fucking suck! Let’s also say that Jack is with us, the pain scientist who knows all there is to know about how the brain creates pain, as well as everything that these sadistic bastards do to inflict it upon me, though he has a genetic defect such that he can only ever feel neutral to amazingly good and so has never felt anything that even approaches “pain”. I can see how one might argue that he could think about the difference between feeling neutral versus wonderful and so posit pain as an opposite that thus feels “bad” rather than the good that he knows so well, but unless he were to actually feel pain (like Mary who ultimately sees red), I don’t think he’d be able to grasp my actual misery. In any case it seems to me that knowing everything about the neurology that creates my horrible existence, wouldn’t even provide the backwards conceptual association between pleasure versus not. The neurology behind what I experience should not give him any sense of my pain because epistemological understandings should not create ontological experiences.
i’m not quite following the argument here but suspect it’s trading on an ambiguity concerning “memory of”. Do I have a memory of Columbus arriving in North America in 1492. In one sense of “memory of” I clearly do not, since I wasn’t born until 1969. In another sense I obviously do, since I learned it a long time ago, and still remember it. A crucial question is which kind of memory is involved in knowing what x is like. There are lots and lots of cases in which it’s quite clear that you know what x is like even though you have no history of contact with x. I know what it would be like to look at a perfect forgery of the mona lisa, although I haven’t seen one yet. SOOOO, what reason is there for a physicalist to believe the experience of red must be different in this regard? I know lots of people find it to be intuitive that this must be true. But what argument is there for believing that intuition? I haven’t encountered one yet that isn’t based on some fallacy like begging the question or equivocation.
Yes it should be important to refine the concept of memory here since everything we can say one “knows” will in some sense exist in the form of a memory. For the moment I consider it useful to define this broadly as “past consciousness that in some sense remains for present use”. And as you’ve mentioned, you’ve never had a consciousness of Columbus actually arriving in America, but rather various memories which suggest that he once arrived. Maybe it was a hoax? Regardless it’s the various suggestions you’ve encountered over the years that have given you such a memory rather than such an event itself. That would be the ontology which provided your epistemology. This doesn’t suggest to me that if I knew the brain mechanisms which create an experience of red (epistemology) then that would also give me the experience of seeing red (ontology). I’d think you’d need a positive argument that one must lead to the other to build a good case against mine, though in the end I can’t fathom a natural possibility.
On the Mona Lisa, it seems to me that unless I’m looking directly at the original, then any version that I do see will be a copy. But reasonable copies ought to give me a good sense of the original. In the examples of an amazing color scientists who never sees color, as well as an amazing pain scientist who never feels anything except good, I don’t see how they’d get a reasonable sense of the red or the pain that they never actually experience, that is until they do.
Perhaps you’ll say that I’m begging the question by defining knowledge of what exists differently from experience of what exists? But did I arbitrarily define these terms this way, or are they generally useful to define this way?
A possible response to Swamp Mary (similar to the point by Bill Robinson) is that her KWIL to experience red is still only latent, just like Mary’s in her black and white room. She only would KWIL if she activated that memory by recalling the apple in her mind’s eye. Or to put the concern another way, what if Mary (and therefore Swamp Mary) had aphantasia? Would aphantasic Swamp Mary really KWIL to see red (before her next encounter with a red thing)? At her next encounter, her reaction might be “oh right, *now* I remember WIL to see red”, but is that really any different from Mary’s supposed “so *that’s* WIL to see red” reaction when she emerges from the room?
What reason is there for thinking “latent” knowledge isn’t knowledge? I know all sorts of things over long stretches that are punctuated by episodes of me being unconscious. Having knowledge is like having a memory; whereas the direct analog to deploying that knowledge is recall, remembering, or recognizing. Imagery seems a red herring when discussing knowledge here, as one would need to know what to imagine before they imagined it.
Thanks! I guess my point - not particularly earth-shaking - just comes down to Swamp Mary not being a knockout punch for those that hold to a distinction between Russellian knowledge by acquaintance and knowledge by description. I'm not actually one of them, but just playing devil's advocate. Also I enjoyed giving Mary aphantasia just for fun :-)
what it’s a knockout punch for isn’t Russellian knowledge by acquaintance (KA), but instead the related view that KA is part of the analysis of knowing what it’s like. Swamp Mary is one of several counterexamples to such analyses.
