(This is the first post of a new Sequence, Highly Advanced Epistemology 101 for Beginners, setting up the Sequence Open Problems in Friendly AI.  For experienced readers, this first post may seem somewhat elementary; but it serves as a basis for what follows.  And though it may be conventional in standard philosophy, the world at large does not know it, and it is useful to know a compact explanation.  Kudos to Alex Altair for helping in the production and editing of this post and Sequence!)

I remember this paper I wrote on existentialism. My teacher gave it back with an F. She’d underlined true and truth wherever it appeared in the essay, probably about twenty times, with a question mark beside each. She wanted to know what I meant by truth.
-- Danielle Egan

I understand what it means for a hypothesis to be elegant, or falsifiable, or compatible with the evidence. It sounds to me like calling a belief ‘true’ or ‘real’ or ‘actual’ is merely the difference between saying you believe something, and saying you really really believe something.
-- Dale Carrico

What then is truth? A movable host of metaphors, metonymies, and; anthropomorphisms: in short, a sum of human relations which have been poetically and rhetorically intensified, transferred, and embellished, and which, after long usage, seem to a people to be fixed, canonical, and binding.
-- Friedrich Nietzche

The Sally-Anne False-Belief task is an experiment used to tell whether a child understands the difference between belief and reality. It goes as follows:

The child sees Sally hide a marble inside a covered basket, as Anne looks on.

Sally leaves the room, and Anne takes the marble out of the basket and hides it inside a lidded box.

Anne leaves the room, and Sally returns.

The experimenter asks the child where Sally will look for her marble.

Children under the age of four say that Sally will look for her marble inside the box. Children over the age of four say that Sally will look for her marble inside the basket.

(Attributed to:  Baron-Cohen, S., Leslie, L. and Frith, U. (1985) ‘Does the autistic child have a “theory of mind”?’, Cognition, vol. 21, pp. 37–46.)

Human children over the age of (typically) four, first begin to understand what it means for Sally to lose her marbles - for Sally's beliefs to stop corresponding to reality. A three-year-old has a model only of where the marble is. A four-year old is developing a theory of mind; they separately model where the marble is and where Sally believes the marble is, so they can notice when the two conflict - when Sally has a false belief.

Any meaningful belief has a truth-condition, some way reality can be which can make that belief true, or alternatively false. If Sally's brain holds a mental image of a marble inside the basket, then, in reality itself, the marble can actually be inside the basket - in which case Sally's belief is called 'true', since reality falls inside its truth-condition. Or alternatively, Anne may have taken out the marble and hidden it in the box, in which case Sally's belief is termed 'false', since reality falls outside the belief's truth-condition.

The mathematician Alfred Tarski once described the notion of 'truth' via an infinite family of truth-conditions:

The sentence 'snow is white' is true if and only if snow is white.

The sentence 'the sky is blue' is true if and only if the sky is blue.

When you write it out that way, it looks like the distinction might be trivial - indeed, why bother talking about sentences at all, if the sentence looks so much like reality when both are written out as English?

But when we go back to the Sally-Anne task, the difference looks much clearer: Sally's belief is embodied in a pattern of neurons and neural firings inside Sally's brain, three pounds of wet and extremely complicated tissue inside Sally's skull. The marble itself is a small simple plastic sphere, moving between the basket and the box. When we compare Sally's belief to the marble, we are comparing two quite different things.

(Then why talk about these abstract 'sentences' instead of just neurally embodied beliefs? Maybe Sally and Fred believe "the same thing", i.e., their brains both have internal models of the marble inside the basket - two brain-bound beliefs with the same truth condition - in which case the thing these two beliefs have in common, the shared truth condition, is abstracted into the form of a sentence or proposition that we imagine being true or false apart from any brains that believe it.)

Some pundits have panicked over the point that any judgment of truth - any comparison of belief to reality - takes place inside some particular person's mind; and indeed seems to just compare someone else's belief to your belief:

So is all this talk of truth just comparing other people's beliefs to our own beliefs, and trying to assert privilege? Is the word 'truth' just a weapon in a power struggle?

For that matter, you can't even directly compare other people's beliefs to our own beliefs. You can only internally compare your beliefs about someone else's belief to your own belief - compare your map of their map, to your map of the territory.

Similarly, to say of your own beliefs, that the belief is 'true', just means you're comparing your map of your map, to your map of the territory. People usually are not mistaken about what they themselves believe - though there are certain exceptions to this rule - yet nonetheless, the map of the map is usually accurate, i.e., people are usually right about the question of what they believe:

And so saying 'I believe the sky is blue, and that's true!' typically conveys the same information as 'I believe the sky is blue' or just saying 'The sky is blue' - namely, that your mental model of the world contains a blue sky.


If the above is true, aren't the postmodernists right? Isn't all this talk of 'truth' just an attempt to assert the privilege of your own beliefs over others, when there's nothing that can actually compare a belief to reality itself, outside of anyone's head?

(A 'meditation' is a puzzle that the reader is meant to attempt to solve before continuing. It's my somewhat awkward attempt to reflect the research which shows that you're much more likely to remember a fact or solution if you try to solve the problem yourself before reading the solution; succeed or fail, the important thing is to have tried first . This also reflects a problem Michael Vassar thinks is occurring, which is that since LW posts often sound obvious in retrospect, it's hard for people to visualize the diff between 'before' and 'after'; and this diff is also useful to have for learning purposes. So please try to say your own answer to the meditation - ideally whispering it to yourself, or moving your lips as you pretend to say it, so as to make sure it's fully explicit and available for memory - before continuing; and try to consciously note the difference between your reply and the post's reply, including any extra details present or missing, without trying to minimize or maximize the difference.)



The reply I gave to Dale Carrico - who declaimed to me that he knew what it meant for a belief to be falsifiable, but not what it meant for beliefs to be true - was that my beliefs determine my experimental predictions, but only reality gets to determine my experimental results. If I believe very strongly that I can fly, then this belief may lead me to step off a cliff, expecting to be safe; but only the truth of this belief can possibly save me from plummeting to the ground and ending my experiences with a splat.

Since my expectations sometimes conflict with my subsequent experiences, I need different names for the thingies that determine my experimental predictions and the thingy that determines my experimental results. I call the former thingies 'beliefs', and the latter thingy 'reality'.

You won't get a direct collision between belief and reality - or between someone else's beliefs and reality - by sitting in your living-room with your eyes closed. But the situation is different if you open your eyes!

Consider how your brain ends up knowing that its shoelaces are untied:

  • A photon departs from the Sun, and flies to the Earth and through Earth's atmosphere.
  • Your shoelace absorbs and re-emits the photon.
  • The reflected photon passes through your eye's pupil and toward your retina.
  • The photon strikes a rod cell or cone cell, or to be more precise, it strikes a photoreceptor, a form of vitamin-A known as retinal, which undergoes a change in its molecular shape (rotating around a double bond) powered by absorption of the photon's energy. A bound protein called an opsin undergoes a conformational change in response, and this further propagates to a neural cell body which pumps a proton and increases its polarization.
  • The gradual polarization change is propagated to a bipolar cell and then a ganglion cell. If the ganglion cell's polarization goes over a threshold, it sends out a nerve impulse, a propagating electrochemical phenomenon of polarization-depolarization that travels through the brain at between 1 and 100 meters per second. Now the incoming light from the outside world has been transduced to neural information, commensurate with the substrate of other thoughts.
  • The neural signal is preprocessed by other neurons in the retina, further preprocessed by the lateral geniculate nucleus in the middle of the brain, and then, in the visual cortex located at the back of your head, reconstructed into an actual little tiny picture of the surrounding world - a picture embodied in the firing frequencies of the neurons making up the visual field. (A distorted picture, since the center of the visual field is processed in much greater detail - i.e. spread across more neurons and more cortical area - than the edges.)
  • Information from the visual cortex is then routed to the temporal lobes, which handle object recognition.
  • Your brain recognizes the form of an untied shoelace.

And so your brain updates its map of the world to include the fact that your shoelaces are untied. Even if, previously, it expected them to be tied!  There's no reason for your brain not to update if politics aren't involved. Once photons heading into the eye are turned into neural firings, they're commensurate with other mind-information and can be compared to previous beliefs.

Belief and reality interact all the time. If the environment and the brain never touched in any way, we wouldn't need eyes - or hands - and the brain could afford to be a whole lot simpler. In fact, organisms wouldn't need brains at all.

So, fine, belief and reality are distinct entities which do intersect and interact. But to say that we need separate concepts for 'beliefs' and 'reality' doesn't get us to needing the concept of 'truth', a comparison between them. Maybe we can just separately (a) talk about an agent's belief that the sky is blue and (b) talk about the sky itself.

Maybe we could always apply Tarski's schema - "The sentence 'X' is true iff X" - and replace every invocation of the word 'truth' by talking separately about the belief and the reality.

Instead of saying:

"Jane believes the sky is blue, and that's true",

we would say:

"Jane believes 'the sky is blue'; also, the sky is blue".

Both statements convey the same information about (a) what we believe about the sky and (b) what we believe Jane believes.

And thus, we could eliminate that bothersome word, 'truth', which is so controversial to philosophers, and misused by various annoying people!

Is that valid? Are there any problems with that?

Suppose you had a rational agent, or for concreteness, an Artificial Intelligence, which was carrying out its work in isolation and never needed to argue politics with anyone.

This AI has been designed [note for modern readers: this is back when AIs were occasionally designed rather than gradient-descended, the AI being postulated is not a large language model] around a system philosophy which says to separately talk about beliefs and realities rather than ever talking about truth.

This AI (let us suppose) is reasonably self-aware; it can internally ask itself "What do I believe about the sky?" and get back a correct answer "I believe with 98% probability that it is currently daytime and unclouded, and so the sky is blue."  It is quite sure that this probability is the exact statement stored in its RAM.

Separately, the AI models that "If it is daytime and uncloudy, then the probability that my optical sensors will detect blue out the window is 99.9%.  The AI doesn't confuse this proposition with the quite different proposition that the optical sensors will detect blue whenever it believes the sky is blue. The AI knows that it cannot write a different belief about the sky to storage in order to control what the sensor sees as the sky's color.

The AI can differentiate the map and the territory; the AI knows that the possible states of its RAM storage do not have the same consequences and causal powers as the possible states of sky.

If the AI's computer gets shipped to a different continent, such that the AI then looks out a window and sees the purple-black of night when the AI was predicting the blue of daytime, the AI is surprised but not ontologically confused. The AI correctly reconstructs that what must have happened was that the AI internally stored a high probability of it being daytime; but that outside in actual reality, it was nighttime. The AI accordingly updates its RAM with a new belief, a high probability that it is now nighttime.

The AI is already built so that, in every particular instance, its model of reality (including its model of itself) correctly states points like:

  • If I believe it's daytime and cloudless, I'll predict that my sensor will see blue out the window;
  • It's a certain fact that I currently predict I'll see blue, but that's not the same fact as it being a certainty as to what my sensor will see when I look outside;
  • If I look out the window and see purple-black, then in reality it was nighttime, and I will then write that it's nighttime into my RAM;
  • If I write that it's nighttime into my RAM, this won't change whether it's daytime and it won't change what my sensor sees when it looks out the window.

All of these propositions can be stated without using the word 'truth' in any particular instance.

Will a sophisticated but isolated AI benefit further from having abstract concepts for 'truth' and 'falsehood' in general - an abstract concept of map-territory correspondence, apart from any particular historical cases where the AI thought it was daytime and in reality it was nighttime? If so, how?

Meditation: If we were dealing with an Artificial Intelligence that never had to interact with other intelligent beings, would it benefit from an abstract notion of 'truth'?


Reply: The abstract concept of 'truth' - the general idea of a map-territory correspondence - is required to express ideas such as:

  • Generalized across possible maps and possible cities, if your map of a city is accurate, navigating according to that map is more likely to get you to the airport on time.
  • In general, to draw a true map of a city, someone has to go out and look at the buildings; there's no way you'd end up with an accurate map by sitting in your living-room with your eyes closed trying to imagine what you wish the city would look like.
  • In abstract generality:  True beliefs are more likely than false beliefs to make correct experimental predictions, so if (in general) we increase our credence in hypotheses that make correct experimental predictions, then (in general) our model of reality should become incrementally more true over time.

This is the main benefit of talking and thinking about 'truth' - that we can generalize rules about how to make maps match territories in general; we can learn lessons that transfer beyond particular skies being blue.

You can sit on a chair without having an abstract, general, quoted concept of sitting; cats, for example, do this all the time. One use for having a word "Sit!" that means sitting, is to communicate to another being that you would like them to sit. But another use is to think about "sitting" up at the meta-level, abstractly, in order to design a better chair.

You don't need an abstract meta-level concept of "truths", aka map-territory correspondences, in order to look out at the sky and see purple-black and be surprised and change your mind; animals were doing that before they had words at all. You need it in order to think, at one meta-level up, about crafting improved thought processes that are better at producing truths. That, only one animal species does, and it's the one that has words for abstract things.

Next in main sequence:

Complete philosophical panic has turned out not to be justified (it never is). But there is a key practical problem that results from our internal evaluation of 'truth' being a comparison of a map of a map, to a map of reality: On this schema it is very easy for the brain to end up believing that a completely meaningless statement is 'true'.

Some literature professor lectures that the famous authors Carol, Danny, and Elaine are all 'post-utopians', which you can tell because their writings exhibit signs of 'colonial alienation'. For most college students the typical result will be that their brain's version of an object-attribute list will assign the attribute 'post-utopian' to the authors Carol, Danny, and Elaine. When the subsequent test asks for "an example of a post-utopian author", the student will write down "Elaine". What if the student writes down, "I think Elaine is not a post-utopian"? Then the professor models thusly...

...and marks the answer false.

After all...

The sentence "Elaine is a post-utopian" is true if and only if Elaine is a post-utopian.


Now of course it could be that this term does mean something (even though I made it up).  It might even be that, although the professor can't give a good explicit answer to "What is post-utopianism, anyway?", you can nonetheless take many literary professors and separately show them new pieces of writing by unknown authors and they'll all independently arrive at the same answer, in which case they're clearly detecting some sensory-visible feature of the writing.  We don't always know how our brains work, and we don't always know what we see, and the sky was seen as blue long before the word "blue" was invented; for a part of your brain's world-model to be meaningful doesn't require that you can explain it in words.

On the other hand, it could also be the case that the professor learned about "colonial alienation" by memorizing what to say to his professor.  It could be that the only person whose brain assigned a real meaning to the word is dead.  So that by the time the students are learning that "post-utopian" is the password when hit with the query "colonial alienation?", both phrases are just verbal responses to be rehearsed, nothing but an answer on a test.

