I wouldn’t guess that I have a big disagreement with you once we translate everything into each other’s preferred way of speaking. I think you’re pointing at something important and right. But I sure won’t habitually call it “the problem of the criterion”, or use any of the explanatory devices presented here.
So, let’s see what you think of my take.
I think “The Problem of the Criterion” as framed here (and, I guess, most places) creates a heavy temptation to conflate several different issues.
The problem of responding to the skeptic. The way the problem of the criterion is phrased uses “knowledge”, which tempts one to think of perfect knowledge. This makes the problem seem acute, because plausibly, nothing can be known with certainty. But this makes it sound as if the alternative to solving the problem of the criterion is falling into complete skepticism. For Bayesians (and some other epistemologies), it would be fine if nothing is known with certainty, and indeed, plausibly nothing is known with certainty. This makes the problem of the criterion sound irrelevant, for these camps. And it is, provided we understand the problem of the criterion to be something like “how can we know anything with certainty?”
The epistemic circularity of foundationalism. This is the recursion problem you mention. This is only a problem if you think each belief should have a justification, and these justifications should never form a circle, and should also avoid forming an infinite chain. But there is no mathematical structure which contains no circles, no infinite chains, and also no unjustified orphan truths! This is intuitively a real problem, since we can separately justify intuitions that each belief should be justified, that circular justifications don’t count, and that infinite backward chains of justification also don’t count. I think Eliezer’s answer to this conundrum is correct (the one which you classify as pragmatism), although I think there are some details to work out. (IE, I don’t think most people walk away from that essay with a working classifier to distinguish between bad circular arguments and acceptable ones, because it doesn’t give one; it focuses on making the case that specific instances of apparently circular reasoning are fine and good.)
The naturalistic question, where does knowledge come from? There’s a phenomenon in the world: sometimes people (and animals) seem to know things (and, sometimes not). How does this happen? How can we accurately describe and analyze the phenomenon? Again, Bayesians will tend to think they have a good answer. But this time the Bayesians will be roused to answer with vim and vigor; for this is an issue of key interest to them. Whereas problem #1 seems like a silly distraction to them.
The normative question: what should we consider ourselves to know, and how should we manage belief? This is the core question of epistemic rationality.
You also mention several other distinct problems which “are just the problem of the criterion in disguise”:
You seem to find it encouraging that the Problem of the Criterion mixes all of these problems together in one big pot. I think it’s quite problematic. In particular, it means that people who think they have solved one problem here will think they’ve solved the others. It also tempts impossibility proofs for solving one problem to be applied to the others.
I think you start out discussing the normative problem:
Oh, you used a ruler? How do you know the ruler is accurately measuring 3 inches?
This how do you know sounds like a justification question, rather than an empirical how-did-you-come-to-know; although it’s impossible to tell for sure without specific disambiguation (this is part of the problem).
But then you mix in the recursion problem. At the beginning of the section “The Problem”, your characterization of The Problem is as follows:
How do you know something is true? To know if something is true, you have some method, a criterion, by which you assess its veracity. But how do you know this criterion is itself true? Oh, you have a method, i.e. a criterion, by which you assess its veracity.
Oops! Infinite recursion detected!
So which problem is it?
If it’s the literal, empirical content of the first question asked, how do you know something is true, then we investigate by trying to remember how we know specific things, noticing when we learn new things, and inventing theoretical frameworks (such as positivism, or fallibilism, or probability theory, or neuroscience) which attempt to characterize the process. The recursive issue isn’t so much a problem as it is more data—as we arrive at tentative conclusions about how we come to know things, we can check whether our theories properly account for how we would learn those theories themselves.[1]
If it’s the normative version of the first question, how should we know things (or perhaps, how can we justify knowledge), then the recursion problem is perhaps an argument for why the question is difficult—but not the question itself. On this reading, any answer to the criterion problem must deal with the recursion problem, but dealing with the recursion problem is not in itself sufficient; we might be able to dismiss the recursion problem while lacking any reply to the full problem.
If it’s the recursive problem itself, then the criterion problem is really only a problem for foundationalists; perhaps it should be remembered as the main problem with foundationalism.
If it’s the problem of how we come to know anything (know with certainty), then the skeptical response seems totally reasonable and indeed correct. In this case the recursion problem sounds like just one way to argue for the skeptical conclusion.[2]
In the field of epistemology, the problem of the criterion is an issue regarding the starting point of knowledge. This is a separate and more fundamental issue than the regress argument found in discussions on justification of knowledge.[1]
So it would seem that the regress argument is kept distinct from the problem of the criterion, leaving my other three interpretations on the table.
However, the Internet Encyclopedia of Philosophy begins its article like this:
A popular form of the Problem of the Criterion can be raised by asking two seemingly innocent questions: What do we know? How are we to decide in any particular case whether we have knowledge? One quickly realizes how troubling the Problem of the Criterion is because it seems that before we can answer the first question we must already have an answer to the second question, but it also seems that before we can answer the second question we must already have an answer to the first question. That is, it seems that before we can determine what we know we must first have a method or criterion for distinguishing cases of knowledge from cases that are not knowledge. Yet, it seems that before we can determine the appropriate criterion of knowledge we must first know which particular instances are in fact knowledge. So, we seem to be stuck going around a circle without any way of getting our epistemological theorizing started.
Which, indeed, sounds like the regress argument.
And a bit later in the article, a severe ambiguity is confirmed:
Further, there is some ambiguity in Chisholm’s own discussions of the Problem of the Criterion as to whether the problem presented by the Problem of the Criterion is the meta-epistemological problem of determining when we have knowledge or the epistemological problem of determining what is true. So, there is a difficulty in determining exactly what problem the Problem of the Criterion is supposed to pose.
The fact that Chisholm’s discussion oscillates between these two versions of the Problem of the Criterion and the fact that he seems to be aware of the two versions of the problem help make it clear that perhaps there is no such thing as the Problem of the Criterion. Perhaps the Problem of the Criterion is rather a set of related problems. This is something that many philosophers since Chisholm, and Chisholm himself (see his 1977), have noted.