Let’s meet Harry, Mary’s cellmate.
They both were locked in the room for all those years and both acquired the same BLAD. They’re released at the same time and see the same apple. Then they’re told to close their eyes and describe what they’ve seen.
Mary, as mentioned, has aphantasia. Harry, on the other hand, has hyperphantasia. Let’s stipulate that they have the same knowledge by description (KD) of the apple - that is, Mary can describe the apple and its properties (as well as all the physical facts about what happened in her brain, etc when she saw the apple) in the same detail as Harry when asked. One might be inclined to say that, in addition to their shared KD, Harry also has (or seems to have) KA of the apple (well I might be inclined to say that… not sure if you would?) Am I understanding you right that, for purposes of KWIL, when they close their eyes there isn’t an essential difference between them? (Or please set me straight)
Thank you again - I’m a fan of your work and grateful for the dialog!
thank you! I’ve enjoyed it too. and you’ve interpreted me correctly
Thanks - you're a rock star! (And I mean a cool rock star, like Tom Petty, not an annoying one like Chris Martin. There are other philosophers of mind who remind me of Chris Martin, but I won't name any names.)
Shots fired!
I take this as a breach of the terms of our truce, and so regrettably have had no choice but to respond in kind.
https://disagreeableme.substack.com/p/failed-mary
Nice discussion, but I’m not satisfied. The main problem is that you have Mary knowing red as a highly complex, brain event property, but there is nothing about red in ordinary experience that remotely suggests any such complexity. So, Mary may know a lot about red, but there is still something she doesn’t know, namely how red appears to those who have ordinary experiences caused by ripe strawberries, bullfighters capes, etc.; (or, for those who like the WIL formulation, she doesn’t know WIL for trichromats outside the room to see a ripe strawberry).
Swamp Mary still has duplicates of the neural effects that seeing something red caused in Mary. If (post-release) Mary can generate a lively memory image of a ripe strawberry, so can Swamp Mary; which does not support physicalism unless consideration of Mary alone does. If (post-release) Mary cannot generate such an image, she may still have a capacity to classify a new experience correctly as ‘red’. If so, Swamp Mary will also have that capacity. – My conclusion is that Swamp Mary can’t really add anything to what can or can’t be shown by considering Mary herself.
Bill Robinson
I just don't grant anti-physicalists the initial presumption that anything about "ordinary experience" lends itself towards the view that there's anything "it's like" (whatever that means) to see red in the first place. That is, I don't just deny that people have qualia, I deny that the notion that we have qualia/phenomenal states is part of ordinary experience. I think the belief that we do was invented and perpetuated by philosophers.
As such, in the case of Mary, I think the thought experiment will typically appeal to people influenced by certain philosophical traditions, and that its appeal is an artifact of training and enculturation, not some shared feature of our psychology.
big agree although my take on “is like” is slightly different. knowing what x is like is one and the same as knowing something or other about x. So, if I know that ruffles have ridges, then I know what ruffles are like. If I know that water is wet, then i know what water is like. This works exactly the same when you plug “seeing red” or “experiencing red” in for x. So, if I know that Russel saw red on a Tuesday, or that seeing red makes your c-fibers fire, I know what red is like. Somewhere around here my opponent will say something like “no, stop, you’re not using those words in our special, proprietary, Nagel-blessed way” at which point I automatically win.
thanks, Bill! However, lots of ordinary (non-future-science) experience reveals that red is complex. all reds are complexes of hue, saturation, and brightness. red resembles orange more than it resembles green, and such relational facts about red are implausible to suggest are merely accidentental. red is a secondary color in cmyk mixture schemes (and is mixable from magenta and yellow). shades of red seems darker than equiluminent shades of yellow. future science is unnecessary for buoying up the premise about knowing phenomenal red by description
Yes, red has _some_ complexity -- hue, saturation, and brightness. (The other things you mention are its relations, not its internal complexity. E.g., I am brother of John Robinson, but that, _pace Leibniz_ doesn't make me more complex than I would be without him.) However, the kind of complexity Mary must be aware of is myriad neural firing rates, and rates of change of those rates, and so forth. Red does not have _that_ kind (or, degree) of complexity (otherwise, Ancient and Medieval philosophers would have written about it). --- As to knowing by description, I think congenitally blind people often know a great deal _about_ what red is, but like pre-release Mary, they don't know what red _is_. Yes, physicalists can consistently claim to know a great deal _about_ red, but they do not have the actual property, red, in their inventory of what there is.