The two phrases don't feel "disconnected" individually because they're connected to each other - post-utopianism has the apparent consequence of colonial alienation, and if you ask what colonial alienation implies, it means the author is probably a post-utopian.  But if you draw a circle around both phrases, they don't connect to anything else.  They're floating beliefs not connected with the rest of the model. And yet there's no internal alarm that goes off when this happens. Just as "being wrong feels like being right" - just as having a false belief feels the same internally as having a true belief, at least until you run an experiment - having a meaningless belief can feel just like having a meaningful belief.

(You can even have fights over completely meaningless beliefs.  If someone says "Is Elaine a post-utopian?" and one group shouts "Yes!" and the other group shouts "No!", they can fight over having shouted different things; it's not necessary for the words to mean anything for the battle to get started.  Heck, you could have a battle over one group shouting "Mun!" and the other shouting "Fleem!"  More generally, it's important to distinguish the visible consequences of the professor-brain's quoted belief (students had better write down a certain thing on his test, or they'll be marked wrong) from the proposition that there's an unquoted state of reality (Elaine actually being a post-utopian in the territory) which has visible consquences.)

One classic response to this problem was verificationism, which held that the sentence "Elaine is a post-utopian" is meaningless if it doesn't tell us which sensory experiences we should expect to see if the sentence is true, and how those experiences differ from the case if the sentence is false.

But then suppose that I transmit a photon aimed at the void between galaxies - heading far off into space, away into the night. In an expanding universe, this photon will eventually cross the cosmological horizon where, even if the photon hit a mirror reflecting it squarely back toward Earth, the photon would never get here because the universe would expand too fast in the meanwhile. Thus, after the photon goes past a certain point, there are no experimental consequences whatsoever, ever, to the statement "The photon continues to exist, rather than blinking out of existence."

And yet it seems to me - and I hope to you as well - that the statement "The photon suddenly blinks out of existence as soon as we can't see it, violating Conservation of Energy and behaving unlike all photons we can actually see" is false, while the statement "The photon continues to exist, heading off to nowhere" is true. And this sort of question can have important policy consequences: suppose we were thinking of sending off a near-light-speed colonization vessel as far away as possible, so that it would be over the cosmological horizon before it slowed down to colonize some distant supercluster. If we thought the colonization ship would just blink out of existence before it arrived, we wouldn't bother sending it.

It is both useful and wise to ask after the sensory consequences of our beliefs. But it's not quite the fundamental definition of meaningful statements. It's an excellent hint that something might be a disconnected 'floating belief', but it's not a hard-and-fast rule.

You might next try the answer that for a statement to be meaningful, there must be some way reality can be which makes the statement true or false; and that since the universe is made of atoms, there must be some way to arrange the atoms in the universe that would make a statement true or false. E.g. to make the statement "I am in Paris" true, we would have to move the atoms comprising myself to Paris. A literateur claims that Elaine has an attribute called post-utopianism, but there's no way to translate this claim into a way to arrange the atoms in the universe so as to make the claim true, or alternatively false; so it has no truth-condition, and must be meaningless.

Indeed there are claims where, if you pause and ask, "How could a universe be arranged so as to make this claim true, or alternatively false?", you'll suddenly realize that you didn't have as strong a grasp on the claim's truth-condition as you believed. "Suffering builds character", say, or "All depressions result from bad monetary policy." These claims aren't necessarily meaningless, but they're a lot easier to say, than to visualize the universe that makes them true or false. Just like asking after sensory consequences is an important hint to meaning or meaninglessness, so is asking how to configure the universe.

But if you say there has to be some arrangement of atoms that makes a meaningful claim true or false...

Then the theory of quantum mechanics would be meaningless a priori, because there's no way to arrange atoms to make the theory of quantum mechanics true.

And when we discovered that the universe was not made of atoms, but rather quantum fields, all meaningful statements everywhere would have been revealed as false - since there'd be no atoms arranged to fulfill their truth-conditions.

Meditation: What rule could restrict our beliefs to just propositions that can be meaningful, without excluding a priori anything that could in principle be true?

  • Meditation Answers - (A central comment for readers who want to try answering the above meditation (before reading whatever post in the Sequence answers it) or read contributed answers.)
  • Mainstream Status - (A central comment where I say what I think the status of the post is relative to mainstream modern epistemology or other fields, and people can post summaries or excerpts of any papers they think are relevant.)


Part of the sequence Highly Advanced Epistemology 101 for Beginners

Next post: "Skill: The Map is Not the Territory"

New Comment
543 comments, sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

I just realized that since I posted two comments that were critical over a minor detail, I should balance it out by mentioning that I liked the post - it was indeed pretty elementary, but it was also clear, and I agree about it being considerably better than The Simple Truth. And I liked the koans - they should be a useful device to the readers who actually bother to answer them.


Human children over the age of (typically) four, first begin to understand what it means for Sally to lose her marbles - for Sally's beliefs to stop corresponding to reality.

was a cute touch.


Thank you for being positive.

I've been recently thinking about this, and noticed that despite things like "why our kind can't cooperate", we still focus on criticisms of minor points, even when there are major wins to be celebrated.

(The 'Mainstream Status' comment is intended to provide a quick overview of what the status of the post's ideas are within contemporary academia, at least so far as the poster knows. Anyone claiming a particular paper precedents the post should try to describe the exact relevant idea as presented in the paper, ideally with a quote or excerpt, especially if the paper is locked behind a paywall. Do not represent large complicated ideas as standard if only a part is accepted; do not represent a complicated idea as precedented if only a part is described. With those caveats, all relevant papers and citations are much solicited! Hopefully comment-collections like these can serve as a standard link between LW presentations and academic ones.)

The correspondence theory of truth is the first position listed in the Stanford Encyclopedia of Philosophy, which is my usual criterion for saying that something is a solved problem in philosophy. Clear-cut simple visual illustration inspired by the Sally-Anne experimental paradigm is not something I have previously seen associated with it, so the explanation in this post is - I hope - an improvement over what's standard.

Alfred Tarski is a famous mat... (read more)

This is a great post. I think the presentation of the ideas is clearer and more engaging than the sequences, and the cartoons are really nice. Wild applause for the artist.

I have a few things to say about the status of these ideas in mainstream philosophy, since I'm somewhat familiar with the mainstream literature (although admittedly it's not the area of my expertise). I'll split up my individual points into separate comments.

Alfred Tarski is a famous mathematician whose theory of truth is widely known.

Summary of my point: Tarski's biconditionals are not supposed to be a definition of truth. They are supposed to be a test of the adequacy of a proposed definition of truth. Proponents of many different theories claim that their theory passes this test of adequacy, so to identify Tarski's criterion with the correspondence theory is incorrect, or at the very least, a highly controversial claim that requires defense. What follows is a detailed account of why the biconditionals can't be an adequate definition of truth, and of what Tarski's actual theory of truth is.

Describing Tarski's biconditionals as a definition of truth or a theory of truth is misleading. The relevant paper is ... (read more)

I've slightly edited the OP to say that Tarski "described" rather than "defined" truth - I wish I could include more to reflect this valid point (indeed Tarski's theorems on truth are a lot more complicated and so are surrounding issues, no language can contain its own truth-predicate, etc.), but I think it might be a distraction from the main text. Thank you for this comment though!

The latest Rationally Speaking post looks relevant: Ian Pollock describes aspects of Eliezer's view as "minimalism" with a link to that same SEP article. He also mentions Simon Blackburn's book, where Blackburn describes minimalists or quietists as making the same point Eliezer makes about collapsing "X is true" to "X" and a similar point about the usefulness of the term "truth" as a generalisation (though it seems that minimalists would say that this is only a linguistic convenience, whereas Eliezer seems to have a slightly difference concept of it in that he wants to talk in general about how we get accurate beliefs).
Thanks for this whole comment. In particular, My gut instinct is deflationist, but I don't see this view as being opposed to "correspondence". The alleged conflict is dubious at best. Stanford Encyclopedia of Philosophy writes: Emphasis added: the italicized premise is false. Explanation is a cognitive feat, and the same fact (even if the identity is a necessary one) can be cognized in different ways. (Such explanations occur frequently enough in mathematics, I think.) The SEP author anticipates my objection and writes: It is open to them to argue that "because" does not create a hyper-intensional context, but it is much more plausible that it does. So until a good argument comes along, mark me down as a correspondence deflationist.
It's vogue to defend correspondence because 1) it sounds like common sense and 2) it signals rejection of largely discredited instrumentalism. But surely a correspondence theorist should have a theory of the nature of the correspondence. How does a proposition or a verbal string correspond to a state of reality? By virtue of what is it a correct description? We can state a metalinguistic relationship about "Snow is white," but how does this locution hook onto the actual world? Correspondence theorists think this is a task for a philosophical theory of reference. (Such as in an account where "torekp" refers to you by virtue of the "christening event" of your creating the account and causal connections therefrom.) Deflationists are apt to say it is ultimately a technical problem in the psychology of language.
Interesting. I am inclined to replicate my compatibility claim at this level too; i.e., the technical solution in the psychology of language will be a philosophical theory of reference (as much as one needs) as well. I'd be interested in references to any of the deflationist discussions of reference you have in mind.

I haven't particularly run across any philosophy explicitly making the connection from the correspondence theory of truth to "There are causal processes producing map-territory correspondences" to "You have to look at things in order to draw accurate maps of them, and this is a general rule with no exception for special interest groups who want more forgiving treatment for their assertions". I would not be surprised to find out it existed, especially on the second clause.

Depends on what you mean by "explicitly". Many correspondence theorists believe that an adequate understanding of "correspondence" requires an understanding of reference -- how parts of our language are associated with parts of the world. I think this sort of idea stems from trying to fill out Tarski's (actual) definition of truth, which I discussed in another comment. The hope is that a good theory of reference will fill out Tarski's obscure notion of satisfaction, and thereby give some substance to his definition of truth in terms of satisfaction.

Anyway, there was a period when a lot of philosophers believed, following Saul Kripke and Hilary Putnam, that we can understan... (read more)

Speaking as the author of Eliezer's Sequences and Mainstream Academia...

Off the top of my head, I also can't think of a philosopher who has made an explicit connection from the correspondence theory of truth to "there are causal processes producing map-territory correspondences" to "you have to look at things to draw accurate maps of them..."

But if this connection has been made explicitly, I would expect it to be made by someone who accepts both the correspondence theory and "naturalized epistemology", often summed up in a quote from Quine:

The stimulation of his sensory receptors is all the evidence anybody has had to go on, ultimately, in arriving at his picture of the world. Why not just see how this construction really proceeds? Why not settle for psychology? ...Epistemology, or something like it, simply falls into place as a chapter of psychology and hence of natural science.

(Originally, Quine's naturalized epistemology accounted only for this descriptive part of epistemology, and neglected the normative part, e.g. truth conditions. In the 80s Quine started saying that the normative part entered into naturalized epistemology through "the t... (read more)

It's not that clear to me in what sense mainstream academia is a unified thing which holds positions, even regarding questions such as "what fields are legitimate". Saying that something is known in mainstream academia seems suspiciously like saying that "something is encoded in the matter in my shoelace, given the right decryption schema. OTOH, it's highly meaningful to say that something is discoverable by someone with competent 'google-fu"

Strongly seconded. Hell, some "Mainstream" scientists are working on big-money research project that attempt to prove that there's a worldwide conspiracy attempting to convince people that global warming exists so as to make money off of it. Either they're all sell-outs, something which seems very unlikely, or at least some of them actually disagree with some other mainstream scientists, who see the "Is there real global warming?" question as obviously resolved long ago.
Agree with all this.

Off the top of my head, I also can't think of a philosopher who has made an explicit connection from the correspondence theory of truth to "there are causal processes producing map-territory correspondences" to "you have to look at things to draw accurate maps of them..."

OK, I defended the tweet that got this response from Eliezer as the sort of rhetorical flourish that gets people to actually click on the link. However, it looks like I also underestimated how original the sequences are - I had really expected this sort of thing to mirror work in mainstream philosophy.