This means the recursion problem does have an interesting role to play in the empirical inquiry. The recursion becomes an interesting test case. Perhaps it becomes a big obstacle to some theories. For example, I’ve heard the accusation that Popperian falsifiability theory isn’t itself falsifiable. But perhaps this is simply because it’s better viewed as a normative theory of science, rather than an empirical one.
The empirical-recursive problem of the criterion for Bayesian behavioral scientists (aka economists) is the question of whether we can empirically justify bayesian rationality as a behavioral theory (aka, behavioral economics).
The empirical-recursive problem of the criterion for neuroscience would be whether a particular neuroscientific theory can account for neuroscience as an activity of the brain. (EG, some connectionist theories which require pre-existing concept cells for every concept will fall short, because a pre-existing ‘connectionism cell’ is arguably too implausible.)
In this case, the recursion problem can be properly highlighted by appealing to Godel’s incompleteness theorem, since Godel’s theorem does apply to this kind of crisp 100% knowledge (provided we think it meets the other conditions for the theorem).
However, the kind of skeptic I’m endorsing (a moderate bayesian skeptic, who believes in probabilistic knowledge) has no business making the recursion-problem argument, since they don’t accept the foundationalist underpinnings of such an argument.
But perhaps that’s fine; there’s a long tradition in skepticism of referring to arguments as ladders which you use to climb up to the skeptical vantage point, but then kick down, having no further use for them.
After all, the bayesian moderate-skeptic is simply trying to argue that logical certainty doesn’t exist in practice. So long as they’re arguing about absurdities, they might as well also assume the absurd foundationalist assumptions.
But it’s a bit rude of them if they try to sneak the assumptions in without explicitly flagging them.
First, this is a great, really thoughtful comment.
My initial thought is that I’m doing a bad job in this post of pointing at the thing that really matters, which is why I think we may see different framings as useful. I see the problem of the criterion as a particular instance of exposing a deeper, fundamental meta-fact about our world, which is that uncertainty is fundamental to our existence and all knowledge is teleological. I don’t think all that comes through here because when I wrote this it was my first time really trying to explain fundamental uncertainty, so I only go so far in explaining it. I got a bit too excited a confused the problem of the criterion for fundamental uncertainty itself.
That (1) seems boring to Bayesians et al. seems fine to me because they already buy that things are fundamentally uncertain, although I think in practice most aspiring Bayesians don’t really know why, which bleeds into (2). I see (2) as important because knowledge does, in fact, ground out in some foundation, and we try hard to make sure that grounding is correct by requiring justifications for our grounding. Personally, the question I’m most interested in is (3), and then considering what implications that has for (1), (2), and (4), though I don’t really have much to say about (4) here other than that one should be fundamentally uncertain and that I think in practice many people trying to work out (4) get confused because they imagine they can achieve justified true belief, or more often with aspiring Bayesians justified true meta-belief even though all object-level beliefs are uncertain.
So my I have to admit my motivations when writing this post were a bit of a mix of yelling at rationalists I meet who are performing rationality but missing this deeper thing that I was also missing out on for a long time and trying to really clarify what this deeper thing is that’s going on.
I know you’re reading some of my other posts, but let me see if I can explain how I actually think about the problem of the criterion today:
We and every dynamic process can be thought of as made up of lots of feedback circuits, and most of those circuits are negative feedback circuits (positive feedback, where it exists, is often wrapped inside a negative feedback circuit because otherwise the universe would be boring and already tiled with something like paperclips). These circuits get information by observing the world and then generating a signal to control the circuit. The process of generating the signal is necessarily lossy, and it’s because we’re going to compress information when generating the signal. As a result, the circuit makes a “choice” about what to signal on, the signal is thus teleological because the signal depends on what purpose the signal has (purpose is an idea in cybernetics I won’t expand here, but basically it’s what the circuit is designed to do), so all the knowledge (information inside the circuits) we have is both uncertain due to compression (and maybe something something quantum mechanics observation, but I don’t think we need to go there) and teleological because circuits impose teleology through their design.
I’ll admit, this is all sort of a weird way to explain it because it hinges on this unusual cybernetic model of the world, so I try to explain myself in other ways that are more familiar to folks. Maybe that’s the wrong approach, but I think it’s reasonable to come at this thing outside in (which is what I do in my book) by addressing the surface level uncertainty first and then peeling back the layers to find the fundamental uncertainty. This post is weird because it’s a bit in the middle.
(I’ll also note there’s an even deeper version of how I think about fundamental uncertainty and teleology today, but I’m not sure I can put it in words because it comes from meditation experience and I can only talk about it in mysterious ways. I think trying to talk about that way of thinking about it is probably more confusing than helpful.)
IMHO the cybernetic picture isn’t weird or uncommon; naively, I expect it to get less pushback.
I see (2) as important because knowledge does, in fact, ground out in some foundation, and we try hard to make sure that grounding is correct by requiring justifications for our grounding.
I think this is what I most want to push back on. My own sense is that you are confused about this. On my understanding, you seem to simultaneously believe that the core foundationalist assumptions make sense, and also believe an impossibility argument which shows them to be inconsistent. This doesn’t make sense to me.
My formalization here is only one possible way to understand the infinite-regress problem (although I think it does a good job of capturing the essence of it) -- but, in this formalization, the contradiction is really direct, which makes it seem pretty silly.
I also think the contradictory axioms do capture two intuitions which, like, beginning philosophy majors might endorse.
So I think the infinite regress problem should be explained to beginning philosophers as a warning against these naive assumptions about justification. (And indeed, this is how I was taught.)
But that’s what it is to me. It seems to be something else for you. Like a paradox. You write of proving the impossibility of solution, rather than resolving the problem. You write that we should “hold the problem”. Like, in some sense it is still a problem even after it has been solved.