I get that that’s an antiphyscalist view of what red is, but it can’t be an input to a non-question-begging argument against physicalism. Anyway, I don’t know what the internal/external distinction would mean wrt color aside from essential/accidental. And more to the point, if one’s descriptions nail down a color’s essential properties, one cannot be accused of leaving out something essential .
I don’t think it’s question-begging to take it as a datum that when we are, in ordinary circumstances, having an experience caused by, e.g., looking at ripe strawberries, there is nothing in our consciousness that is anywhere near as complex as what would have to be in pre-release Mary’s thoughts of neural event structures, if those structures were to have sufficient complexity to make them correlates of, say, red in particular and not just, say, some sensory input or other. I agree that internal/external for a property like red probably is the same as essential/accidental (provided that causes are taken as contingently related to their effects). But “if one’s descriptions nail down a color’s essential properties” is a big if. It would be question-begging against my view to assume that that conditional is satisfied, because on my view, what pre-release Mary knows about is, e.g., the causes of color experiences and not the essences of colors at all.
The claims that are striking me as question begging against the physicalist are, first and foremost your earlier “[physicalists] do not have the actual property, red, in their inventory of what there is” but possibly I’m misunderstanding whether that’s supposed to be part of your argument or just a statement of your view. I didn’t think the complexity claims were specifically question-begging, but we still disagree about them (and it’s an interesting disagreement). How many properties, essential or otherwise, does someone need to know about x to know what x is like? It’s implausible that one needs to need to know all of the properties. I know what it’s like to taste Cabernet, although a sommelier probably knows way more about what it’s like. Probably if someone knew zero essential properties of x and only accidental properties, it might be a stretch to say that they know what x is like. So back to pre-release Mary. Without violating the rules of the thought experiment, I can allow her all sorts experiences, as long as none of them are what both sides of the debate would call “red experiences”. So Mary can know all sorts of stuff “by acquaintance” via seeing other hues, etc. She also knows ALL sorts of cool correlational stuff about which neural modulations go along with which experiences, just none that acquaint her with so-called “red experience”. Both sides of the debate are obligated to describe this in neutral non-question-begging ways. Just like the physicalist cannot assert at this point that these correlations are actually identities, nor can the anti physicalist assert that these are mere, non-identity, correlations. The main residual disagreement between the two of us seems to be whether the complexity pre-release Mary accesses is complexity enough for knowing what it’s like to have a red experience. Your saying she has nowhere near enough needs to be bolstered by something non-question-begging. Why does it fall short? Simply because it’s not acquaintance? Or is there more that can be said?
“Physicalists do not have the actual property, red, in their inventory of what there is” is a statement of my view, and absolutely not a premise.
I’d like to hear more about just how you disagree with the claim that ordinary experiences of, e.g. red do not have any hint of the degree of complexity that would have to be in the content of pre-release Mary’s thoughts about the neural activation events that are the red experiences that she knows others have on occasions of their being exposed to, e.g., ripe strawberries.
I don’t think sommeliers know any more than I do about what it’s like for me to experience sipping cabernet. (I do think their training may well have caused their experience while sipping cabernet to be different from mine – although not entirely dissimilar.)
I think the original experiment had pre-release Mary in a room with only black, white and shades of gray. I agree that one might get to know what it would be like to experience Hume’s missing shade of blue, in the conditions he imagined; but I don’t think pre-release Mary comes even close to having that kind of background.
I make no appeal to acquaintance; I have a paper titled ‘Dispensing with Experiential Acquaintance’. https://www.william-s-robinson-consciousness.com/s/DispExpAcqPostprint.docx.