Although I wouldn't think of this particular thing as being an invention on his part - I'm not sure I've read that particular chain of thought before, but all the elements of the chain are things I've known for years. However I think it illustrates the strength of Eliezer's writing well. It's a perfectly legitimate sequence of thought steps that leads in a clear and obvious way to the right answer. It's not new thought, but a particularly clear way of expressing something that many people have thought and written about in a less organised way. To clarify - there are times when Eliezer is inventive - for example his work on CEV - but this isn't one of those places. I know I'm partly arguing about the meaning of "inventive", but I don't think we're doing him a favor here by claiming this is an example of his inventiveness when there are much better candidates.
Karl Popper did so explicitly, thoroughly and convincingly in The Logic of Scientific Discovery. Pretty influential, and definitely a part of "Mainstream Academia." Here's an interesting, if lengthy, footnote to Chapter 84 - Remarks Concerning the use of the concepts 'True' and 'Corroborated'. A (short) footnote of my own: Popper's writings have assumed the status of mere "background knowledge", which is a truly great achievement for any philosopher of science. However, The Logic of Scientific discovery is a glorious book which deserves to be even more widely read. Part I of the book spans no more than 30 pages. It's nothing short of beautiful. PDF here.
Could you please quote the part of Popper's book that makes the explicit connection from the correspondence theory of truth to "there are causal processes producing map-territory correspondences" to "you have to look at things to draw accurate maps of them..."?
Right, this is the obvious next question. I started looking for the appropriate "sound bites" yesterday, but encountered a bit of difficulty in doing so, as I shall explain. Popper's embrace of (Tarskian) correspondence theory should be at least somewhat clear from the footnote I quoted above. It seems clear to me, from my recount of the book that "you have to look at things to draw accurate maps of them" is one of the chief aims, and one of the central claims of the book; a claim which is defended, by a lengthy, but quite convincing and unusually successful argument - the premises to which are presented only one at a time, and quite meticulously over at least several chapters, so I'm not exactly sure how to go about quoting only the "relevant parts". My claim that his argument was convincing and successful, is based on the historical observation that popperian falsificationism (the hypothetico-deductive framework) won out over the then quite prevalent logical positivist / verificationist view, to such an extent that it quickly became the default mode of Science, a position it has held, mostly uncontested, ever since, and therefore is barely worthy of mention today. Except when it is, that is; when one encounters problems that are metaphysical (according to Popper), such as Susskind's String Landscape of perhaps 10^500 vacuua, the small (but significant) observed value of the cosmological constant, the (seemingly fine tuned) value of the fine structure constant, and other observations that may require anthropic i.e. metaphysical explanations, since these problems are seemingly not decidable inside of standard, i.e. popperian science. I feel faced with a claim similar to "I don't believe any mathematician has convincingly proven Fermat's last theorem." To which I reply: Andrew Wiles (1995) The obvious next question is: "Can you please quote the part where he proves the theorem?" This is unfortunately somewhat involved, as the entire 109 page paper tries and succed
A (very) quick attempt, perhaps this will suffice? (Let me know if not. ) I begin with the tersest possible defense of my claim that Popper argued that "you actually have to look at things to draw accurate maps of them...", even though this particular example is particularily trivial: Page 19: To paraphrase: You have to look actually out the window to discover whether it is raining or not. ---------------------------------------- Continuing, page 16: (Oops, comment too long.)
(Continued) Page 20: [a number of indicative, but not decisive quotes omitted] ---------------------------------------- I had hoped to find some decisive sound bite in part one, which is a brief discussion of the epistemological problems facing any theory of scientific method, and an outline of Popper's framework, but it looks like I shall have to go deeper. Will look into this over the weekend. I also found another, though much more recent candidate, David Deutsch in The Beginning of Infinity, Chapter 1 on "The Reach of Explanations". Tough I'm beginning to suspect that although they both point out that "you have to look at things to draw accurate maps of them...", and describe "causal processes producing map-territory correspondences" (for example, between some state of affairs and the output of some scientific instument) both Deutsch and Popper seem to have omitted what one may call the "neuroscience of epistemology." (Where the photon reflects off your shoelace, gets absorbed by your retina, leading to information about the configuration of the world becoming entangled with some corresponding state of your brain, and so on.) This is admittedly quite a crucial step, which Yudkowsky's explanation does cover, and which I cannot recall to have seen elsewhere.
Here's a quote from Perry Anderson's recent (highly critical) essay on Gandhi:
Trying to include mainstream academia other than philosophy, and going off your blog post "The Second Law of Thermodynamics, and Engines of Cognition", it seems the idea of the general rule that you have to look at and interact with things to form accurate beliefs about them was largely due to Leo Szilard in his 1939 paper "On the Decrease in Entropy in a Thermodynamic System by the Intervention of Intelligent Beings" which analyzed Maxwell's demon thought experiment and introduced the Szilard engine and the entropy cost of gaining information. You gave a more Bayesian analysis than Szilard in that post, possibly going off Jaynes' work in statistical mechanics, like his 1957 papers "Information Theory and Statistical Mechanics" parts one and two, which are the earliest mention of Liouville's theorem I can find in that context. Does Pearl have anything to throw in the mix, like a fancy rule about concluding a past causal interaction when you see corresponding maps and cities?
2Eliezer Yudkowsky
DevilWorm and pragmatist point to the "reliabilism" school of philosophy (http://en.wikipedia.org/wiki/Reliabilism & http://plato.stanford.edu/entries/reliabilism). Clicking on either link reveals arguments concerned mainly with that old dispute over whether the word "knowledge" should be used to refer to "justified true belief". Going on the wording I'm not even sure whether they're considering how photons from the Sun are involved in correlating your visual cortex to your shoelaces. But it does increase the probability of a precedent - does anyone have something more specific? (A lot of the terminology I've seen so far is tremendously vague, and open to many interpretations...) Incidentally, there might be an even higher probability of finding some explicit precedent in a good modern AI book somewhere?
It might be too obvious to be worth mentioning. If you're actually building (narrow) AI devices like self-driving cars, then of course your car has to have a way of sensing things round about it if it's going to build a map of its surroundings. This fact should be turned into an SMBC cartoon.
That's what I was thinking. Maybe in something like Knowledge Representation and Reasoning.
AI books tend to assume that one pretty explicitly. For those of a more philosophical bent, some might say something like "The world pushes back", but it's not like anyone doing engineering is in the business of questioning whether the external world exists.
Epistemology and the Psychology of Human Judgment (badger's summary) seems relevant, as one of the things they do is attack reliabilism's uselessness. I don't recall any direct precedents, but it's been a while since I read it.
Bishop & Trout call their approach "strategic reliabilism." A short summary is here. It's far more Yudkowskian than normal reliabilism. LWers may also enjoy their paper The Pathologies of Standard Analytic Epistemology.
That was a pretty cool paper. I don't think I've ever seen SPRs in a philosophy paper before.
For the curious, I interviewed Michael Bishop a couple years ago.
Process reliabilism maybe? Defines the "justified" part in "justified true belief" as the belief being formed by a reliable truth-producing process. From the Stanford Encyclopedia of Philosophy article: Whatever a "causal theory of knowing" is. But it sounds like the kind of thing you're talking about.

I don't like the "post-utopian" example. I can totally expect differing sensory experiences depending on whether a writer is post-utopian or not. For example, if they're post-utopian, when reading their biography I would more strongly expect reading about them having been into utopian ideas when they were young, but having then changed their mind. And when reading their works, I would more strongly expect seeing themes of the imperfectability of the world and weltschmerz.

8Eliezer Yudkowsky
I've edited the OP to try and compartmentalize off the example a bit more.
Do you also think the label "Impressionist painter" is meaningless?
7Eliezer Yudkowsky
I have no idea what Impressionism is (I am not necessarily proud of this ignorance, since for all I know it does mean something important). Do you think that a panel of artists would be able to tell who was and wasn't "Impressionist" and mostly agree with each other? That does seem like a good criterion for whether there's sensory data that they're reacting to.

Apparently even computers agree with those judgments (or at least cluster "impressionists" in their own group - I didn't read the paper, but I expect that the cluster labels were added manually).

ETA: Got the paper. Excerpts:

The dataset includes 994 paintings representing 34 painters, such that each painter has at least 19 images in the dataset. The painters represent several different schools of art such as Early, High, and Northern Renaissance, Mannerism, Baroque, Rococo, Romanticism, Impressionism, Post and Neo Impressionism, Abstract Expressionism, Surrealism, and Fauvism, as commonly defined by art historians. The images were downloaded from various online sources, and normalized to a size of 640,000 pixels while preserving the original aspect ratio. The paintings that were selected for the experiment are assumed to be all in their original condition.

[...] To make the analysis more meaningful for comparing similarities between artistic styles of painters, we selected for each painter paintings that reflect the signature artistic style of that painter. For instance, in Wassily Kandinsky collection we included only paintings representing his abstract expressionism sig

... (read more)

I'm no art geek, but Impressionism is an art "movement" from the late 1800s. A variety of artists (Monet, Renoir, etc) began using similar visual styles that influenced what they decided to paint and how they depicted images.

Art critics think that artistic "movements" are a meaningful way of analyzing paintings, approximately at the level of usefulness that a biologist might apply to "species" or "genus." Or historian of philosophy might talk about the school of thought know today as "Logical Positivism."

Do you think movements is a reasonable unit of analysis (in art, in literature, in philosophy)? If no, why not? If yes, why are you so hostile to the usage of labels like "post-utopian" or "post-colonialist"?

The pictures made within an artistic movement have something similar. We should classify them by that something, not only by the movement. Although the name of the movement can be used as a convenient label for the given cluster of picture-space. If I give you a picture made by unknown author, you can't classify it by author's participation in given movements. But you can classify it by the contents of the picture itself. So even if we use the movement as a label for the cluster, it is better if we can also describe typical properties of picture within that cluster. Just like when you find a random dog on a street, you can classify it as "dog" species, without taking a time machine and finding out whether the ancestors of this specific dogs really were domesticated wolves. You can teach "dogs are domesticated wolves" at school, but this is not how you recognize dogs in real life. So how exactly would you recognize "impressionist" paintings, or "post-utopian" books in real life, when the author is unknown? Without teaching this, you are not truly teaching impressionism or post-utopianism. (In case of "impressionism", my rule of thumb is that the picture looks nice and realistic from distance, but when you stand close to it, the details become somehow ugly. My interpretation of "impressionism" is: work of authors who obviously realized that milimeter precision for a wall painting is an overkill, and you can make pictures faster and cheaper if you just optimize it for looking correct from a typical viewing distance.)
I agree with you that there are immediately obvious properties that I use to classify an object into a category, without reference to various other historical and systemic facts about the object. For example, as you say, I might classify a work of art as impressionist based on the precision with which it is rendered, or classify an animal as a dog based on various aspects of its appearance and behavior, or classify food as nutritious based on color, smell, and so forth. It doesn't follow that it's somehow better to do so than to classify the object based on the less obvious historical or systemic facts. If I categorize an object as nutritious based on those superficial properties, and later perform a lab analysis and discover that the object will kill me if I eat it, I will likely consider my initial categorization a mistake. If I share your rule of thumb about "impressionism", and then later realize that some works of art that share the property of being best viewed from a distance are consistently classed by art students as "pointilist" rather than "impressionist", and I further realize that when I look at a bunch of classed-as-pointilist and classed-as-impressionist paintings it's clear to me that paintings in each class share a family resemblance that they don't share with paintings in the other class, I will likely consider my initial rule of thumb a mistake. Sometimes, the categorization I perform based on properties that aren't immediately apparent is more reliable than the one I perform "in real life."
1Eliezer Yudkowsky
Is this actually a standard term? I was trying to make up a new one, without having to actually delve into the pits of darkness and find a real postmodern literary term that doesn't mean anything.

Maybe you should reconsider picking on an entire field you know nothing about?

I'm not saying this to defend postmodernism, which I know almost nothing about, but to point out that the Sokal hoax is not really enough reason to reject an entire field (any more than the Bogdanov affair is for physics).

I'm pointing out that you're neglecting the virtues of curiosity and humility, at least.

And this is leaving aside that there is no particular reason for "post-utopian" to be a postmodern as opposed to modern term; categorizing writers into movements has been a standard tool of literary analysis for ages (unsurprisingly, since people love putting things into categories).


At this point, getting in cheap jabs at post-modernism and philosophy wherever possible is a well-honored LessWrong tradition. Can't let the Greens win!

I don't think you can avoid the criticism of "literary terms actually do tend to make one expect differing sensory experiences, and your characterization of the field is unfair" simply by inventing a term which isn't actually in use. I don't know whether "post-utopian" is actually a standard term, but yli's comment doesn't depend on it being one.

Well, there are a lot of hits for "post-utopian" on Google, and they don't seem to be references to you.
4Eliezer Yudkowsky
I think there were fewer Google references back when I first made up the word... I will happily accept nominations for either an equally portentous-sounding but unused term, or a portentous-sounding real literary term that is known not to mean anything.

Has anyone ever told you your writing style is Alucentian to the core? Especially in the way your municardist influences constrain the transactional nuances of your structural ephamthism.

[This comment is no longer endorsed by its author]Reply
9Eliezer Yudkowsky
This looks promising. Is it real, or did you verify that the words don't mean anything standard?

Alucentian, municardist, and structural ephamthism don't mean anything, though Municard is trademarked. Between Louise Rosenblatt's Transactional Theory in literary criticism and Transactional analysis in psychotherapy, there's probably someone who could define "transactional nuances" for you, though it's certainly not a standard phrase.

[This comment is no longer endorsed by its author]Reply

Coming up with a made up word will not solve this problem. If the word describes the content of the author's stories then there will be sensory experiences that a reader can expect when reading those stories.

I think the idea is that the hypothetical teacher is making students memorize passwords instead of teaching the meaning of the concept.
post-catalytic psycho-elemental anti-ludic anarcho-hegemonic desublimational
"Cogno-intellectual" was the catchphrase for this when I was in school. See Abrahams et al.:
To see the word used spectacularly, check out this paper: www.es.ele.tue.nl/~tbasten/fun/rhetoric_logic.pdf
LW comments use the Markdown syntax.
Was that meant to be a link?
It was. I can't get the 'show help' menu to pop-up, so I feel frustratingly inept right now. :)
Put the text you want to display in square brackets, and the URL you want to go to in regular brackets. That should do it.
Anti-ludic has meaning, though. It means "against playfulness". Nobody may have used it yet, but that doesn't mean that you can't combine roots to make a new and meaningful word.
I don't think literature has any equivalent to metasyntactic variables. Still, placeholder names might help - perhaps they are examples of "post-kadigan" literature?
I think those might all be real terms.
I think most literature teachers I've had would ignore the question entirely and use all those terms anyway with whatever meaning they thought fits best.
I have no idea, I just interpreted it in an obvious way.
I share this interpretation, but I always figured in Eliezer's examples the hypothetical professor was so obsessed with passwords or sounding knowledgeable that they didn't bother to teach the meaning of 'post-utopian', and might even have forgotten it. Or they were teaching to the test, but if this is a college class there is no standard test, so they're following some kind of doubly-lost purpose. Or it could be that the professor is passing down passwords they were taught as a student themselves. A word must have had some meaning when it was created, but if most people treat it as a password it won't constrain their expectations. Also, I like that the comment system correctly interpreted my use of underbars to mean italics. I've been using that convention in plaintext for 15 years or so, glad to see someone agrees with it!

There are some kinds of truths that don't seem to be covered by truth-as-correspondence-between-map-and-territory. (Note: This general objection is well know and is given as Objection 1 in SEP's entry on Correspondence Theory.) Consider:

  1. modal truths if one isn't a modal realist
  2. mathematical truths if one isn't a mathematical Platonist
  3. normative truths

Maybe the first two just argues for Platonism and modal realism (although I note that Eliezer explicitly disclaimed being a modal realist). The last one is most problematic to me, because some kinds of normative statements seem to be talking about what one should do given some assumed-to-be-accurate map, and not about the map itself. For example, "You should two-box in Newcomb's problem." If I say "Alice has a false belief that she should two-box in Newcomb's problem" it doesn't seem like I'm saying that her map doesn't correspond to the territory.

So, a couple of questions that seem open to me: Do we need other notions of truth, besides correspondence between map and territory? If so, is there a more general notion of truth that covers all of these as special cases?

I think a more general notion of truth could be defined as correspondence between a map and any structure. If you define a structure using axioms and are referencing that structure, then you can talk about the correspondence properties of that reference. This at least cover both mathematical structures and physical reality.

If I say "Alice has a false belief that she should two-box in Newcomb's problem" it doesn't seem like I'm saying that her map doesn't correspond to the territory.

The problem with Alice's belief is that it is incomplete. It's like saying "I believe that 3 is greater than" (end of sentence).

Even incomplete sentences can work in some contexts where people know how to interpret them. For example if we had a convention that all sentences ending with "greater than" have to be interpreted as "greater than zero", then in given context the sentence "3 is greater than" makes sense, and is true. It just does not make sense outside of this context. Without context, it's not a logical proposition, but rather a proposition template.