(Perhaps the seeming contradiction is merely due to the way the criterion problem conflates multiple problems; EG, the naturalistic question of where knowledge comes from is still a live question after the justification-infinite-regress problem has been resolved.)
It makes sense to me to try to spell out the consequences of the infinite-regress problem; they may be complex and non-obvious. But for me this should involve questioning the naive assumptions of justification, and figuring out what it is they were trying to do. On my analysis, a reasonable place to go from there is the tiling agents problem and Vingean reflection. This is a more sophisticated picture of the problems rational agents run into when philosophizing about themselves, because it admits that you don’t need to already know how you know in order to know—you are already whatever sort of machine you are, and the default outcome is that you keep running as you’ve run. You’re not at risk of all your knowledge evaporating if you can’t justify it. However, there is a big problem of self-trust, and how you can achieve things over time by cooperating with yourself. And there’s an even bigger problem if you do have the ability to self-modify; then your knowledge actually might evaporate if you can’t justify it.
But this problem is to a large degree solved by the lack of a probabilistic Lob’s theorem. This means we are in a much better situation with respect to self-trust, vingean reflection, and the tiling agents problem—so long as we can accept probabilistic fallibility rather than needing perfect self-trust.
So in effect, the answer to (a newer and significantly more sophisticated version of) the recursive justification problem is the same as the answer to skepticism which you agreed was right and proper—namely, embracing probabilism rather than requiring certainty.
I expect you’ll agree with this conclusion; the part where I think we have some sort of fuzzy disagreement is the part where you still say things like “I see (2) as important because knowledge does, in fact, ground out in some sort of foundation, and we try to make sure that grounding is correct by requiring justifications for our grounding.” I think there are some different possible ways to spell out what you meant here, but I don’t think the best ones connect with (2) very much, since (2) is based on some silly-in-hindsight intuitions about justification.
So I think maybe there’s something goin on here where I’m taking too much for granted that formal systems are the best way to figure out what to believe about the world to get an accurate picture of it, and so long as you’re figuring out what to believe using some formal system then you’re forced to ground it in some assumed foundation that is not itself justified. This seems important because it means there’s a lot of interesting stuff going on with those assumptions and how they get chosen such that they cause the rest of the system to produce beliefs that track reality.
But it sounds like you think this is a confused view. I admit I feel a bit differently about things than the way they’re presented in the post, but not entirely, as my comment reflects. I think when I wrote the post I really couldn’t see my way around how to know things without being able to formally justify them. Now I better understand there can be knowing without belief that knowledge is justified or true, but I do think there’s something to the idea that I am still too much captured by my own ontology and not fully seeing it as I see through it, and this causes me to get a bit mixed up in places.
The bit about self-trust is really interesting. To get a little personal and psychological, I think the big difference between me ~4 years ago and me now is that ~4 years ago I learned to trust myself in some big important way that goes deeper than surface level self-trust. On the surface this looks like trusting my feelings and experiences, but to a certain extent anyone can do this if you just give them feedback that says trust those things more than other things. But I’m talking about something a level deeper where I trust the process by which those things get generated to do what they do and that they’re doing their best (in the sense that the universe is deterministic and we’re all always doing our best/worst because it could have been no other way).
If I’m honest I’ve really struggled to understand what’s going on with Lob’s theorem and why it seems like a big deal to everyone. But it seems like from what you’re saying it’s related to the issue that got me started on things, which is how do you deal with the problem of verifying that another agent (which could be yourself at another time) shares your values and is thus in some way aligned with you. If that’s the case maybe I already grasp the fundamentals of this obstacle Lob seems to create and have just failed to realize that’s what folks were talking about when they talk about Lob with respect to aligning AI.
@abramdemski Wanted to say thanks again for engaging with my posts and pointing me towards looking again at Lob. It’s weird: now that I’ve taken so time to understand it, it’s just what in my mind was already the thing going on with Godel, just I wasn’t doing a great job of separating out what Godel proves and what the implications are. As presented on its own, Lob didn’t seem that interesting to me so I kept bouncing off it as something worth looking at, but now I realize it’s just the same thing I learned from GEB’s presentation of Peano arithmetic and Godel when I read it 20+ years ago.
When I go back to make revisions to the book, I’ll have to reconsider including Godel and Lob somehow in the text. I didn’t because I felt like it was a bit complicated and I didn’t really need to dig into it since I think there’s already a bit too many cases where people use Godel to overreach and draw conclusions that aren’t true, but it’s another way to explain these ideas. I just have to think about if Godel and Lob are necessary: that is, do I need to appeal to them to make my key points, or are these things that are better left as additional topics I can point folks at but not key to understanding the intuitions I want them to develop.
I’ve heard Lob remarked that he would never have published if he realized earlier how close his theorem was to just Godel’s second incompleteness theorem; but I can’t seem to entirely agree with Lob there. It does seem like a valuable statement of its own.
I agree, Godel is dangerously over-used, so the key question is whether it’s necessary here. Other formal analogs of your point include Tarski’s undefinability, and the realizablility / grain-of-truth problem. There are many ways to gesture towards a sense of “fundamental uncertainty”, so the question is: what statement of the thing do you want to make most central, and how do you want to argue/illustrate that statement?
I think Eliezer’s answer to this conundrum is correct (the one which you classify as pragmatism),
I think pragmatism has the problems that the OP points out: it’s necessarily a partial solution , because not everything has a clear winning condition.
I don’t think most people walk away from that essay with a working classifier to distinguish between bad circular arguments and acceptable ones
Embracing circularity is different from pragmatism. Indeed, without the classifier you mention , it’s completely impractical!
Embracing circularity is different from pragmatism.
This is why I said “the one which you classify as pragmatism”—Gordon Worley calls it pragmatist because Eliezer’s notion of rationality is ultimately about winning; but that is quite far from being a good summary of Eliezer’s specific argument in Where Recursive Justification Hits Bottom.
I think pragmatism has the problems that the OP points out: it’s necessarily a partial solution , because not everything has a clear winning condition.