I think it’s a phenomenological datum that when you look at a ripe strawberry, you don’t thereby have anything in your consciousness that has remotely the degree of complexity that would have to be in pre-release Mary’s thoughts, if those thoughts were adequate to be specific to the brain events that people (or just one person whose brain scans have been intensively studied) have had when looking at ripe strawberries, bullfighters’ capes, and so forth. I think that’s obvious, but there is also an argument that our Ancient and Medieval forebears were certainly smart, and none of them gave any hint that they found experiences to have the kind of complexity that would have to be in pre-release Mary’s thoughts.
I agree that statements like ‘red is more like orange than it is like blue’ are necessary truths, but I don’t think that knowing a humongous number of such statements would bring anyone closer to knowing what any of the mentioned colors is (or, WIL to experience them). I think congenitally blind people could know a huge number of such truths and not know what red is.
Typically, when I think about what it means to "know what it's like," to have some experience, I interpret that as meaning something like, "I can simulate the experience in my head (e.g., I can see the color red in my mind's eye) in a way that feels similar enough to having the real experience," so I can point to that and say, "That's what it's like." The Mary's Room argument only works by conflating this sense of KWIL to a sense that's more related to knowing some specific proposition about the experience of seeing red. The latter interpretation seems to be the one you have, and on this interpretation, physicalism does imply that Mary should be able to KWIL to see red even before seeing it herself, but, as you explained in the article, this is no problem for physicalists. There's no reason to think she couldn't KWIL, unless you already reject physicalism, and on physicalism, it seem perfectly reasonable how she would come to know.
Under the former interpretation, it's not true that learning all the physical facts about color perception would let her KWIL to see the color red, since, after all, there's no necessary connection between knowing some information and being able to perform a mental simulation that generates a particular experience at will. Even if the information you know is about how that mental simulation works neurologically, that doesn't magically grant you the ability to just make it happen. But of course, all this mean that, under this interpretation, physicalism doesn't imply that Mary would KWIL to see the color red before she actually sees it.
So we have two interpretations. On one of them, the first premise of the Mary's Room argument, that Mary learns some new fact when seeing red for the first time that she couldn't have learned by reading about it, is false, while the second premise that physicalism implies that she should be able to learn WIL to see red just by reading about it is true at least in principle. On the second interpretation, the first premise is true, but the second premise is false. On no interpretation are both premises of the argument true, so Mary's Room fails to disprove physicalism.
I am a physicalist and a physician, but I don't agree at all. I love Mary too, but because it illustrates the following:
1. differing physical causes (coloured light, monochromatic light) lead to different physical effects (brain states). It really needn't be more complicated than that. Changing the stimuli changes the downward effects, if there is any sensitivity to the change in stimuli. It's incredible to me how frequently this is missed.
2. Mary is an expert on ordinary color vision in ordinary humans. She hasn't a clue about her personal brain state - a brain that has been deprived of colour stimuli entirely. She has a very different brain than you or me. If she were a super scientist specialising in her own brain, then she could predict her brains future physical states. Which is obviously not the same as being in that state - but still, she could, in principle, predict what she would say if she hadn't made the prediction (since predicting is a physical process this affects the future physical states).
3. Mary certainly doesn't see the world in black-and-white. She sees no lack of colour in her room, a lack which is crucial for the B&W "feel". And if she can distinguish colours properly when let out (questionable), then she certainly won't experience the redness of red that you or I are familiar with. Every experience is created in light of previous experience.
3 is a really good and under-appreciated point.
2, “…she hasn’t a clue…,” is flat out false.
1 is a red herring. I don’t think any party to the debate is denying or missing 1. Seems instead like a totally obvious point that has no bearing on the discussion.
Thank you for the article and for your response to my somewhat sloppy comment, you deserved better.
3. Thanks, I'm glad to hear! Have you or anyone else pushed this point? (I presume I'm not the first). I just posted "Mary in the Coloured Room" (https://markslight.substack.com/p/mary-in-the-coloured-room) which doesn't quite push this angle exactly, but a related one. I'm soon going to post "Mary Returns", where she several decades after release, with no memory of experiencing any lack of colour in the room, returns to it. Check it out if you're interested.