Similarly, the sentence "you should X" is meaningful in contexts which provide additional explanation of what "should" means. For a consequentialist, the meaning of "you should" is "maximizes your utility". For a theist, it could mean "makes Deity happy". For both of them, the meaning of "should" is obvious, and within their contexts, they are right. The se... (read more)

At the risk of nitpicking: "Makes Deity happy" sounds to me like a very specific interpretation of "utility", rather than something separate from it. I can't picture any context for the phrase "P should X" that doesn't simply render "X maximizes utility" for different values of the word "utility". If "make Deity happy" is the end goal, wouldn't "utility" be whatever gives you the most efficient route to that goal?
Utility has a single, absolute, unexpressible meaning. To say "X gives me Y utility" is pointless, because I am making a statement about qualia, which are inherently incommunicable - I cannot describe the quale "red" to a person without a visual cortex, because that person is incapable of experiencing red (or any other colour-quale). "X maximises my utility" is implied by the statements "X maximises my deity's utility" and "maximising my deity's utility maximises my utility", but this is not the same thing as saying that X should occur (which requires also that maximisng your own utility is your objective). Stripped of the word "utility", your statement reduces to "The statement 'If X is the end goal, and option A is the best way to achieve X, A should be chosen' is tautologous", which is true because this is the definition of the word "should".
Michael Lynch has a functionalist theory of truth (described in this book) that responds to concerns like yours. His claim is that there is a "truth role" that is constant across all domains of discourse where we talk about truth and falsity of propositions. The truth role is characterized by three properties: 1. Objectivity: The belief that p is true if and only if with respect to the belief that p, things are as they are believed to be. 2. Norm of belief: It is prima facie correct to believe that p if and only if the proposition that p is true. 3. End of inquiry: Other things being equal, true beliefs are a worthy goal of inquiry. Lynch claims that, in different domains of discourse, there are different properties that play this truth role. For instance, when we're doing science it's plausible that the appropriate realizer of the truth role is some kind of correspondence notion. On the other hand, when we're doing mathematics, one might think that the truth role is played by some sort of theoretical coherence property. Mathematical truths, according to Lynch, satisfy the truth role, but not by virtue of correspondence to some state of affairs in our external environment. He has a similar analysis of moral truths. I'm not sure whether Lynch's particular description of the truth role is right, but the functionalist approach (truth is a functional property, and the function can be performed by many different realizers) is very attractive to me.
Me too, thanks for this.
I think Yudkowsky is a Platonist, and I'm not sure he has a consistent position on modal realism, since when arguing on morality he seemed to espouse it: see his comment here. I don't think that "You should two-box in Newcomb's problem." is actually a normative statement, even if it contains a "should": you can rephrase it epistemically as "If you two-box in Newcomb's problem then you will maximize your expected utility". Therefore, if you say "Alice has a false belief that if she two-boxes in Newcomb's problem then she will maximize her expected utility" you are saying that her belief doesn't correspond to the mathematical constructs underlying Newcomb's problem. If you take the Platonist position that mathematical constructs exist as external entities ("the territory"), then yes, you are saying that her map doesn't correspond to the territory.
Well, sure, a utilitarian can always "rephrase" should-statements that way; to a utilitarian what "X should Y" means is "Y maximizes X's expected utility." That doesn't make "X should Y" not a normative statement, it just means that utilitarian normative statements are also objective statements about reality. Conversely, I'm not sure a deontologist would agree that you can rephrase one as the other... that is, a deontologist might coherently (and incorrectly) say "Yes, two-boxing maximizes expected utility, but you still shouldn't do it."
I think you are conflating two different types of "should" statements: moral injunctions and decision-theoretical injunctions. The statement "You should two-box in Newcomb's problem" is normally interpreted as a decision-theoretical injunction. As such, it can be rephrased epistemically as "If you two-box in Newcomb's problem then you will maximize your expected utility". But you could also interpret the statement "You should two-box in Newcomb's problem" as the moral injunction "It is morally right for you to two-box in Newcomb's problem". Moral injunctions can't be rephrased epistemically, at least unless you assume a priori that there exist some external moral truths that can't be further rephrased. The utilitarianist of your comment is doing that. His actual rephrasing is "If you two-box in Newcomb's problem then you will maximize the expected universe cumulative utility". This assumes that: * This universe cumulative utility exists as an external entity * The statement "It is morally right for you to maximize the expected universe cumulative utility" exists as an external moral truth.
0Wei Dai
Thanks for the link. That does seem inconsistent. This comment should help you understand why I disagree. Does it make sense?
I don't claim that all injunctions can be rephrased as epistemic statements. I claim that decision-theoretic injunctions can be rephrased as epistemic statements. Moral injunctions can't.
-2Wei Dai
I'm confused by your reply because the comment I linked to tried to explain why I don't think "You should two-box in Newcomb's problem" can be rephrased as an epistemic statement (as you claimed earlier). Did you read it, and if so, can you explain why you disagree with its reasoning? ETA: Sorry, I didn't notice your comment in the other subthread where you gave your definitions of "decision-theoretic" vs "moral" injunctions. Your reply makes more sense with those definitions in mind, but I think it shows that the comment I linked to didn't get my point across. So I'll try it again here. You said earlier: A causal decision theorist (C) and an evidential decision theorist (E) have different definitions of "maximize your expected utility", and so when C says to E "you should two-box in Newcomb's problem" he is not just saying "If you two-box in Newcomb's problem then you will maximize your expected utility according to the CDT formula" since E wouldn't care about that. So my point is that "you should two-box in Newcomb's problem" is usually not a "decision-theoretical injunction" in your sense of the phrase, but rather a normative statement as I claimed.
I was assuming implicitely that we were talking in the context of EDT. In general, you can say "Two-boxing in Newcomb's problem is the optimal action for you", where the definition of "optimal action" depends on the decision theory you use. If you use EDT, then "optimal action" means "maximizes expected utility", hence the statement above is false (that is, it is inconsistent with the axioms of EDT and Newcomb's problem). If you use CDT, then "optimal action" means "maximizes expected utility under a causality assumption". Hence the statement above is technically true, although not very useful, since the axioms that define Newcomb's problem specifically violate the causality assumption. So, which decision theory should you use? An answer like "you should use the decision theory that determines the optimal action without any assumption that violates the problem constraints" seems irreducible to an epistemic statement. But is that actually correct? If you are studing actual agents, then the point is moot, since these agents already have a decision theory (in practice it will be an approximation of either EDT or CDT, or something else), but what if you want to improve yourself, or build an artificial agent? Then you evaluate the new decision theory according to the decision theory that you already have. Then, assuming that in principle your current decision theory can be described epistemically, you can say, for instance: "A decision theory that determines the optimal action without any assumption that violates the problem constraints is optimal for me". If you want to suggest a decision theory to somebody who is not you, you can say: "A decision theory that determines the optimal action without any assumption that violates the problem constraints is optimal for you", or, more properly but less politely: "You using a decision theory that determines the optimal action without any assumption that violates the problem constraints are optimal for me".
1Wei Dai
I had similar thoughts before, but eventually changed my mind. Unfortunately it's hard to convince people that their solution to some problem isn't entirely satisfactory without having a better solution at hand. (For example, this post of mine pointing out a problem with using probability theory to deal with indexical uncertainty sat at 0 points for months before I made my UDT post which suggested a different solution.) So instead of trying harder to convince people now, I think I will instead try harder to figure out a better answer by myself (and others who already share my views).
It seems that way to me. Specifically, in that case I think you're saying that Alice (wrongly) expects that her decision is causally independent from the money Omega put in the boxes, and as such thinks that her expected utility is higher from grabbing both boxes.
I don't think 2 is answered even if you say that the mathematical objects are themselves real. Consider a geometry that labels "true" everything that follows from its axioms. If this geometry is consistent, then we want to say that it is true, which implies that everything it labels as "true", is. And the axioms themselves follow from the axioms, so the mathematical system says that they're true. But you can also have another valid mathematical system, where one of those axioms is negated. This is a problem because it implies that something can be both true and not true. Because of this, the sense in which mathematical propositions can be true can't be the same sense in which "snow is white" can be true, even if the objects themselves are real. We have to be equivocating somewhere on "truth".
It's easy to overcome that simply by being a bit more precise - you are saying that such and such a proposition is true in geometry X. Meaning that the axioms of geometry X genuinely do imply the proposition. That this proposition may not be true in geometry Y has nothing to do with it. It is a different sense of true in that it isn't necessarily related to sensory experience - only to the interrelationships of ideas.
You are tacitly assuming that Platonists have to hold that what is formally true (proveable, derivable from axioms) is actuallty true. But a significant part of the content of Platonism is that mathematical statements are only really true if they correspond to the organisation of Plato's heaven. Platonists can say, "I know you proved that, but it isn't actually true". So there are indeed different notions of truth at play here. Which is not to defend Platonism. The notion of a "real truth" which can't be publically assessed or agreed upon in the way that formal proof can be is quite problematical.
He says that counterfactuals do have a truth value, though IMO he's a bit vague about what that is (or maybe it's me who can't fully understand what he says).

She’d underlined true and truth wherever it appeared in the essay, probably about twenty times, with a question mark beside each. She wanted to know what I meant by truth.

She should hand back the paper with the note, "What do you mean by 'mean'?"

If someday the vast majority of people decided that what is known as "blue" should be renamed "snarffle" then eventually it would cease to be blue. Instead it would be snarffle because that is the belief. But that doesn't change the reality that it is the wavelength 475 nm. Human beliefs determine how we interpret information, not reality.

If the above is true, aren't the postmodernists right?

I do wish that you would say "relativists" or the like here. Many of your readers will know the word "postmodernist" solely as a slur against a rival tribe.


Actually, "relativist" isn't a lot better, because it's still pretty clear who's meant, and it's a very charged term in some political discussions.

I think it's a bad rhetorical strategy to mock the cognitive style of a particular academic discipline, or of a particular school within a discipline, even if you know all about that discipline. That's not because you'll convert people who are steeped in the way of thinking you're trying to counter, but because you can end up pushing the "undecided" to their side.

Let's say we have a bright young student who is, to oversimplify, on the cusp of going down either the path of Good ("parsimony counts", "there's an objective way to determine what hypothesis is simpler", "it looks like there's an exterior, shared reality", "we can improve our maps"...) or the path of Evil ("all concepts start out equal", "we can make arbitrary maps", "truth is determined by politics" ...). Well, that bright young student isn't a perfectly rational being. If the advocates for Good look like they're being jerks and mocking the advocates for Evil, that may be enough to push that person down the path of Evil.

Wulky Wilkinson is the mind killer. Or so it seems to me.

I agree with your point about rhetoric, but I think you give post-modern thought too little credit. First of all, Sturgeon's law says 90% of everything is crap. I can't understand why you think this statement is post-modern - or why you think it is wrong. Luminiferous Aether was possibly correct - until we tested the proposition, what basis did we have to say that ~P was better than P? This has clear flavors of post-modernism - and is false as stated. But I think someone like Foucault would want the adjective social thrown in there a bit. Given that, the diversity of cultures throughout history is some evidence that the proposition could be true - depending on what caveats we place on / how we define "arbitrary." Kuhn and Feyerabend have not always been clear on how anti-scientific realist they intended to be, but I think a proposition like "Scientific models are socially mediated" is plausible - unless Kuhn and Feyerabend totally screwed up their history. Again, post-modern flavored. And again, if we add the word "social" to the front, the statement is likely true. For example, people once thought social class (nobility, peasant, merchant) was very morally relevant. Now, not so much.
With the first item, "all concepts start out equal", consider that Occam's Razor says we should prefer simpler concepts. With "we can make arbitrary maps", I don't see how adding the word "social" in there anywhere makes it any better. Although there are many different cultures, the space of possible culture-models or culture-maps is much larger still, and so if we're trying to model how a culture works we can't just pick a map arbitrarily. Same issue applies to "truth is determined by politics". A political theory is a hypothesis about what conditions will create a given sort of society. Some political theories are better than others at such predictions.

I presume that the point with "social" is that, even if some political theories are better than others, the extent to which different theories are accepted or believed by the population at large is also strongly affected by social factors. Which, again, is an idea that has been discussed on LW a lot, and is generally accepted here...

Also, (guessing from my discussions with smart humanities people) it's saying that supposedly neutral and impartial research by scientists will be affected by a large number of (social) biases, some of them conscious, some of them unconscious, and this can have a big impact on which theory is accepted as the best and the most "experimentally tested" one. Again, not exactly a heretical belief on LW.

Ironically, I always thought that many of the posts on LW were using scientific data to show what my various humanities friends had been saying all along.

Particularly since many LWers believe things like: or
Why is the former false?
Hrm? Who said those were false? My point was that these are ideas that are popular in LW and basically true, but that most LWers don't acknowledge are post-modern in origin. The first statement is a basic takeaway from Kuhn and Feyerabend. The second is basic History of Sexuality from Foucault.
Oh, sorry, didn't get your point. I think the first statement has been reinvented often, by people who read enough Kelvin quotes. The second statement is just bizarre. Clearly many people are helped by their meds. Does feeding random psych meds to random freaks produce an increase in quality of life, or at least a wide enough spread that there's a large group that gets a stable improvement? Or are you just claiming the weaker version: symptoms make sense and are treated, but all statements of the form "patients with this set of symptoms form a cluster, and shall be labeled Noun Phrase Disorder" are false? I would claim some diagnoses are reasonable, e.g. Borderline Personality with clearly forms a cluster among bloggers who talk about their mental health. And those that aren't (a whole lotta paraphilias, and ways to cut up umbrella terms) tend to change fast anyway.

Psychology has made significant strides in response to criticism from the post-modernists. The post-modern criticism of mental health treatment is much less biting than it once was.

Still, for halo effect reasons, we should be careful.

The larger point is that Eliezer's reference to post-modernism is simply a Boo Light and deserves to be called out as such.

Your link does not support your claim that post-modernists had an effect. ---------------------------------------- Fubarobfusco may have a point about boo lights, but this large thread you have spawned distracts from it and thus undercuts him. In the long run, praising postmodernists may be a good approach to diffusing boo lights, but if you want to do that, make a separate post. In the short term, doing so distracts from the point. Whether postmodernists said useful things is not relevant to whether they said what Eliezer attributes to them and is not relevant to how the audience reacts to that attribution.
Many people can effectively be kept out of trouble and made easier for caretakers or relatives to care for via mild sedation. This is fairly clearly the function of at least a significant portion of psychiatric medication.
Systematic execution of the old guard doesn't count as scientific progress? Hmm, or does it?
Someone is trying to set up a strawman. Kuhn didn't advocate violent overthrow of the scientific establishment - he simply noted that generational change was an under-appreciated part of the change of scientific orthodoxy.
Someone is just trying to make a joke.
The prose wasn't quite as good as the joke's intent, so part of the effect was lost. Still, it made me smile, FWIW :P
The difference is that post-modernists believe that something like this is true for all science and use this to justify this state of affairs in psychology, whereas LWers believe that this is not an acceptable state of affairs and should be fixed. Edit: Also as MizedNuts pointed out, the diagnoses do try to cut reality at the joints, they just frequently fail due to social signaling interfering with seeking truth.
First, if physical anti-realism is true to some extent, then it is true to that extent. By contrast, if Kuhn and Feyerabend messed up the history, then physical anti-realists have no leg to stand on. People can stand what is true, for they are already enduring it. Second, folks like Foucault were at the forefront of the argument that unstated social norm enforcement via psychological diagnosis was far worse than explicit social norm enforcement. They certainly don't argue that the current state of affairs in psychology was (or is) justifiable.
Citation appreciated. Foucault was specifically trying to improve the standards of psychiatric care.