I don’t think Eliezer’s specific argument has this problem, since his advice has little to do with winning conditions. His advice is that when considering very abstract questions such as the foundations of epistemology, we have to consider these questions in the same way we consider any other questions, IE using the full strength of our epistemology.
Embracing circularity is different to both if the above, and no answer to the the problem if the criterion, since coherence is a criterion.
I think there are a bunch of related problems being labeled as the criterion problem.
One of them is the question: what is the criterion?
The possible answers to this question are (of course) proposed criteria.
For this version of the criterion problem, your objection seems absurd. Of course a proposed answer to the criterion problem is going to be a criterion! That doesn’t disqualify an answer! Rather, that is a necessary characteristic of an answer! We can’t disqualify proposed criteria simply for being criteria!
Another one of them is the infinite-recursion problem Gordon mentions (also known as the regress argument). This argument rests on the foundationalist intuitions that all beliefs need to be justified, and infinite chains of justification (such as circular justification) don’t count.
Coherence seems to also qualify as a possible answer to this version of the criterion problem, addressing the problem by pointing to which of the contradictory assumptions we should give up.
I think rejecting these assumptions is a legitimate sort of answer to the problem, since the criterion problem isn’t just supposed to be a problem within foundationalist theories—it’s not like rejecting those assumptions misses the point of the problem. Rather, the regress argument is an argument made to all thinking beings, for whom the foundationalist assumptions are tacitly supposed to be compelling. So pointing out which assumptions are not, in fact, compelling seems like a good type of reply.
So overall, I don’t buy your argument that coherence is “no answer” to the criterion problem. It seems like a reasonable sort of answer to pursue, for at least two of the central forms of the criterion problem.
One of them is the question: what is the criterion?
Another is: does it work? Another is : does anything work?
For this version of the criterion problem, your objection seems absurd. Of course a proposed answer to the criterion problem is going to be a criterion!
Circularity isn’t just a criterion, it’s one of the unholy trinity of criteria …particularism, regress and Circularity .. that are know to have problems. The problem with Circularity is that it justifies too much, not too little.
Coherence is either circularity, in which case it has the same problems; or something else, in which case it needs to be spelt out.
I don’t think EY has spelt out a fourth criterion that works. Where Regress Stops is particularism.
Coherence seems to also qualify as a possible answer to this version of the criterion problem, addressing the problem by pointing to which of the contradictory assumptions we should give up.
What do you mean by coherence?
So overall, I don’t buy your argument that coherence is “no answer” to the criterion problem. It seems like a reasonable sort of answer to pursue, for at least two of the central forms of the criterion problem.
Circularity isn’t just a criterion, it’s one of the unholy trinity of criteria …particularism, regress and Circularity .. that are know to have problems. The problem with Circularity is that it justifies too much, not too little.
My take on “justifies too much” is that coherentism shouldn’t just say coherence is the criterion, in the sense of anything coherent is allowed to fly. Instead, the (subjectively) correct approach is (something along the lines of) coherence with the rest of my beliefs.
What do you mean by coherence?
Coherence is one of the six response types cited in the OP. I was also using ‘circularity’ and ‘coherence’ somewhat synonymously. Sorry for the confusion.
When I took undergrad philosophy courses, “coherentism” and “foundationalism” were two frequently-used categories representing opposite approaches to justifiation/knowledge. Foundationalism is the idea that everything should be built up from some kind of foundation (so here we could classify it as particularism, probably). Coherentism is basically the negation of that, but is generally assumed to replace building-up-from-foundations with some notion of coherence. Neither approach is very well-specified; they’re supposed to be general clusters of approaches, as opposed to full views.
More specifically, with respect to my previous statement:
Another one of them is the infinite-recursion problem Gordon mentions (also known as the regress argument). This argument rests on the foundationalist intuitions that all beliefs need to be justified, and infinite chains of justification (such as circular justification) don’t count.
Coherence seems to also qualify as a possible answer to this version of the criterion problem, addressing the problem by pointing to which of the contradictory assumptions we should give up.
Here, we can interpret coherentism specifically as the acceptance of circular logic as valid (the inference “A, therefore A” is a valid one, for example—note that ‘validity’ is about the argument from premise to conclusion, and does not imply accepting the conclusion). This resolves the infinite regress problem by accepting regress, whereas particularism instead rejects the assumption that every statement needs a justification.
Circularity isn’t just a criterion, it’s one of the unholy trinity of criteria …particularism, regress and Circularity .. that are know to have problems. The problem with Circularity is that it justifies too much, not too little.
Coherence is either circularity, in which case it has the same problems; or something else, in which case it needs to be spelt out.
I don’t think EY has spelt out a fourth criterion that works. Where Regress Stops is particularism.
I think the details of these views matter a whole lot. Which kind of particularism? How is it justified? Does the motivation make sense? Or which coherentism, etc.
The arguments about the Big Three are vague and weak due to the lack of detail. Chisholm makes his attacks on each approach, which sound convincing as-stated, but which seem much less relevant when you’re dealing with a fully-fleshed-out view with specific arguments in its favor.
(But note that for Chisholm, the three are: particularism, methodism, skepticism. So his categorization is different from yours.)
You classify Eliezer as particularism, I see it as a type of circular/coherentist view, Gordon Worley sees need to invent a “pragmatism” category. The real answer is that these categories are vague and Eliezer’s view is what it is, regardless of how you label it.
Also note that I prefer to claim Eliezer’s view is essentially correct without claiming that it solves the problem of the criterion, since I find the criterion problem to be worryingly vague.
My take on “justifies too much” is that coherentism shouldn’t just say coherence is the criterion, in the sense of anything coherent is allowed to fly. Instead, the (subjectively) correct approach is (something along the lines of) coherence with the rest of my beliefs.
That’s how coherence usually works. If no new belief is accepted unless consistent with existing beliefs, then you don’t get as much quodlibet as under pure circular justification, but you don’t get convergence on a single truth either.