2. Fair enough. Of course she can have a clue! My bad. But as far as I can see the point I was trying to make stands: there's a crucial difference between a super scientist specialised in the olfactory and gustatory neurophysiology of sommeliers or a super scientist specialised in the olfactory and gustatory neurophysiology of people who av never ever smelled or tasted wine. And if you're a super scientist specialised in sommelier neurophysiology, you're an expert at predicting sommelier reactions and responses to various wines. There's simply no way you can accurately predict your own response to a particular wine, from your immense knowledge of sommelier neurophysiology. Or what am I missing?
1. Okay. Maybe it's a red herring among physicalists (I'm in no position to say), but is it really, in general? There must be something obvious I'm missing here. Knowledge is physical, and aquiring knowledge is a physical process brought about by physical inputs. I don't see how it could ever be a threat to physicalism that input A (black letters on a white background) does not put Mary's brain in the same state as input B (being released). To me, the word "knowledge" here is a red herring, just confusing matters. For Mary to be able to acquire the same brain state as in post-release, she would either have to perform brain surgery on herself to put her brain in exactly that state, or she would have some strange neurophysiology that does the equivalent - in which case she would be very non-human and she might not end up KWIL to see red for ordinary humans. As a mere physician not well versed in philosophy, I am of course humble to the fact that I could very well be misunderstanding you. To me, at least, non-physicalists seems to miss this distinction all the time.
Re 3:
Definitely see this:
https://www.cambridge.org/core/books/abs/qualitative-consciousness/theres-something-about-david/ACCEE53B0F26CC9019F8E1579507762E
But maybe also this:
https://philarchive.org/rec/MANP-8
Re: 2 and 1. I can put myself in the right brain state to know all sorts of things about Jupiter without flying all the way to Jupiter or even looking at Jupiter through a telescope. What reason is there, besides a hunch, for a physicalist to believe knowing what it’s like to see red is relevantly different? I’ve given arguments for why the hunch is false. All the arguments I’ve seen so far for believing the hunch is true are long-winded restatements of the hunch, not positive evidence for the truth of the hunch. Stay tuned for another post where I spell this out even further.
3: Thank you very much! I'm afraid I can't access the papers. I will try from work.
2 and 1 - Thank you for your response! I have read your arguments in this post and done my best to understand them. I must say that it seems to me that if you agree with 3, then that is not consistent with your objections to 2 and 1.
But first: I'm humble to my limitations due to having never taken a single philosophy course so far (I'm starting in May). But my perhaps narrow physician-physicalist perspective is that you in this post and responses to me are ignoring or not properly addressing important physical and biological objections to your view. As long as Mary is even remotely human, which is reasonable to presume, black text on white paper cannot put her brain in the same state as building a repertoire of "automatic" visual colour data processing that you and I have trained over many years. This is not just a hunch. It's a reasonable conclusion based on what we know about physical, biological, cognitive systems.
Thus, Mary cannot KWIL for ordinary humans to see red by any other means than building that repertoire. Similarly, you cannot train a co-joined LLM+visual processing AI to recognise and report on colours merely by training it with text token data. The LLM may report that it can imagine seeing something red, but it's not going to be able to draw or recognise something red reliably. That requires experience (training). Ultimately I think these are physical/biological/computational issues, not philosophical ones. Although what counts as "human" and "KWIL" are of course philosophical issues.
Re: 3 and applying it to 2,1:
As I see it, everything is conceptual and relational, including colour experience and KWIL to see a red apple. KWIL red apple requires not only to be acquainted with red stimuli and apples, but also with grey, blue, green and yellow stimuli, and other fruits and objects. It requires the associations to fire, blood, blushing, ripe apples, and it requires the "unlikeness" to green grass, the blue sky, bananas, and the unlikeness to the absence of any visual stimuli. Likewise, imagining a shade of blue one has never seen requires acquaintance with not only other blue shades, but with all the other colours, and with with the concept of shades, to begin with. We have evolved and trained the concept of shade discrimination and we cannot just imagine it without ever having seen any shades.