This post is better than the simple truth and I will be linking to it more often, even though this isn't as funny.

Nice illustrations.

EDIT: Reworded in praise-first style.


The other day Yvain was reading aloud from Feser and I said I wished Feser would read The Simple Truth. I don't think this would help quite as much.

The Simple Truth sought to convey the intuition that truth is not just a property of propositions in brains, but of any system successfully entangled with another system. Once the shepherd's leveled up a bit in his craftsmanship, the sheep can pull aside the curtain, drop a pebble into the bucket, and the level in the bucket will remain true without human intervention.

Good point.

I also really enjoyed this post, and specifically thought that the illustrations were much nicer than what's been done before.

However, I did notice that out of all the illustrations that were made for this post, there were about 8 male characters drawn, and 0 females. (The first picture of the Sally-Anne test did portray females, but it was taken from another source, not drawn for this post like the others.) In the future, it might be a good idea to portray both men AND women in your illustrations. I know that you personally use the "flip a coin" method for gender assignment when you can, but it doesn't seem like the illustrator does (There IS a 0.3% chance that the coin flips just all came up "male" for the drawings)

The specs given to the illustrator were stick figures. I noticed the male prevalence and requested some female versions or replacement with actual stick figures.

In the light of the illustrations' lack of gender variety it's strange that they do have a variety of skin and hair colors.
I hadn't noticed about their sex, but I did notice that they all seem to be children and no adults (EDIT: except the professor in the last picture). (BTW, the character with dark hair, pale skin, red T-shirt and blue trousers doesn't obviously look masculine to me; it might as well be a female child (too young to have boobs).)
7Eliezer Yudkowsky

nit to pick: Rod and cone cells don't send action potentials.

2Eliezer Yudkowsky
Can you amplify? I'd thought I'd looked this up.

Photoreceptor cells produce graded potential, not action potential. It goes through a bipolar cell and a ganglion cell before finally spiking, in a rather processed form.

2Eliezer Yudkowsky
Ah, thanks!

Koan answers here for:

What rule could restrict our beliefs to just propositions that can be meaningful, without excluding a priori anything that could in principle be true?

I dislike the "post utopian" example, and here's why:

Language is pretty much a set of labels. When we call something "white", we are saying it has some property of "whiteness." NOW we can discuss wavelengths and how light works, or whatnot, but 200 years ago, they had no clue. They could still know that snow is white, though. At the same time, even with our knowledge of how colors work, we can still have difficulties knowing exactly where the label "white" ends, and grey or yellow begins.

Say I'm carving up music-space. I can pretty easily classify the differences between Classical and Rap, in ways that are easy to follow. I could say that classical features a lot of instrumentation, and rap features rhythmic language, or something. But if I had lots of people spending all their lives studying music, they're going to end up breaking music space into much smaller pieces. For example, dub step and house.

Now, I can RECOGNIZE dubstep when I hear it, but if you asked me to teach you what it was, I would have difficulties. I couldn't necessarily say "It's the one that goes, like, WOPWOPWOPWOP iiinnnnnggg" if I'm a learned professor, so I'... (read more)

I think Eliezer is taking it as a given that English college professors who talk like that are indeed talking without connection to anticipated experience. This may not play effectively to those he is trying to teach, and as you say, may not even be true.

-4Eliezer Yudkowsky
In particular, "post-utopian" is not a real term so far as I know, and I'm using it as a stand-in for literary terms that do in fact have no meaning. If you think there are none of those, Alan Sokal would like to have a word with you.

There's a sense in which a lot of fuzzy claims are meaningless: for example, it would be hard for a computer to evaluate "Socrates is kind" even if the computer could easily evaluate more direct claims like "Socrates is taller than five feet". But "kind" isn't really meaningless; it would just be a lot of work to establish exactly what goes into saying "kind" and exactly where the cutoff point between "kind" and "not so kind" is.

I agree that literary critical terms are fuzzy in the same sense as "kind", but I don't think they're necessarily any more fuzzy. For example, replacing "post-utopian" with its likely inspiration "post-colonial", I don't know much about literature, but I feel pretty okay designating Salman Rushdie as "post-colonial" (since his books very often take place against the backdrop of the issues surrounding British decolonization of India) and J. K. Rowling as "not post-colonial" (since her books don't deal with issues surrounding decolonization at all.)

Likewise, even though "post-utopian" was chosen specifically to be meaningless, I can say with... (read more)

I liked your comment and have a half-formed metaphor for you to either pick apart or develop:

LW/ rationalist types tend towards hard sciences. This requires more System 2 reasoning. Their fields are like computer programs. Every step makes sense, and is understood.

Humanities tends toward more System 1 pattern recognition. This is more akin to a neural network. Even if you are getting the "right" answer, it is coming out of a black box.

Because the rationalist types can't see the algorithm, they assume it can't be "right".


I like your idea and upvoted the comment, but I don't know enough about neural networks to have a meaningful opinion on it.

I like the idea that this comment produces in my mind. But nitpickingly, a neural network is a type of computer program. And most of the professional bollocks-talkers of my acquaintance think very hard in system-two like ways about the rubbish they spout. It's hard to imagine a system-one academic discipline. Something like 'Professor of telling whether people you are looking at are angry', or 'Professor of catching cricket balls'.... I wonder if you might be thinking more of the difference between a computer program that one fully understands (a rare thing indeed), and one which is only dimly understood, and made up of 'magical' parts even though its top level behaviour may be reasonably predictable (which is how most programmers perceive most programs).
Well, in the case of answers to questions like that in the humanities what does the word 'right' actually mean? If we say a particular author is 'post utopian' what does it actually mean for the answer to that question to be 'yes' or 'no'? It's just a classification that we invented. And like all classification groups there is a set of rules characteristics that mean that the author is either post utopian or not. I imagine it as a checklist of features which gets ticked off as a person reads the book. If all the items in the checklist are ticked then the author is post utopian. If not then the author is not. The problem with this is that different people have different items in their checklist and differ in their opinion on how many items in the list need to be checked for the author to be classified as post utopian. You can pick any literary classification and this will be the case. There will never be a consensus on all the items in the checklist. There will always be a few points that everybody does not agree on. This makes me think that objectively speaking there is not 'absolutely right' or 'absolutely wrong' answer to a question like that. In hard science on the other hand. There is always an absolutely right answer. If we say: "Protons and neutrons are oppositely charged." There is an answer that is right because no matter what my beliefs, experiment is the final arbiter. Nobody who follows through the logical steps can deny that they are oppositely charged without making an illogical leap. In the literary classification, you or your neural network can go through logical steps and still arrive at an answer that is not the same for everybody. EDIT: I meant "protons and electrons are oppositely charged" not "protons and neutrons". Sorry!
One: Protons and neutrons aren't oppositely charged. Two: You're using particle physics as an example of an area where experiment is the final arbiter; you might not want to do that. Scientific consensus has more than a few established beliefs in that field that are untested and border on untestable.
Honestly, he'd be hard pressed to find a field that has better tested beliefs and greater convergence of evidence. The established beliefs you mention are a problem everywhere, and pretty much no field is backed with as much data as particle physics.
Fair enough; I had wanted to say that but don't have sufficiently intimate awareness of every academic field to be comfortable doing so. I think it works just as well to illustrate that we oughtn't confuse passing flaws in a field with fundamental ones, or the qualities of a /discipline/ with the qualities of seeking truth in a particular domain.
Press the Show help button to figure out how to italisize and bold and all that.
Was this intended to be a response to a different comment?
No, it's just that FluffyC used slashes to indicate that the word in the middle was to be italisized, so she probably hadn't read the help section, and I thought that reading the help section would, well, help FluffyC.
Oh Whoops! I mean protons and electrons! Silly mistake!
I don't think that the fact that everyone having a different checklist is the point. In this perfect, hypothetical world, everyone has the same checklist. I think that the point is that the checklist is meaningless, like having a literary genre called y-ism and having "The letter 'y' constitutes 1/26th of the text" on the checklist. Even if we can identify y-ism with our senses, the distinction is doesn't "mean" anything. It has zero application outside of the world of y-ism. It floats.
That is an important point. It is not so easy to come up up with a criterion of "meaningfulness" that excludes the stuff rationalists don't like, but doens't exclude a lot of everyday terninology at the same time. I could add that others have their own criteria of "meaningfulness". Humanities types aren't very bothered about questions like how many moons saturn has, because it doens't affect them or their society. The common factor seems to both kinds of "meaningfullness" is that they amount to "the stuff I personally consider to be worth bothering about". A concern with objective meaningfullness is still a subjective concern.
FWIW, the Moby Dick example is less stupid than you paint it, given the recurrence of whiteness as an attribute of things special or good in western culture - an idea that pre-dates the invention of race. I think a case could be made out that (1) the causality runs from whiteness as a special or magical attribute, to its selection as a pertinent physical feature when racism was being invented (considering that there were a number of parallel candidates, like phrenology, that didn't do so well memetically), and (2) in a world that now has racism, the ongoing presence of valuing white things as special has been both consciously used to reinforce it (cf the KKK's name and its connotations) and unconsciously reinforces it by association,

FWIW, the Moby Dick example is less stupid than you paint it, given the recurrence of whiteness as an attribute of things special or good in western culture - an idea that pre-dates the invention of race.

I can't resist. I think you should read Moby Dick. Whiteness in that novel is not used as any kind of symbol for good:

This elusive quality it is, which causes the thought of whiteness, when divorced from more kindly associations, and coupled with any object terrible in itself, to heighten that terror to the furthest bounds. Witness the white bear of the poles, and the white shark of the tropics; what but their smooth, flaky whiteness makes them the transcendent horrors they are? That ghastly whiteness it is which imparts such an abhorrent mildness, even more loathsome than terrific, to the dumb gloating of their aspect. So that not the fierce-fanged tiger in his heraldic coat can so stagger courage as the white-shrouded bear or shark.

If you want to talk about racism and Moby Dick, talk about Queequeg!

Not that white animals aren't often associated with good things, but this is not unique in western culture:

So in spring, when appears the constellation Visakha, the Bodhisatwa, under the appearance of a young white elephant of six defenses, with a head the color of cochineal, with tusks shining like gold, perfect in his organs and limbs, entered the right side of his mother, and she, by means of a dream, was conscious of the fact.

If that's your criteria, you could use some stand-in for computer science terms that have no meaning.

WMSCI, the World Multiconference on Systemics, Cybernetics and Informatics, is a computer science and engineering conference that has occurred annually since 1995. [...] WMSCI attracted publicity of a less favorable sort in 2005 when three graduate students at MIT succeeded in getting a paper accepted as a "non-reviewed paper" to the conference that had been randomly generated by a computer program called SCIgen

I think you are playing to what you assume are our prejudices.

Suppose X is a meaningless predicate from a humanities subject. Suppose you used it, not a simulacrum. If it's actually meaningless by the definition I give elsewhere in the thread, nobody will be able to name any Y such that p(X|Y) differs from p(X|¬Y) after a Bayesian update. Do you actually expect that, for any significant number of terms in humanities subjects, you would find no Y, even after grumpy defenders of X popped up in the thread? Or did you choose a made-up term so as to avoid flooding the thread with Y-proponents? If you expect people to propose candidates for Y, you aren't really expecting X to be meaningless.

The Sokal hoax only proves one journal can be tricked by fake jargon. Not that bona fide jargon is meaningless.

I'm sure there's a lot of nonsense, but "post-utopian" appears to have a quite ordinary sense, despite the lowness of the signal to noise ratio of some of those hits. A post-utopian X (X = writer, architect, hairdresser, etc.) is one who is working after, and in reaction against, a period of utopianism, i.e. belief in the perfectibility of the world by man. Post-utopians today are the people who believe that the promises of science have been found hollow, and ruin and destruction are all we have to look forward to.

We're all utopians here.

By this definition, wouldn't the belief that science will not lead to perfection but we can still look forward to more of what we already have (rather than ruin and destruction) be equally post-utopian?
Not as I see the word used, which appears to involve the sense of not merely less enthusiastic than, but turning away from. You can't make a movement on the basis of "yes, but not as sparkly".
Pity. "It will be kind of like it is now" is an under-utilized prediction.
Dunno, Futurama is pretty much entirely based on that.

What would he have to say? The Sokal Hoax was about social engineering, not semantics.

"Post-utopian" is a real term, and even in the absence of examples of its use, it is straightforward to deduce its (likely) meaning, since "post-" means "subsequent to, in reaction to" and "utopian" means "believing in or aiming at the perfecting of polity or social conditions". So post-utopian texts are those which react against utopianism, express skepticism at the perfectibility of society, and so on. This doesn't seem like a particularly difficult idea and it is not difficult to identify particular texts as post-utopian (for example, Koestler's Darkness at Noon, Huxley's Brave New World, or Nabokov's Bend Sinister). So I think you need to pick a better example: "post-utopian" doesn't cut it. The fact that you have chosen a weak example increases my skepticism as to the merits of your general argument. If meaningless terms are rife in the field of English literature, as you seem to be suggesting, then it should be easy for you to pick a real one. (I made a similar point in response to your original post on this subject.)

There is the literature professor's belief, the student's belief, and the sentence "Carol is 'post-utopian'". While the sentence can be applied to both beliefs, the beliefs themselves are quite different beasts. The professor's belief is something that carve literature space in a way most other literature professors do. Totally meaningful. The student's belief, on the other hand, is just a label over a set of authors the student have scarcely read. Going a level deeper, we can find an explanation for this label, which turns out to be just another label ("colonial alienation"), and then it stops. From Eliezer's main post (emphasis mine) :

Some literature professor lectures that the famous authors Carol, Danny, and Elaine are all 'post-utopians', which you can tell because their writings exhibit signs of 'colonial alienation'. For most college students the typical result will be that their brain's version of an object-attribute list will assign the attribute 'post-utopian' to the authors Carol, Danny, and Elaine.