You classify Eliezer as particularism, I see it as a type of circular/coherentist view, Gordon Worley sees need to invent a “pragmatism” category.
Having re-read Where Recursive Justfication Hits Bottom, I think EY says different things in different places. If he is saying that our Rock Bottom assumptions are actually valid, that’s particularism. If he is saying that we are stuck with them, however bad they are it’s not particularism, and not a solution to epistemology.
He also offers a kind of circular defense of induction, which I don’t think amounts to full fledged circularity,because you need some empirical data to kick things off. The Circular Justifcation of Induction isn’t entirely bad, but it’s yet another partial solution, because it’s limited to the empirical and quantifiable. If you explicitly limit everyone to the empirical and quantifiable, that’s logical.positivism, but EY say he’s isn’t a logical positivist.
Also note that I prefer to claim Eliezer’s view is essentially correct without claiming that it solves the problem of the criterion, since I find the criterion problem to be worryingly vague
What are his views, and what is it correct about?
Having read them, I still dont know.
I think the details of these views matter a whole lot. Which kind of particularism? How is it justified? Does the motivation make sense? Or which coherentism, etc
It matters both ways. The solution needs to be clear, if there is one.
“Usually” being the key here. To me, the most interesting coherence theories are broadly bayesian in character.
but you don’t get convergence on a single truth either.
I’m not sure what position you’re trying to take or what argument you’re trying to make here—do you think there’s a correct theory which does have the property of convergence on a single truth? Do you think convergence on a single truth is a critical feature of a successful theory?
I don’t think it’s possible to converge on the truth in all cases, since information is limited—EG, we can’t decide all undecidable mathematical statements, even with infinite time to think. Because this is a fundamental limit, though, it doesn’t seem like a viable charge against an individual theory.
If he is saying that our Rock Bottom assumptions are actually valid, that’s particularism. If he is saying that we are stuck with them, however bad they are it’s not particularism, and not a solution to epistemology.
I don’t think he’s saying either of those things exactly. He is saying that we can question our rock bottom assumptions in the same way that we question other things. We are not stuck with them because we can change our mind about them based on this deliberation. However, this deliberation had better use our best ideas about how to validate or reject ideas (which is, in a sense, circular when what we are analyzing is our best ideas about how to validate or reject ideas). Quoting from Eliezer:
So what I did in practice, does not amount to declaring a sudden halt to questioning and justification. I’m not halting the chain of examination at the point that I encounter Occam’s Razor, or my brain, or some other unquestionable. The chain of examination continues—but it continues, unavoidably, using my current brain and my current grasp on reasoning techniques. What else could I possibly use?
And later in the essay:
If one of your current principles does come up wanting—according to your own mind’s examination, since you can’t step outside yourself—then change it! And then go back and look at things again, using your new improved principles.
See, he’s not saying they are actually valid, and he’s not saying we’re stuck with them. He’s just recommending using your best understanding in the moment, to move forward.
however bad they are it’s not particularism, and not a solution to epistemology.
I’m also not sure what “a solution to epistemology” means.
Quoting from Eliezer:
But this doesn’t answer the legitimate philosophical dilemma: If every belief must be justified, and those justifications in turn must be justified, then how is the infinite recursion terminated?
So I think it’s fair to say that Eliezer is trying to answer the regress argument. That’s the specific question he is trying to answer. Observing that his position is “not a solution to epistemology” does little to recommend against it.
He also offers a kind of circular defense of induction, which I don’t think amounts to full fledged circularity,because you need some empirical data to kick things off.
I think another subtlety which might differentiate it from what’s ordinarily called circular reasoning is a use/mention distinction. He’s recommending using your best reasoning principles, not assuming them like axioms. This is what I had in mind at the beginning when I said that his essay didn’t hand the reader a procedure to distinguish his recommendation from the bad kind of circular reasoning—a use/mention distinction seems like a plausible analysis of the difference, but he at least doesn’t emphasize it. Instead, he seems to analyze the situation as this-particular-case-of-circular-logic-works-fine:
So, at the end of the day, what happens when someone keeps asking me “Why do you believe what you believe?”
At present, I start going around in a loop at the point where I explain, “I predict the future as though it will resemble the past on the simplest and most stable level of organization I can identify, because previously, this rule has usually worked to generate good results; and using the simple assumption of a simple universe, I can see why it generates good results; and I can even see how my brain might have evolved to be able to observe the universe with some degree of accuracy, if my observations are correct.”
But then… haven’t I just licensed circular logic?
Actually, I’ve just licensed reflecting on your mind’s degree of trustworthiness, using your current mind as opposed to something else.
I take this position to be accurate so far as it goes, but somewhat lacking in providing a firm detector for bad vs good circular logic. Eliezer is clearly aware of this:
I do think that reflective loops have a meta-character which should enable one to distinguish them, by common sense, from circular logics. But anyone seriously considering a circular logic in the first place, is probably out to lunch in matters of rationality; and will simply insist that their circular logic is a “reflective loop” even if it consists of a single scrap of paper saying “Trust me”. Well, you can’t always optimize your rationality techniques according to the sole consideration of preventing those bent on self-destruction from abusing them.
“Usually” being the key here. To me, the most interesting coherence theories are broadly bayesian in character.
Bayesianism is even more explicit about the need for compatibility with existing beliefs, I’m priors.
I’m not sure what position you’re trying to take or what argument you’re trying to make here
I don’t think there is a single theory that achieves every desideratum (including minimality of unjustified assumptions). Ie. Epistemology is currently unsolved.
do you think there’s a correct theory which does have the property of convergence on a single truth?
I think convergence is a desideratum.
I don’t know of a theory that achieves all desiderata. Theres any number of trivial theories that can converge, but do nothing else.
I don’t think it’s possible to converge on the truth in all cases, since information is limited—EG, we can’t decide all undecidable mathematical statements, even with infinite time to think. Because this is a fundamental limit, though, it doesn’t seem like a viable charge against an individual theory.