For these reasons, when Mary is captive, she does not know what black and white looks like to ordinary people. And if someone gave her a red apple in her room, she would still not know what a red apple looks like to us - as her brain would not "automatically" see the likeness with ripe raspberries, blood, and the unlikeness to green grass. It wouldn't just come to her without the appropriate past experience. The same goes when she is released. She learns many new things, but even if she could distinguish red from green, she will not learn what redness looks like to ordinary people (which is, of course, not identical between any two persons). Redness cannot be separated into how it looks on one hand and how we react to how it looks on the other. (If you and I have different favourite colours, that is enough to determine that we do not see colours the same way).
KWIL to see a Japanese flag without having seen one requires familiarity with concepts of roundness, rectangles, white, red, all the other colours which are absent, flags and countries and how they relate to each other. And of course, reporting a successful imagining of flag does ensure the successful imagining of a flag.
You sure can learn a whole bunch of things about Jupiter without even seeing an image of Jupiter. But I think your hunch-objection goes both ways. Even if you then go to Jupiter, and report "yes, this experience is exactly as I imagined it", if brain scans tell a different story, I will trust the brain scan more. That goes for my own self-reports too, of course. But I think you cannot separate out KWIL to see redness from remembering redness.
There's a slightly autistic but I suspect important distinction to be made between KWIL to imagine the japanese flag and KWIL to have seen it. I think we lump these together, for good cognitive/functional reasons, but they're not the same thing. Not sure.
I take it that it follows from your position that if Mary was congenitally blind she could still KWIL to see a red apple?
Looking forward to your next post on the topic! And I just saw you're into meditation too, gonna read that post! thanks.
yes. there is no knowledge that isn’t objective knowledge. in other words, there is no knowledge that requires some particular kind of sensory experience to be had by the knower of that knowledge .
Okay, I see. I find no other way of interpreting you than that you claim than this: I can, in principle, be born into a completely dark, silent and odorless room, kept alive through intravenous nutrition, yet I can by reading Braille somehow train my visual, auditory, gustatory and olfactory cortex so as to be able to distinguish all sorts of shapes, colours, movement, tastes, smells, voices, etc. When I am released I am not at all surprised by anything (provided I have prepared properly), and can navigate the world like anyone else, communicate, recognise colours, flavours, smells, etc, recognise celebrity faces and voices, and so forth. Correct?
In my view, this where I think philosophy that is not rooted in the reality of biology and neuroscience risks being very misleading. Philosophy, knowledge, cognition, all of that is always physical and biological. I would think you agree with that, and I see you're a cognitive scientist too! And as someone who hasn't studied philosophy, I get that I'm probably missing something. But I'm sincerely very puzzled and I fail to see you addressing this! Anyway, thanks for trying to get me to understand and for your engagement. It's quite unique. Namaste! (first time I use that word)
Where was this when I was asking you philosophy questions on twitter? Very well written, also wasn’t aware of your Swamp Mary paper, shall give it a read. (Maestro btw)
The answer literally is “twitter”. Seriously: the present blog post started as a thread on that hell site, see: https://x.com/petemandik/status/1748036918527263042 (Hi, Maestro!)
First, just wanted to point out that doing the Mary experiment is completely do-able today, with a slight variation. If the only light available to Mary is via sodium lamps everything she sees will be shades of orange, including her blood, etc. There was a museum-type place in San Francisco (forget the name, Exploratorium?) that had such a room. On one wall there was what appeared to be a black & white (well, orange) photo of the city. They provided flashlights for visitors, and if you shined one on the wall you’d see it was in full color.
Second, what would happen if the first colored thing Mary sees is an apple which someone has dyed blue? (Think Dennett came up with that.)
Finally, I recently wrote a piece describing how to give AI a qualia of blue. Would be interested in what you think of it if you have the time. https://jamesofseattle.com/2025/02/01/how-to-create-an-ai-with-blue-qualia/
*
We’ve long been able to upload images to ChatGPT; no need relevant to the current question to bolt on an additional “vision system”. It can answer all sorts of questions for you about how the pictured apple seems, what it’s like, etc. I have no idea, though, what would motivate anyone saying that it (or anyone) has got anything direct, infallible, or ineffable.