  1. The professor have a meaningful belief.
  2. Unable to express it properly (it may not be his fault), gives a mysterious explanation.
  3. That mysterious expl
... (read more)
If that is what Eliezer meant, then it was confusing to use an example for which many people suspect that the concept itself is not meaningful. It just generates distraction, like the "Is Nixon a pacifist?" example in the original Politics is the mind-killer post (and actually,the meaningfulness of post-colonialism as a category might be a political example in the wide sense of the word). He could have used something from physics like "Heat is transmitted by convention", or really any other topic that a student can learn by rot without real understanding.
I don't think Eliezer meant all what I have written (edit: yep, he didn't). I was mainly analysing (and defending) the example to death, under Daenerys' proposed assumption that the belief in the professor's head is not floating. More likely, he picked something familiar that would make us think something like "yeah, if those are just labels, that's no use".¹ By the way is there any good example? Something that (i) clearly is meaningful, and (ii) let us empathise with those who nevertheless extract a floating belief out of it? I'm not sure. I for one don't empathise with the students who merely learn by rot, for I myself don't like loosely connected belief networks: I always wanted to understand. Also, Eliezer wasn't very explicit about the distinction between a statement, embodied in text, images, or whatever our senses can process, and belief, embodied in a heap of neurons. But this post is introductory. It is probably not very useful to make the distinction so soon. More important is to realize that ideas are not floating in the void, but are embodied in a medium: paper, computers… and of course brains. [1] We're not familiar to "post-utopianism" and "colonial alienation" specifically, but we do know the feeling generated by such literary mumbo jumbo.
Thank you! Your post helped me finally to understand what it was that I found so dissatisfying with the way I'm being taught chemistry. I'm not sure right now what I can do to remedy this, but thank you for helping me come to the realization.
If the teacher does not have a precise codification of what makes a writer "post-utopian", then how should he teach it to students? I would say the best way is a mix of demonstrating examples ("Alice is not a post-utopian; Carol is a post-utopian"), and offering generalizations that are correlated with whether the author is a post-utopian ("colonial alienation"). This is a fairly slow method of instruction, at least in some cases where the things being studied are complicated, but it can be effective. While the student's belief may not yet be as well-formed as the professor's, I would hesitate to call it meaningless. (More specifically, I would agree denotatively but object connotatively to such a classification.) I would definitely not call the belief useless, since it forms the basis for a later belief that will be meaningful. If a route to meaningful, useful belief B goes through "meaningless" belief A, then I would say that A is useful, and that calling A meaningless produces all the wrong sorts of connotations.
The example assumed bad teaching based on rote learning. Your idea might actually work. (Edit: oops, you're probably aware of that. Sorry for the noise)
To over-extend your metaphor, dubstep is electronic music with a breakbeat and certain BPM. Bassnectar described it in an inverview once as hip-hop beats at half time in breakbeat BPMs. It's really easy to tell the difference between dubstep and house, because dubstep has a broken kick..kickSNARE beat, while house has a 4/4 kick.kick.kick.kick beat. (Interestingly, the dubstep you seem to describe is what people who listened to earlier dubstep commonly call "brostep," and was inspired by one Rusko song ("Cockney Thug," if I remember correctly).) The point I mean to make by this is that most concepts do have system 2 algorithms that identify them, even if most people on LW would disagree with the social groups that advance those concepts. I have many friends and comrades that are liberal arts students, and most of the time, if they said something like "post-utopian" or "colonial alienation" they'd have a coherent system-2 algorithm for identifying which authors or texts are more or less post-utopian. Really, I agree that this is a bad example, because there are two things going on: the students have to guess the teacher's password (which is the same as if you had Skirllex teaching MUSC 202: Dubstep Identification, and only accepted "songs with that heavy wobble bass shit" as "real dubstep, bro"), and there's an alleged unspoken conspiracy of academics to have a meaningless classifier (which is maybe the same as subgenres of hard noise music, where there truly is no difference between typical songs in each subgenre, and only artist self-identification or consensus among raters can be used as a grouping strategy). As others have said better than me, the Sokal affair seems to be better evidence of how easy it is to publish a bad paper than it is evidence that postmodernism is a flawed field.
Example: an irishman arguing with a mongolian over what dragons look like.
When the Irishman is a painter and the Mongolian a dissatisfied customer, does their disagreement have meaning?
In that case, they're arguing about the wrong thing. Their real dispute is that the painting isn't what the Mongolian wanted as a result of a miscommunication which neither of them noticed until one of them had spent money (or promised to) and the other had spent days painting. So, no, even in that situation, there's no such thing as a dragon, so they might as well be arguing about the migratory patterns of unicorns.
While the English profs may consistently classify writing samples as post utopian or not, the use of the label "post utopian" should be justified by the english meanings of "post" and "utopian" in some way. "Post" and "utopian" are concepts with meaning, they're not just nonsense sounds available for use as labels. If you have no conceptual System 1 algorithm for "post utopian", and just have some consistent System 2 algorithm, it's a conceptual confusion to use compound of labels for concepts that may have nothing at all to do with your underlying System 2 defined concept. Likely the confusion serves an intellectually dishonest purpose, as in euphemism. When you see this kind of nonsense, there is some politically motivated obfuscation nearby.

A set of beliefs is not like a bag of sand, individual beliefs unconnected with each other, about individual things. They are connected to each other by logical reasoning, like a lump of sandstone. Not all beliefs need to have a direct connection with experience, but as long as pulling on the belief pulls, perhaps indirectly, on anticipated experience, the belief is meaningful.

When a pebble of beliefs is completely disconnected from experience, or when the connection is so loose that it can be pulled around arbitrarily without feeling the tug of experience, then we can pronounce it meaningless. The pebble may make an attractive paperweight, with an intricate structure made of elements that also occur in meaningful beliefs, but that's all it can be. Music of the mind, conveying a subjective impression of deep meaning, without having any.

For the hypothetical photon disappearing in the far-far-away, no observation can be made on that photon, but we have other observations leading to beliefs about photons in general, according to which they cannot decay. That makes it meaningful to say that the far away photon acts in the same way. If we discovered processes of photon decay, it would still be meaningful, but then we would believe it could be false.

Interesting idea. But how did you know how to phrase your original beliefs about photons? You could just have easily decided to describe photons as "photons obey Maxwell's equations up to an event horizon and case to exist outside of it". You could then add other beliefs like "nothing exists outside of the event horizon" which are incompatible with the photon continuing to exist. In other words, your beliefs cannot afford to be independent of one another, but you could build two different belief systems, one in which the photon continues to exist and one in which it does not, that make identical predictions about experiences. Is it meaningful to ask which of these belief systems is true?
Systems of belief are more like a lump of sandstone than a pile of sand, but they are also more like a lump of sandstone, a rather friable lump, than a lump of marble. They are not indissoluble structures that can be made in arbitrary shapes, the whole edifice supported by an attachment at one point to experience. Experience never brought hypotheses such as you suggest to physicists' attention. The edifice as built has no need of it, and it cannot be bolted on: it will just fall off again.
But these hypotheses have just be brought to our attention - just now. In fact the claim that these hypotheses produce indistinguishable physics might even be useful. If I want to simulate my experiences, I can save on computational power by knowing that I no longer have to keep track of things that have gone behind an event horizon. The real question is why the standard set of beliefs should be more true or meaningful than this new one. A simple appeal to what physicists have so far conjectured is not in general sufficient.
Which meaningful beliefs to consider seriously is an issue separate from the original koan, which asks which possible beliefs are meaningful. I think we are all agreeing that a belief about the remote photon's extinction or not is a meaningful one.
I don't see how you can claim that the belief that the photon continues to exist is a meaningful belief without also allowing the belief that the photon does not continue to exist to be a meaningful belief. Unless you do something along the lines of taking Kolmogorov complexity into account, these beliefs seem to be completely analogous to each other. Perhaps to phrase things more neutrally, we should be asking if the question "does the photon continue to exist?" is meaningful. On the one hand, you might want to say "no" because the outcome of the question is epiphenomenal. On the other hand, you would like this question to be meaningful since it may have behavioral implications.
They're both meaningful. There are reasons to reject one of them as false, but that's a separate issue.
OK. I think that I had been misreading some of your previous posts. Allow me the rephrase my objection. Suppose that our beliefs about photons were rewritten as "photons not beyond an event horizon obey Maxwell's Equations". Making this change to my belief structure now leaves beliefs about whether or not photons still exist beyond an event horizon unconnected from my experiences. Does the meaningfulness of this belief depend on how I phrase my other beliefs? Also if one can equally easily produce belief systems which predict the same sets of experiences but disagree on whether or not the photon exists beyond the event horizon, how does this belief differ from the belief that Carol is a post-utopian?
Dunno about “meaningful”, but the model with lower Kolmogorov complexity will give you more bang for the buck.
Your view reminds me of Quine's "web of belief" view as expressed in "Two Dogmas of Empiricism" section 6: Quine doesn't use Bayesian epistemology, unfortunately because I think it would have helped him clarify and refine his view. One way to try to flesh this intuition out is to say that some beliefs are meaningful by virtue of being subject to revision by experience (i.e. they directly pay rent), while others are meaningful by virtue of being epistemically entangled with beliefs that pay rent (in the sense of not being independent beliefs in the probabilistic sense). But that seems to fail because any belief connected to a belief that directly pays rent must itself be subject to revision by experience, at least to some extent, since if A is entangled with B, an observation which revises P(A) typically revises P(B), however slightly.

If a person with access to the computer simulating whichever universe (or set of universes) a belief is about could in principle write a program that takes as input the current state of the universe (as represented in the computer) and outputs whether the belief is true, then the belief is meaningful.

(if the universe in question does not run on a computer, begin by digitizing your universe, then proceed as above)

That has the same problem as atomic-level specifications that become false when you discover QM. If the Church-Turing thesis is false, all statements you have specified thus become meaningless or false. Even using a hierarchy of oracles until you hit a sufficient one might not be enough if the universe is even more magical than that.

But that's only useful if you make it circular. Taking you more strictly at your word than you mean it the program could just return true for the majority belief on empirically non-falsifiable questions. Or it could just return false on all beliefs including your belief that that is illogical. So with the right programs pretty much arbitrary beliefs pass as meaningful. You actually want it to depend on the state of the universe in the right way, but that's just another way to say it should depend on whether the belief is true.
2Scott Alexander
That's a problem with all theories of truth, though. "Elaine is a post-utopian author" is trivially true if you interpret "post-utopian" to mean "whatever professors say is post-utopian", or "a thing that is always true of all authors" or "is made out of mass". To do this with programs rather than philosophy doesn't make it any worse. I'm suggesting is that there is a correspondence between meaningful statements and universal computer programs. Obviously this theory doesn't tell you how to match the right statement to the right computer program. If you match the statement "snow is white" to the computer program that is a bunch of random characters, the program will return no result and you'll conclude that "snow is white" is meaningless. But that's just the same problem as the philosopher who refuses to accept any definition of "snow", or who claims that snow is obviously black because "snow" means that liquid fossil fuel you drill for and then turn into gasoline. If your closest match to "post-utopian" is a program that determines whether professors think someone is post-utopian, then you can either conclude that post-utopian literally means "something people call post-utopian" - which would probably be a weird and nonstandard word use the same way using "snow" to mean "oil" would be nonstandard - or that post-utopianism isn't meaningful.
Yeah, probably all theories of truth are circular and the concept is simply non-tabooable. I agree your explanation doesn't make it worse, but it doesn't make it better either.
Doesn't this commit you to the claim that at least some beliefs about whether or not a particular Turing machine halts must be meaningless? If they are all meaningful and your criterion of meaningfulness is correct, then your simulating computer solves the halting problem. But it seems implausible that beliefs about whether Turing machines halt are meaningless.
Input->Black box->Desired output. "Black box" could be replaced with"magic." How would your black box work in practice?
That doesn't help us decide whether there are stars outside the cosmological horizon.
7Scott Alexander
I feel like writing a more intelligent reply than "Yes it does", so could you explain this further?
Suppose we are not living in a simulation. We are to digitize our universe. Do we make our digitization include stars outside the cosmological horizon? By what principle do we decide? (I suppose you could be asking us to actually digitize the universe, but we want a principle we can use today.)
0Scott Alexander
Well, if the universe actually runs on a computer, then presumably that computer includes data for all stars, not just the ones that are visible to us. If the universe doesn't run on a computer, then you have to actually digitize the universe so that your model is identical to the real universe as if it were on a computer, not stop halfway when it gets too hard or physically impossible. I don't think any of these principles will actually be practical. Even the sense-experience principle isn't useful. It would classify "a particle accelerator the size of the Milky Way would generate evidence of photinos" as meaningful, but no one is going to build a particle accelerator the size of the Milky Way any more than they are going to digitize the universe. The goal is to have a philosophical tool, not a practical plan of action.
Oh, I didn't express myself clearly in the last paragraph of the grandparent. Don't worry, I'm not trying to demand any kind of practical procedure. I think we're on the same page. However: I don't think we can really say that in general. Perhaps if the computer stored the locations and properties of stars in an easy-to-understand way, like a huge array of floating-point numbers, and we looked into the computer's memory and found a whole other universe's worth of extra stars, with spatial coordinates that prevent us from ever interacting with them, then we would be comfortable saying that those stars exist but are invisible to us. But what if the computer compresses star location data, so the database of visible stars looks like random bits? And then we find an extra file in the computer, which is never accessed, and which is filled with random bits? Do we interpret those as invisible stars? I claim that there is no principled, objective way of pointing to parts of a computer's memory and saying "these bits represent stars invisible to the simulation's inhabitants, those do not". I'm suspicious of the phrase "identical to the real universe as if it were on a computer". It seems like a black box. Suppose we commission a digital model of this universe, and the engineer in charge capriciously programs the computer to delete information about any object that passes over the cosmological horizon. But they conscientiously program the computer to periodically archive snapshots of the state of the simulation. It might look like this model does not contain spaceships that have passed over the cosmological horizon. But the engineer points out that you can easily extrapolate the correct location of the spaceship from the archived snapshots — the initial state and the laws of physics uniquely determine the present location of the spaceship beyond the cosmological horizon, if it exists. The engineer claims that the simulation actually does contain the spaceship outside the co
Why should it if its purpose is to simulate reality for humans? What's wrong with a version of The Truman Show?
0Scott Alexander
Because since everything would be a simulation, "all stars" would be identical in meaning with "all stars that are being simulated" and with "all stars for which the computer includes data".
In a Truman Show situation, the simulators would've shown us white pin-pricks for thousands of years, and then started doing actual astrophysics simulations only when we got telescopes.
A variant of Löb's theorem, isn't it? Edit: Downvoted because the parallels are too obvious, or because the comparison seems too contrived? "E"nquiring minds want to know ...