That’s not the core problem: there are reasons to believe that convergence can’t be achieved, even if everyone has access to the same finite pool of information. The problem of the criterion is one of them… if there is fundamental disagreement about the nature of truth and evidence, then agents that fundamentally differ won’t converge in finite time.
“Usually” being the key here. To me, the most interesting coherence theories are broadly bayesian in character.
Bayesianism is even more explicit about the need for compatibility with existing beliefs, ie. priors.
I’m not sure what position you’re trying to take or what argument you’re trying to make here
I don’t think there is a single theory that achieves every desideratum (including minimality of unjustified assumptions). Ie. Epistemology is currently unsolved.
do you think there’s a correct theory which does have the property of convergence on a single truth?
I think convergence is a desideratum.
I don’t know of a theory that achieves all desiderata. Theres any number of trivial theories that can converge, but do nothing else. There’s lots of partial theories as well, but it’s not clear how to make the tradeoffs.
I don’t think it’s possible to converge on the truth in all cases, since information is limited—EG, we can’t decide all undecidable mathematical statements, even with infinite time to think. Because this is a fundamental limit, though, it doesn’t seem like a viable charge against an individual theory.
That’s not the core problem: there are reasons to believe that convergence can’t be achieved, even if everyone has access to the same finite pool of information. The problem of the criterion is one of them… if there is fundamental disagreement about the nature of truth and evidence, then agents that fundamentally differ won’t converge in finite time.
We are not stuck with them because we can change our mind about them based on this deliberation. However, this deliberation had better use our best ideas about how to validate or reject ideas (which is, in a sense, circular when what we are analyzing is our best ideas about how to validate or reject ideas
Yes...the circularity of the method weighs against convergence in the outcome.
He is saying that we can question our rock bottom assumptions in the same way that we question other things. We are not stuck with them because we can change our mind about them based on this deliberation. However, this deliberation had better use our best ideas about how to validate or reject ideas
Out best ideas relatively might not be good enough absolutely. In that passage he is sounding like a Popperism, but the Popperism approach is particularly unable to achieve convergence.
“What else could I possibly use?”
Doesn’t imply that what he must use is any good in absolute terms.
I don’t think there is a single theory that achieves every desideratum (including minimality of unjustified assumptions). Ie. Epistemology is currently unsolved.
I was never arguing against this. I broadly agree. However, I also think it’s a poor problem frame, because “solve epistemology” is quite vague. It seems better to be at least somewhat more precise about what problems one is trying to solve.
Out best ideas relatively might not be good enough absolutely. In that passage he is sounding like a Popperism, but the Popperism approach is particularly unable to achieve convergence.
Well, just because he says some of the things that Popper would say, doesn’t mean he says all of the things that Popper would say.
do you think there’s a correct theory which does have the property of convergence on a single truth?
I think convergence is a desideratum. I don’t know of a theory that achieves all desiderata. Theres any number of trivial theories that can converge, but do nothing else. There’s lots of partial theories as well, but it’s not clear how to make the tradeoffs.
I don’t think it’s possible to converge on the truth in all cases, since information is limited—EG, we can’t decide all undecidable mathematical statements, even with infinite time to think. Because this is a fundamental limit, though, it doesn’t seem like a viable charge against an individual theory.
That’s not the core problem: there are reasons to believe that convergence can’t be achieved, even if everyone has access to the same finite pool of information. The problem of the criterion is one of them… if there is fundamental disagreement about the nature of truth and evidence, then agents that fundamentally differ won’t converge in finite time.
If you’ve got an argument that a desideratum can’t be achieved, don’t you want to take a step back and think about what’s achievable? In the quoted section above, it seems like I offer one argument that convergence isn’t achievable, and you pile on more, but you still stick to a position something like, we should throw out theories that don’t achieve it?
Bayesianism is even more explicit about the need for compatibility with existing beliefs, ie. priors.
That’s why I’m using it as an example of coherentism.
IDK what the problem is here, but it seems to me like there’s some really weird disconnect happening in this conversation, which keeps coming back in full force despite our respective attempts to clarify things to each other.
I was never arguing against this. I broadly agree.
But you also said that:-
Also note that I prefer to claim Eliezer’s view is essentially correct
Correct about what? That he has solved epistemology, or that epistemology is unsolved, or what to do in the absence of a solution? Remember , the standard rationalist claim is that epistemology is solved by Bayes. That’s the claim that people like Gordon and David Chapman are arguing against. If you say you agree with Yudkowsky, that is what people are going to assume you mean.
However, I also think it’s a poor problem frame, because “solve epistemology” is quite vague.
I just told you told you what that means “a single theory that achieves every desideratum (including minimality of unjustified assumptions)”
Also note that I prefer to claim Eliezer’s view is essentially correct without claiming that it solves the problem of the criterion, since I find the criterion problem to be worryingly vague
So Yudkowsky is essentially correct about ….something… but not necessarily about the thing this discussion is about.
Well, just because he says some of the things that Popper would say, doesn’t mean he says all of the things that Popper would say.
He says different things in different places, as I said. He says different things in different places, so hes unclear.
If you’ve got an argument that a desideratum can’t be achieved, don’t you want to take a step back and think about what’s achievable
I don’t think all desiderata are achievable by one theory. That’s my precise reason for thinking that epistemology is unsolved.
but you still stick to a position something like, we should throw out theories that don’t achieve it?
I didn’t say that. I haven’t even got into the subject of what to do given the failure of epistemology to meet all its objectives.
What I was talking about specifically was the inability of Bayes to achieve convergence. You seem to disagree, because you were talking about “agreement Bayes”.
Also note that I prefer to claim Eliezer’s view is essentially correct
Correct about what? That he has solved epistemology, or that epistemology is unsolved, or what to do in the absence of a solution? Remember , the standard rationalist claim is that epistemology is solved by Bayes. That’s the claim that people like Gordon and David Chapman are arguing against. If you say you agree with Yudkowsky, that is what people are going to assume you mean.