Uploading an image is not the same thing as experiencing blue right now. But notice the case where the system experiences something, but doesn’t know what. If the system was capable of learning, it might come to call that new experience “red”. But before that learning, how is that experience effable? How is it not direct? How is it not infallible (barring intervention, which would make it hallucination)? BTW, direct, ineffable, and infallible necessarily refer to the system’s perspective, right?
There is some evidence that indicates that indeed people blind from birth and even prenatal more mature fetuses might have visual experiences of color blobs that could resemble those we directly experience (though presumably fetuses may not be entirely deprived of seeing light in certain circumstances, I'd argue). However, even with Mary's precise physicalist knowledge of the brain I cannot see how she would correlate sensory perceptions she does have with any of those blobs as to say which is which, unless there are *very* specific brain functions that correlate phenomenally and spectrally with colors that exclude errors such as color-blindness. In other words Mary would have to know a lot about not just "normal" visual experience that correlates to brain function with respect to colors, but also to know that she is not having an erroneous experience of a kind of color blindness.
"...a complete physical description of seeing red should let someone KWIL it...knowing what it’s like isn’t chained to experiencing it."
Many animals, e.g., butterflies, fish, birds, can see in the ultraviolet spectrum. Can we know (in principle) what it's like for them? On physicalism, there's nothing categorically private about experiences - they're public objects in principle amenable to complete physical description - so on your view I guess the answer would be yes. But it also seems like one would have to instantiate that description to know what seeing ultraviolet is like. Is that necessarily to smuggle in a non-physical ghost?
Can you say some more about “instantiating a description” means? Suppose I know that Jupiter is a gas giant, and thus that to describe Jupiter as a gas giant is to utter a truth. Which of us, me or Jupiter, instantiates the description “is a gas giant”? If I had to guess, it’s Jupiter doing the relevant instantiating here. Is that correct?
Yup. For me to know what seeing ultraviolet is like I'd have to be a system in which the physical description obtained. At least that's my intuition. Reading pages and pages of equations and flow charts and seeing slides and videos of the relevant neural states (e.g., the total conceptual/propositional/quantified description of the NCC of a butterfly seeing in the ultraviolet spectrum) doesn't seem to make me KWIL to see in the ultraviolet spectrum.
It would be nice to have an argument instead of an intuition, given all the reasons to think that the intuition is false.
What are all the reasons to think that the intuition is false?
missing shade of blue; japanese flag; swamp mary. additionally , reasons to think the sensory idea of visible ultraviolet is a complex and thus describable independently of any assumption of physicalism (e.g. being a color, it would be a complex of hue, saturation, and brightness). and more!
So I could know what ultraviolet is like for the butterfly from knowing the exhaustive description of the quantitative parameters involved in its hue, saturation, and brightness, even if I were blind. I don't need to have had any visual experience to know what such experiences are like on your view since KWIL is just to know all the physical facts. This suggests that a system that doesn't have experiences could know what it's like to have them.
Okay, read it.
Firstly, I *do* agree that that Mary's Room doesn't at all disprove physicalism. (FWIW, I have a physicalist metaphysics.) And I do agree that Swamp Mary would know the experience of seeing red because it would be in her memory as obtained from OG Mary.
Where we disagree is about objective versus subjective knowledge. They involve different brain states, and I see no way to accomplish the same brain states involved with subjective experience through objective learning. The only way around it is for Mary to use her expert knowledge to construct a machine capable of generating those brain states. Essentially, to create a VR experience, but that is indistinguishable from having the experience.
Why does Mary have to have subjective experience brain states in order to know what it would be like to have those brain states? It’s the same question we’ve been circling around, now with the phrase “brain states” inserted in it.
Because the only way to have those brain states is through experiencing the phenomena associated with them. You can't induce the retina, visual cortex, and brain activity of seeing red by thinking about it objectively.
FWIW, the subjective/objective difference doesn't AT ALL deny physicalism as far as I'm concerned. It's just that brains are unique in our reality in having both an outside view (the object view) and an inside view (the subjective view). Everything else we study, we only study from the outside.
The infamous Hard Problem is the question of how the hell something can even have subjective experience, and at this point we just don't know enough, but there's no reason to believe the answer goes beyond physicalism. It's just that brains are the most mind-bendingly beyond-complicated systems we've ever encountered.