Before reading other answers, I would guess that a statement is meaningful if it is either implied or refuted by a useful model of the universe - the more useful the model, the more meaningful the statement.

This is incontrovertibly the best answer given so far. My answer was that a proposition is meaningful iff an oracle machine exists that takes as input the proposition and the universe, outputs 0 if the proposition is true and outputs 1 if the proposition is false. However, this begs the question, because an oracle machine is defined in terms of a "black box".
Looking at Furslid's answer, I discovered that my definition is somewhat ambiguous - a statement may be implied or refuted by quite a lot of different kinds of models, some of which are nearly useless and some of which are anything but, and my definition offers no guidance on the question of which model's usefulness reflects the statement's meaningfulness. Plus, I'm not entirely sure how it works with regards to logical contradictions.
Where Recursive Justification Hits Bottom and its comment thread should be interesting to you. In the end, we have to rely on the logical theory of probability (as well as standard logical laws, such as the law of noncontradiction). There is no better choice. Using Bayes' theorem (beginning with priors set by Occam's Razor) tells you how useful your model is.
I think I was unclear. What I was considering was along the following lines: What occurred to me just now, as I wrote out the example, is the idea of simplicity. If you penalize models that add complexity without addition of practical value, the professor's list will be rapidly cut from almost any model more general than "what answer will receive a good grade on this professor's tests?"

For a belief to be meaningful you have to be able to describe evidence that would move your posterior probability of it being true after a Bayesian update.

This is a generalization of falsifiability that allows, for example, indirect evidence pertaining to universal laws.

How about basic logical statements? For example: If P, then P. I think that belief is meaningful, but I don't think I could coherently describe evidence that would make me change it's probability of being true.
Possible counterexample: "All possible mathematical structures exist."
You'd have to define "exist", because mathematical structures in themselves are just generalized relations that hold under specified constraints. And once you defined "exist", it might be easier to look for Bayesian evidence - either for them existing, or for a law that would require them to exist. As a general thing, my definition does consider under-defined assertions meaningless, but that seems correct.
Yeah, I'm not really sure how to interpret "exist" in that statement. Someone that knows more about Tegmark level IV than I do should weigh in, but my intuition is that if parallel mathematical structures exist that we can't, in principle, even interact with, it's impossible to obtain Bayesian evidence about whether they exist.
If we couldn't, even in principle, find any evidence that would make the theory more likely or less, then yeah I think that theory would be correctly labeled meaningless. But, I can immediately think of some evidence that would move my posterior probability. If all definable universes exist, we should expect (by Occam) to be in a simple one, and (by anthropic reasoning) in a survivable one, but we should not expect it to be elegant. The laws should be quirky, because the number of possible universes (that are simple and survivable) is larger than the subset thereof that are elegant.
Why? That assumes the universes are weighted by complexity, which isn't true in all Tegmark level IV theories.
Consider "Elaine is a post-utopian and the Earth is round" This statement is meaningless, at least in the case where the Earth is round, where it is equivalent to "Elaine is a post-utopian." Yet it does constrain my experience, because observing that the Earth is flat falsifies it. If something like this came to seem like a natural proposition to consider, I think it would be hard to notice it was (partly) meaningless, since I could still notice it being updated. This seems to defeat many suggestions people have made so far. I guess we could say it's not a real counterexample, because the statement is still "partly meaningful". But in that case it would be still be nice if we could say what "partly meaningful" means. I think that the situation often arises that a concept or belief people throw around has a lot of useless conceptual baggage that doesn't track anything in the real world, yet doesn't completely fail to constrain reality (I'd put phlogiston and possibly some literary criticism concepts in this category). My first attempt is to say that a belief A of X is meaningful to the extent that it (is contained in / has an analog in / is resolved by) the most parsimonious model of the universe which makes all predictions about direct observations that X would make.
A solution to that particular example is already in logic - the statements "Elaine is a post-utopian" and "the Earth is round" can be evaluated separately, and then you just need a separate rule for dealing with conjunctions.
For every meaningful proposition P, an author should (in theory) be able to write coherently about a fictional universe U where P is true and a fictional universe U' where P is false.

So my belief that 2+2=4 isn't meaningful?