I already addressed this in a previous comment:
I’m also not sure what “a solution to epistemology” means.
Quoting from Eliezer:
But this doesn’t answer the legitimate philosophical dilemma: If every belief must be justified, and those justifications in turn must be justified, then how is the infinite recursion terminated?
So I think it’s fair to say that Eliezer is trying to answer the regress argument.
I am trying to be specific about my claims. I feel like you are trying to pin very general and indefensible claims on me. I am not endorsing everything Eliezer says; I am endorsing the specific essay.
However, I also think it’s a poor problem frame, because “solve epistemology” is quite vague.
I just told you told you what that means “a single theory that achieves every desideratum (including minimality of unjustified assumptions)”
Is there some full list of desiderata which has broad agreement or which you otherwise want to defend??
I feel like any good framing of “solve epistemology” should start by being more precise than “solve epistemology”, wrt what problem or problems are being approached.
but you still stick to a position something like, we should throw out theories that don’t achieve it?
I didn’t say that. I haven’t even got into the subject of what to do given the failure of epistemology to meet all its objectives.
Agreement that you didn’t say that. It’s merely my best attempt at interpretation. I was trying to ask a question about what you were trying to say. It seems to me like one thing you have been trying to do in this conversation is dismiss coherentism as a possible answer, on the argument that it doesn’t satisfy some specific criteria, in particular truth-convergence. I’m arguing that truth-convergence should clearly be thrown out as a criterion because it’s impossible to satisfy (although it can be revised into more feasible criteria). On my understanding, you yourself seemed to agree about its infeasibility, although you seemed to think we should focus on different arguments about why it is infeasible (you said that my argument misses the point). But (on my reading) you continued to seem to reject coherentism on the same argument, namely the truth-convergence problem. So I am confused about your view.
I don’t think all desiderata are achievable by one theory. That’s my precise reason for thinking that epistemology is unsolved.
To reiterate: then why would they continue to be enshrined as The Desiderata?
So Yudkowsky is essentially correct about ….something… but not necessarily about the thing this discussion is about.
I mean, looking back, it was me who initially agreed with the OP [original post], in a top-level comment, that Eliezer’s essay was essentially correct. You questioned that, and I have been defending my position… and now it’s somehow off-topic?
Correct about what? That he has solved epistemology, or that epistemology is unsolved, or what to do in the absence of a solution?
I already addressed this in a previous comment
I don’t see where.
Is there some full list of desiderata which has broad agreement or which you otherwise want to defend??
Certainty. A huge issue in early modern philosophy which has now been largely abandoned.
Completeness. Everything is either true or false, nothing is neither.
Consistency. Nothing is both true and false.
Convergence. Everyone can agree.
Objectivity: Everyone can agree on something that’s actually true
It seems to me like one thing you have been trying to do in this conversation is dismiss coherentism as a possible
I dismissed it as an answer that fulfills all criteria, because it doesn’t fulfil Convergence. I didn’t use the word possible—you did in another context, but I couldn’t see what you meant. If nothing fulfils all criteria, then coherentism could be preferable to approaches with other flaws.
To reiterate: then why would they continue to be enshrined as The Desiderata?
Because all of them individually can be achieved if you make trade offs.
Because things don’t stop being desireable when they are unavailable.
I’m arguing that truth-convergence should clearly be thrown out as a criterion because it’s impossible to satisfy
That isn’t at all clear.
Lowering the bar to whatever you can jump over ,AKA Texas sharpshooting,isn’t a particularly rational procedure. The absolute standard you are or are not hitting relates to what you can consistently claim on the object level: without the possibility convergence on objective truth, you can’t claim that people with other beliefs are irrational or wrong, (at least so long as they hit some targets).
It’s still important to make relative comparisons even if you can’t hit the absolute standard...but it’s also important to remember your relatively best theory is falling short of absolute standards.
Completeness: I don’t need everything to be either true or false, I just need everything to have a justifiable probability.
That’s also Completeness.
Convergence+ObjectivityConvergence
Giving up on convergence has practical consequences. To be Consistent, you need to give up on the stance that your views are a slam dunk, and the other tribe’s views are indefensible.
I wouldn’t guess that I have a big disagreement with you once we translate everything into each other’s preferred way of speaking. I think you’re pointing at something important and right. But I sure won’t habitually call it “the problem of the criterion”, or use any of the explanatory devices presented here.
So, let’s see what you think of my take.
I think “The Problem of the Criterion” as framed here (and, I guess, most places) creates a heavy temptation to conflate several different issues.
The problem of responding to the skeptic. The way the problem of the criterion is phrased uses “knowledge”, which tempts one to think of perfect knowledge. This makes the problem seem acute, because plausibly, nothing can be known with certainty. But this makes it sound as if the alternative to solving the problem of the criterion is falling into complete skepticism. For Bayesians (and some other epistemologies), it would be fine if nothing is known with certainty, and indeed, plausibly nothing is known with certainty. This makes the problem of the criterion sound irrelevant, for these camps. And it is, provided we understand the problem of the criterion to be something like “how can we know anything with certainty?”
The epistemic circularity of foundationalism. This is the recursion problem you mention. This is only a problem if you think each belief should have a justification, and these justifications should never form a circle, and should also avoid forming an infinite chain. But there is no mathematical structure which contains no circles, no infinite chains, and also no unjustified orphan truths! This is intuitively a real problem, since we can separately justify intuitions that each belief should be justified, that circular justifications don’t count, and that infinite backward chains of justification also don’t count. I think Eliezer’s answer to this conundrum is correct (the one which you classify as pragmatism), although I think there are some details to work out. (IE, I don’t think most people walk away from that essay with a working classifier to distinguish between bad circular arguments and acceptable ones, because it doesn’t give one; it focuses on making the case that specific instances of apparently circular reasoning are fine and good.)