I thought Eliezer's story about waking up in a universe where 2+2 seems to equal 3 felt pretty coherent. edit: It seems like the story would be less coherent if it involved detailed descriptions of re-deriving mathematics from first principles. So perhaps ArisKatsaris' definition leaves too much to the author's judgement in what to leave out of the story.
I think that it's a good deal more subtle than this. Eliezer described a universe in which he had evidence that 2+2=3, not a universe in which 2 plus 2 was actually equal to 3. If we talk about the mathematical statement that 2+2=4, there is actually no universe in which this can be false. On the other hand in order to know this fact we need to acquire evidence of it, which, because it is a mathematical truth, we can do without any interaction with the outside world. On the other hand if someone messed with your head, you could acquire evidence that 2 plus 2 was 3 instead, but seeing this evidence would not cause 2 plus 2 to actually equal 3.
On the contrary. Imagine a being that cannot (due to some neurological quirk) directly percieve objects - it can only percieve the spaces between objects, and thus indirectly deduce the presence of the objects themselves. To this being, the important thing - the thing that needs to be counted and to which a number is assigned - is the space, not the object. Thus, "two" looks like this, with two spaces: 0 0 0 Placing "two" next to "two" gives this: 0 0 0 0 0 0 Counting the spaces gives five. Thus, 2+2=5.
I think you misunderstand what I mean by "2+2=4". Your argument would be reasonable if I had meant "when you put two things next to another two things I end up with four things". On the other hand, this is not what I mean. In order to get that statement you need the additional, and definitely falsifiable statement "when I put a things next to b things, I have a+b things". When I say "2+2=4", I mean that in the totally abstract object known as the natural numbers, the identity 2+2=4 holds. On the other hand the Platonist view of mathematics is perhaps a little shaky, especially among this crowd of empiricists, so if you don't want to accept the above meaning, I at least mean that "SS0+SS0=SSSS0" is a theorem in Peano Arithmetic. Neither of these claims can be false in any universe.
I think I understand what CCC means by the being that perceives spaces instead of objects - Peano Arithmetic only exists because it is useful for us, human beings, to manipulate numbers that way. Given a different set of conditions, a different set of mathematical axioms would be employed.
Peano Arithmetic is merely a collection of axioms (and axiom schema), and inference laws. It's existence is not predicated upon its usefulness, and neither are its theorems. I agree that the fact that we actually talk about Peano Arithmetic is a consequence of the fact that it (a) is useful to us (b) appeals to our aesthetic sense. On the other hand, although the being described in CCC's post may not have developed Peano's axioms on their own, once they are informed of these axioms (and modus ponens, and what it means for something to be a theorem), they would still agree that "SS0+SS0=SSSS0" in Peano Arithmetic. In summary, although there may be universes in which the belief "2+2=4" is no longer useful, there are no universes in which it is not true.
I freely concede that a tree falling in the woods with no-one around makes acoustic vibrations, but I think it is relevant that it does not make any auditory experiences. In retrospect, however, backtracking to the original comment, if "2+2=4" were replaced by "not(A and B) = (not A) or (not B)", I think my argument would be nearly untenable. I think that probably suffices to demonstrate that ArisKatsaris's theory of meaningfulness is flawed.
How is it relevant? CCC was arguing that "2+2=4" was not true in some universes, not that it wouldn't be discovered or useful in all universes. If your other example makes you happy that's fine, but I think it would be possible to find hypothetical observers to whom De Morgan's Law is equally useless. For example, the observer trapped in a sensory deprivation chamber may not have enough in the way of actual experiences for De Morgan's Law to be at all useful in making sense of them.
In my opinion, saying "2+2=4 in every universe" is roughly equivalent to saying "1.f3 is a poor chess opening in every universe" - it's "true" only if you stipulate a set of axioms whose meaningfulness is contingent on facts about our universe. It's a valid interpretation of the term "true", but it is not the only such interpretation, and it is not my preferred interpretation. That's all.
If this is the case, then I'm confused as to what you mean by "true". Let's consider the statement "In the standard initial configuration in chess, there's a helpmate in 2". I imagine that you consider this analogous to your example of a statement about chess, but I am more comfortable with this one because it's not clear exactly what a "poor move" is. Now, if we wanted to explain this statement to a being from another universe, we would need to taboo "chess" and "helpmate" (and maybe "move"). The statement then unfolds into the following: "In the game with the following set of rules... there is a sequence of play that causes the game to end after only two turns are taken by each player" Now this statement is equivalent to the first, but seems to me like it is only more meaningful to us than it is to anyone else because the game it describes matches a game that we, in a universe where chess is well known, have a non-trivial probability of ever playing. It seems like you want to use "true" to mean "true and useful", but I don't think that this agrees with what most people mean by "true". For example, there are infinitely many true statements of the form "A+B=C" for some specific integers A,B,C. On the other hand, if you pick A and B to be random really large numbers, the probability that the statement in question will ever be useful to anyone becomes negligible. On the other hand, it seems weird to start calling these statements "false" or "meaningless".
You're right, of course. To a large extent my comment sprung from a dislike of the idea that mathematics possesses some special ontological status independent of its relevance to our world - your point that even those statements which are parochial can be translated into terms comprehensible in a language fitted to a different sort of universe pretty much refutes that concern of mine.
I suppose it depends on how stict you are about what "coherently" means. A fictional universe is not the same as a possible universe and you pobably could write about a universe where you put two apples next to two other apples and then count five apples.
Hmm, I get your point, upvoting -- but I'm not sure that "2+2=4" is meaningful in the same sense that "Bob already had 2 apples and bought 2 more apples, he was now in possession of 4 apples" is meaningful. To the extent that 2+2=4 is just a matter of extending mathematical definitions from Peano Arithmetic, it's as meaningful as saying 1=1 -- less a matter of beliefs, and more of a matter of definitions. And as far as it represents real events occurring, we can indeed imagine surreal fictional universes where if you buy two apples when you have already two apples, you end up in possession of five or six or zero apples...
A variation on this question "what rule could restrict our beliefs to just propositions that can be decided, without excluding a priori anything true?" is known to hopeless in a strong sense. Incidentally I think the phrase "in principle" isn't doing any work in your koan.
Meaningful seems like a odd word to choose, as it contains the answer itself. What rule restricts our beliefs to just propositions that can be meaningful? Why, we could ask ourselves if the proposition has meaning. The "atoms" rule seems fine, if one takes out the word "atoms" and replaces it with "state of the universe," with the understanding that "state" includes both statics and dynamics. Thus, we could imagine a world where QM was not true, and other physics held sway- and the state of that world, including its dynamics, would be noticeably different than ours. And, like daenerys, I think the statement that "Elaine is a post-utopian" can be meaningful, and the implied expanded version of it can be concordant with reality. [edit] I also wrote my koan answers as I was going through the post, so here's 1: And 2:
I very much like your response to (1) - I think the point about having access to a common universe makes it very clear.
Beliefs must pay rent.
Insufficient: the colony ship leaves no evidence.
How about an expanded version: if we could be a timeless spaceless perfect observer of the universe(s), what evidence would we expect to see?
Can you guarantee that a TSPO wouldn't see epiphenomenal consciousness?
Well, no. How is that different from epiphenomenal spaceships? Our modal predicts spaceships but no p-zombies.
I suspect that, if we are born, we already have a first model of physics, a few built-in axioms. As we grow older, we acquire beliefs that are only recursive applications and elaborations of these axioms. I would say that, if a belief can be reduced to this lowest level of abstraction, it is a meaningful belief.
Proposition p is meaningful relative to the collection of possible worlds W if and only if there exist w, w' in W such that p is true in the possible world w and false in the possible world w'. Then the question become: to be able to reason in all generality what collection of possible worlds should one use? That's a very hard question.
They are truisms - in principle they are statements that are entirely redundant as one could in principle work out the truth of them without being told anything. However, principle and practice are rather different here - just because we could in principle reinvent mathematics from scratch doesn't mean that in practice we could. Consequently these beliefs are presented to us as external information rather than as the inevitable truisms they actually are.
"God's-eye-view" verificationism A proposition P is meaningful if and only if P and not-P would imply different perceptions for a hypothetical entity which perceives all existing things. (This is not any kind of argument for the actual existence of a god. Downvote if you wish, but please not due to that potential misunderstanding.)
Doesn't that require such an entity to be logically possible?
No, in fact it works better on the assumption that there is no such entity. If it could be an existing entity, then we could construct a paradoxical proposition, such as P="There exists an object unperceived by anything.", which could not be consistently evaluated as meaningful or unmeaningful. Treating a "perceiver of all existing things" as a purely hypothetical entity--a cognitive tool, not a reality--avoids such paradoxes.
Huh? We're talking past each other here. If there's an all-seeing deity, P is well-formed, meaningful, and false. Every object is perceived by the deity, including the deity itself. If there's no all-seeing deity, the deity pops into hypothetical existence outside the real world, and evaluates P for possible perceiving anythings inside the real world; P is meaningful and likely true. But that's not what I was talking about. I'm talking about logical possibility, not existence. It's okay to have a theory that talks about squares even though you haven't built any perfect squares, and even if the laws of physics forbid it, because you have formal systems where squares exist. So you can ask "What is the smallest square that encompasses this shape?", with a hypothetical square. But you can't ask "What is the smallest square circle that encompasses this shape?", because square circles are logically impossible. I'm having a hard time finding an example of an impossible deity, not just a Turing-underpowered one, or one that doesn't look at enough branches of a forking system. Maybe a universe where libertarian free will is true, and the deity must predict at 6AM what any agent will do at 7AM - but of course I snuck in the logical impossibility by assuming libertarian free will.
Oh, oops. My mental model was this: Consider an all-perceiving entity (APE) such that, for all actually existing X, APE magically perceives X. That's all of the APE's properties -- I'm not talking about classical theism or the God of any particular religion -- so it doesn't look to me like there are logical problems. Mostly agreed. But that's not the GEV verificationism I suggested. The above paragraph takes the form "Evaluate P given APE" and "Evaluate P given no-APE". My suggestion is the reverse; it takes the form "Evaluate APE's perceptions given P" and "Evaluate APE's perceptions given not-P". If the great APE counts as a real thing, what would its set of perceptions be given that there exists an object unperceived by anything? That's simply to build a contradiction: APE sees everything, and there's something APE doesn't see. But if the all-perceiving entity is assumed not to be a real thing, the problem goes away.
Propositions must be able in principle to be connected to a state of how the world could-be, and this connection must be durable over alternate states of basic world identity. That is to say, it should be possible to simulate both states in which the proposition is true, and states in which it is not.
I don't think there can be any such rule.
Internal consistency. Propositions must be non self-contradictory. If a proposition is a conjunction of multiple propositions, then those propositions must not contradict each other.
I think the condition is necessary but not sufficient. How would it deal with the post-utopian example in the article text?
When we try to build a model of the underlying universe, what we're really talking about it is trying to derive properties of a program which we are observing (and a component of), and which produces our sense experiences. Probably quite a short program in its initial state, in fact (though possibly not one limited by the finite precision of traditional Turing Machines). So, that gives us a few rules that seem likely to be general: the underlying model must be internally consistent and mathematically describable, and must have a total K-complexity less than the amount of information in the observable universe (or else we couldn't reason about it). So the question to ask is really "can I imagine a program state that would make this proposition true, given my current beliefs about my organization of the program?" This is resilient to the atoms / QM thing, at least, as you can always change the underlying program description to better fit the evidence. Although, in practice, most of what intelligent entities do can more precisely be described as 'grammar fitting' than 'program induction.' We reason probabalistically, essentially by throwing heuristics at a wall to see what offers marginal returns on predicting future sense impressions, since trying to guess the next word in a sentence by reverse-deriving the original state of the universe-program and iterating it forwards is not practical for most people. That massive mess of semi-rational, anticipatorially-justified rules of thumb is what allows us to reason in the day to day. So a more pragmatic question is 'how does this change my anticipation of future events?' or 'What sense experiences do I expect to have differently as a result of this belief?' It is only when we seek to understand more deeply and generally, or when dealing with problems of things not directly observable, that it is practical to try to reason about the actual program underlying the universe.
I'm pleased to find this post and community; the writing is thoughtful and challenging. I'm not a philosopher, so some of the post waltzes off the edge of my cognitive dance floor, yet without stumbling or missing a beat. Proposing a rule to restrict belief seems problematic; who will enforce the restriction and how will bear on the outcome being "just." So, the only just enforcer can be the individual believer. Perhaps the rule might pertain to the intersection of belief and action: beliefs may not cause actions that limit others' freedom or well-being. Person A believes the sky is blue. Person B complains that person A's belief limits their ability to believe that the sky is green. But person B's complaint is out of bounds, as it's based on B's desire for unanimity, a desire that limits others' freedom. Hmm.
For some reason, I did not find this option here (perhaps it is implied somewhere in the chains): a statement makes sense if, in principle, it is possible to imagine its sensory results in detail. Depends on whether Russell's teapot makes sense, and also suggests that 2+2=3 doesn't make sense.
Restrict propositions to observable references? (Or have a rule about falsifiablility?) The problem with the observable reference rule is that sense can be divorced from reference and things can be true (in principle) even if un-sensed or un-sensable. However, when we learn language we start by associating sense with concrete reference. Abstractions are trickier. It is the case that my sensorimotor apparatus will determine my beliefs and my ability to cross-reference my beliefs with other similar agents with similar sensorimotor apparatus will forge consensus on propositions that are meaningful and true. Falsifiability is better. I can ask another human is Orwell post-Utopian? They can say 'hell no he is dystopian'... But if some say yes and some say no, it seems I have an issue with vagueness which I would have to clarify with some definition of criteria for post-Utopian and dystopian. Then once we had clarity of definition we could seek evidence in his texts. A lot of humanities texts however just leave observable reference at the door and run amok with new combinations of sense. Thus you get unicorns and other forms of fantasy...
All the propositions must be logical consequences of a theory that predicts observation, once you've removed everything you can from the theory without changing its predictions, and without adding anything.
It seems to me that we at least have to admit two different classes of proposition: 1) Propositions that reflect or imply an expectation of some experiences over others. Examples include the belief that the sky is blue, and the belief that we experience the blueness of the sky mediated by photons, eyes, nerves, and the brain itself. 2) Propositions that do not imply a prediction, but that we must believe in order to keep our model of the world simple incomprehensible. An example of this would be the belief that the photon continues to exist after taxes outside of our light cone.
Solomonoff induction! Just kidding.
If I, given a universal interface to a class of sentient beings, but without access to that being's language or internal mind-state, could create an environment for each possible truth value of the statement, where any experiment conducted by a being of that class upon the environment would reflect the environment's programmed truth value of the statement, and that being could form a confidence of belief regarding the statement which would be roughly uniform among beings of that class and generally leaning in the direction of the programmed truth value, then the statement has meaning. In other words, I put on my robe and wizard's cap, and you put on your haptic feedback vest and virtual reality helmet, and you tell me whether Elaine is a Post-Utopian. This should cover propositions whose truth-value might not be knowable by us within our present universe if we can craft the environment such that it is knowable via the interface to the observer. e.g. hyperluminal messaging / teleportation / "pause" mode / "ghost" mode, debug HUDs, etc.
Explicitly assuming realism and reductionism. I think. A meaningful statement is one that claims the "actual reality" lies within a particular well-defined subset of possible worlds, where each possible world is a complete and concrete specification of everything in that universe, to the highest available precision, at the lowest possible level of description, in that universes own ontology. Of course, superhuge uncomputable subsets of possible worlds are not practically useful, so we compress by talking about concepts (like "white", "snow", "5", "post-utopian"), among other things. Unfortunately, once we get into turing-complete compression, we can construct programs (concepts) that do all sorts of stupid stuff (like not halt). Concepts need to be portable between ontologies. This might sink this whole idea. For example, "snow is white" says the One True Reality is within the (unimaginably huge) subset of possible worlds where the substructures that the "snow" concept matches are also matched by the "white" concept. For example "2 + 2 = 5" refers to the subset of possible worlds where the concept generated by the application of the higher-order concept "+" to "2" and "2" will match everything matched by "5". (I unpacked "=" to "concepts match same things", but you don't have to) There's something really neat about these abstract concepts, but they don't seem fundamentally different from other ones. TL;DR: So the rule is "your beliefs should be specified by a probability distribution over exact possible worlds", and I don't know of a compression language for possible world subsets that can't express meaningless concepts (and it probably isn't worth it to look for one).
"A statement can be meaningful if a test can be constructed that will return only one result, in all circumstances, if the statement is true." Consider the satement: If I throw an object off this cliff, then the object will fall. The test is obvious; I can take a wide variety of objects (a bowling ball, a rock, a toy car, and a set of music CDs by ) and throw them off the cliff. I can then note that all of them fall, and therefore improve the probability that the statement is true. I can then take one final object, a helium balloon, and throw it off the cliff; as the balloon rises, however, I have therefore shown that the statement is false. (A more correct version would be "if I throw a heavier-than-air object off this cliff, then the object will fall." It's still not completely true yet - a live pigeon is heavier than air - but it's closer). By this test, however, the statement "Carol is a post-utopian author" is meaningful, as long as there exist some features which are the features of post-modern authors (the features do not need to be described, or even known, as long as their existence can be proven - repeatable, correct classification by a series of artificial neural networks would prove that such features exist).
Here's my first swing at it: A proposition is meaningful if it constrains the predicted observations of any theoretically possible observer. This way, the proposition "the unmanned starship will not blink out of existence when it leaves my light cone" is meaningful because it's possible that there might potentially be an observer nearby who observes the starship not disappear. On the other hand, the statement "The position of this particle is exactly X and its momentum exactly P" is not meaningful under this rule, and that's a feature.
Taboo "theoretically possible".
Hm, how about: "[...] of any observer which our best current theory of how minds work says could exist". So for example, a statement along the lines of "a ghost watches and sees whether or not Mars continues to exist when it passes behind the Sun from Earth's perspective" would have been meaningful a long time ago, but is not meaningful for people today who know a little about brains. This also means that a proposition may be meaningful only because the proposer is ignorant.
Taboo "could". Basically, counter-factual surgery is a lot trickier than you seem to think.
There aren't many threads where I'd let that pass. This is one of them.
My $0.02: A proposition P is meaningful to an observer O to the extent that O can alter its expectations about the world based on P. This doesn't a priori exclude anything that could be true, although for any given observer it might do so. As it should. Not every true proposition is meaningful to me, for example, and some true propositions that are meaningful to me aren't meaningful to my mom. Of course, it doesn't necessarily exclude things that are false, either. (Nor should it. Propositions can be meaningful and false.) For clarity, it's also perhaps worth distinguishing between propositions and utterances, although the above is also true of meaningful utterances.
Maps are models of the territory. And the usefulness of them is often that they make predictions about parts of the territory I haven't actually seen yet, and may have trouble getting to at all. The Sun will come up in the morning. There isn't a leprachaun colony living a mile beneath my house. There aren't any parts of the moon that are made of cheese. I have no problem saying that these things are true, but they are in fact extrapolations of my current map into areas which I haven't seen and may never see. These statements don't meaningfully stand alone, they arise out of extrapolating a map that checks out in all sorts of other locations which I can check. One can then have meaningful certainty about the zones that haven't yet been seen. How does one extrapolate a map? In principle I'd say that you should find the most compressible form - the form that describes the territory without adding extra 'information' that I've assumed from someplace else. The compressed form then leads to predictions over and above the bald facts that go into it. The map should match the territory in the places you can check. When I then make statements that something is "true", I'm making assertions about what the world is like, based on my map. As far as English is concerned, I don't need absolute certainty to say something is true, merely reasonable likelihood. Hence the photon. The most compressible form of our description of the universe is that the parts of space that are just beyond visibility aren't inherently different from the parts we can see. So the photon doesn't blink out over there, because we don't see any such blinking out over here.
If by "meaningful" you mean "either true or false" and by "meaningless" you mean "neither true nor false", then a Platonist and a formalist would disagree about the meaningfulness of the continuum hypothesis. Since I don't know any knockdown argument for either Platonism or formalism, I defy everyone who claims to have a crisp answer to your question, including possibly you.
OK. Here's my best shot at it. Firstly, I don't really like the wording of the Koan. I feel like a more accurate statement of the fundamental problem here is "What rule could restrict our beliefs to propositions that we can usefully discuss whether or not they are true without excluding any statements for which we would like be base our behavior on whether or not they are true." Unfortunately, on some level I do not believe that there is a satisfactory answer here. Though it is quite possible that the problem is with my wanting to base my behavior on the truth of statements whose truth cannot be meaningfully discussed. To start with, let's talk about the restriction about restricting to statements for which we can meaningfully discuss whether or not they are true. Given the context of the post this is relatively straightforward. If truth is an agreement between our beliefs and reality, and if reality is the thing that determines our experiences, then it is only meaningful to talk about beliefs being true if there are some sequences of possible experiences that could cause the belief to be either true or false. This is perhaps too restrictive a use of "reality", but certainly such beliefs can be meaningfully discussed. Unfortunately, I would like to base my actions upon beliefs that do not fall into this category. Things like "the universe will continue to exist after I die" does not have any direct implications on my lifetime experiences, and thus would be considered meaningless. Fortunately, I have found a general transformation that turns such beliefs into beliefs that often have meaning. The basic idea is to instead of asking directly about my experiences to instead use Solomonoff induction to ask the question indirectly. For example, the question above becomes (roughly) "will the simplest model of my lifetime experiences have things corresponding to objects existing at times later than anything correspond to me?" This new statement could be true (as it is wit
Only propositions that constrain our sensory experience are meaningful. If it turns out that the cosmologists are wrong and the universe begins to contract, we will have the opportunity to make contact with the civilization that the colonization starship spawns. The proposition "The starship exists" entails that the probability of the universe contracting and us making contact with the descendants of the passengers of the starship is substantial compared to the probability of the universe contracting.

Counter-example. "There exists at least one entity capable of sensory experience." What constraints on sensory experience does this statement impose? If not, do you reject it as meaningless?

Heh. Okay, this and dankane's similar proposition are good counterexamples.
5Scott Alexander
Least convenient possible world - we discover the universe will definitely expand forever. Now what? Or what about the past? If I tell you an alien living three million years ago threw either a red or a blue ball into the black hole at the center of the galaxy but destroyed all evidence as to which, is there a fact of the matter as to which color ball it was?
"Possible" is an important qualifier there. Since 0 and 1 are not probabilities, you are not describing a possible world.
The comment doesn't lose too much if we take 'definite' to mean 0.99999 instead of 1. (I would tend to write 'almost certainly' in such contexts to avoid this kind of problem.)
Yvain's objection fails if "definitely" means "with probability 0.99999". In that case the conditional probability P( encounter civilization | universe contracts) is well-defined.
Oh, I thought I retracted the grandparent. Nevermind---it does need more caveats in the expression for it to return to being meaningful.
I think it loses its force entirely in that case. Nisan's proposal was a counterfactual, and Yvain's counter was a possible world where that counterfactual cannot obtain. Since there is no such possible world, the objection falls flat.
If this claim is meaningful, isn't Nisan's proposal false?
No. Why would that be?
I suspect that the answer to the alien-ball case may be empirical rather than philosophical. Suppose that there existed quantum configurations in which the alien threw in a red ball, and there existed quantum configurations in which the alien threw in a blue ball, and both of those have approximately equal causal influence on the configuration-cluster in which we are having (approximately) this conversation. In this case, we would happen to be living in a particular type of world such that there was no fact of the matter as to which color ball it was (except that e.g. it mostly wasn't green).
You're right, my principle doesn't work if there's something we believe with absolute certainty. If we later find out that the alien did in fact leave some evidence, and recover that evidence, we'll have an opinion about the color of the ball.
This seems to be avoiding Yvain's question by answering a preferred one. The position expressed so far, combined with the avoidance here would seem to give the answer 'No'.
What about the proposition "the universe will cease to exist when I die" (using some definition of "die" that precludes any future experiences, for example, "die for the last time")? Then the truth of this proposition does not constrain sensory input (because it only makes claims about times after which you have no sensory input), but does have behavioral ramifications if you are, for example, deciding whether or not to write a will.

I don't think EY has chosen the most useful way to proceed on a discussion of truth. He has started from an anecdote where the correspondence theory of truth is the most applicable, and charges ahead developing the correspondence theory.

We call some beliefs true, and some false. True and false are judgments we apply to beliefs - sorting them into two piles. I think the limited bandwidth of a binary split should already be a tip off that we're heading down the wrong path.

In practice, ideas will be more or less useful, with that usefulness varying depending on the specifics of the context of the application of those beliefs. Even taking "belief as predictive model" as given, it's not that a belief is either accurate or inaccurate, but it will be more or less accurate, and so more or less useful, as I've claimed is the general case of interest.

Going back to the instrumental versus epsitemic distinction, I want to win, and having a model that accurately predicts events is only one tool for winning among many. It's a wonderful simulation tool, but not the only thing I can do with beliefs.

If I'm going to sort beliefs into more and less useful, the first thing to do is identif... (read more)