The naturalistic question, where does knowledge come from? There’s a phenomenon in the world: sometimes people (and animals) seem to know things (and, sometimes not). How does this happen? How can we accurately describe and analyze the phenomenon? Again, Bayesians will tend to think they have a good answer. But this time the Bayesians will be roused to answer with vim and vigor; for this is an issue of key interest to them. Whereas problem #1 seems like a silly distraction to them.
The normative question: what should we consider ourselves to know, and how should we manage belief? This is the core question of epistemic rationality.
You also mention several other distinct problems which “are just the problem of the criterion in disguise”:
You seem to find it encouraging that the Problem of the Criterion mixes all of these problems together in one big pot. I think it’s quite problematic. In particular, it means that people who think they have solved one problem here will think they’ve solved the others. It also tempts impossibility proofs for solving one problem to be applied to the others.
I think you start out discussing the normative problem:
This how do you know sounds like a justification question, rather than an empirical how-did-you-come-to-know; although it’s impossible to tell for sure without specific disambiguation (this is part of the problem).
But then you mix in the recursion problem. At the beginning of the section “The Problem”, your characterization of The Problem is as follows:
So which problem is it?
If it’s the literal, empirical content of the first question asked, how do you know something is true, then we investigate by trying to remember how we know specific things, noticing when we learn new things, and inventing theoretical frameworks (such as positivism, or fallibilism, or probability theory, or neuroscience) which attempt to characterize the process. The recursive issue isn’t so much a problem as it is more data—as we arrive at tentative conclusions about how we come to know things, we can check whether our theories properly account for how we would learn those theories themselves.[1]
If it’s the normative version of the first question, how should we know things (or perhaps, how can we justify knowledge), then the recursion problem is perhaps an argument for why the question is difficult—but not the question itself. On this reading, any answer to the criterion problem must deal with the recursion problem, but dealing with the recursion problem is not in itself sufficient; we might be able to dismiss the recursion problem while lacking any reply to the full problem.
If it’s the recursive problem itself, then the criterion problem is really only a problem for foundationalists; perhaps it should be remembered as the main problem with foundationalism.
If it’s the problem of how we come to know anything (know with certainty), then the skeptical response seems totally reasonable and indeed correct. In this case the recursion problem sounds like just one way to argue for the skeptical conclusion.[2]
EDIT
It looks like, according to Wikipedia,
So it would seem that the regress argument is kept distinct from the problem of the criterion, leaving my other three interpretations on the table.
However, the Internet Encyclopedia of Philosophy begins its article like this:
Which, indeed, sounds like the regress argument.
And a bit later in the article, a severe ambiguity is confirmed:
This means the recursion problem does have an interesting role to play in the empirical inquiry. The recursion becomes an interesting test case. Perhaps it becomes a big obstacle to some theories. For example, I’ve heard the accusation that Popperian falsifiability theory isn’t itself falsifiable. But perhaps this is simply because it’s better viewed as a normative theory of science, rather than an empirical one.
The empirical-recursive problem of the criterion for Bayesian behavioral scientists (aka economists) is the question of whether we can empirically justify bayesian rationality as a behavioral theory (aka, behavioral economics).
The empirical-recursive problem of the criterion for neuroscience would be whether a particular neuroscientific theory can account for neuroscience as an activity of the brain. (EG, some connectionist theories which require pre-existing concept cells for every concept will fall short, because a pre-existing ‘connectionism cell’ is arguably too implausible.)
In this case, the recursion problem can be properly highlighted by appealing to Godel’s incompleteness theorem, since Godel’s theorem does apply to this kind of crisp 100% knowledge (provided we think it meets the other conditions for the theorem).
However, the kind of skeptic I’m endorsing (a moderate bayesian skeptic, who believes in probabilistic knowledge) has no business making the recursion-problem argument, since they don’t accept the foundationalist underpinnings of such an argument.
But perhaps that’s fine; there’s a long tradition in skepticism of referring to arguments as ladders which you use to climb up to the skeptical vantage point, but then kick down, having no further use for them.
After all, the bayesian moderate-skeptic is simply trying to argue that logical certainty doesn’t exist in practice. So long as they’re arguing about absurdities, they might as well also assume the absurd foundationalist assumptions.
But it’s a bit rude of them if they try to sneak the assumptions in without explicitly flagging them.
First, this is a great, really thoughtful comment.
My initial thought is that I’m doing a bad job in this post of pointing at the thing that really matters, which is why I think we may see different framings as useful. I see the problem of the criterion as a particular instance of exposing a deeper, fundamental meta-fact about our world, which is that uncertainty is fundamental to our existence and all knowledge is teleological. I don’t think all that comes through here because when I wrote this it was my first time really trying to explain fundamental uncertainty, so I only go so far in explaining it. I got a bit too excited a confused the problem of the criterion for fundamental uncertainty itself.
That (1) seems boring to Bayesians et al. seems fine to me because they already buy that things are fundamentally uncertain, although I think in practice most aspiring Bayesians don’t really know why, which bleeds into (2). I see (2) as important because knowledge does, in fact, ground out in some foundation, and we try hard to make sure that grounding is correct by requiring justifications for our grounding. Personally, the question I’m most interested in is (3), and then considering what implications that has for (1), (2), and (4), though I don’t really have much to say about (4) here other than that one should be fundamentally uncertain and that I think in practice many people trying to work out (4) get confused because they imagine they can achieve justified true belief, or more often with aspiring Bayesians justified true meta-belief even though all object-level beliefs are uncertain.
So my I have to admit my motivations when writing this post were a bit of a mix of yelling at rationalists I meet who are performing rationality but missing this deeper thing that I was also missing out on for a long time and trying to really clarify what this deeper thing is that’s going on.
I know you’re reading some of my other posts, but let me see if I can explain how I actually think about the problem of the criterion today:
We and every dynamic process can be thought of as made up of lots of feedback circuits, and most of those circuits are negative feedback circuits (positive feedback, where it exists, is often wrapped inside a negative feedback circuit because otherwise the universe would be boring and already tiled with something like paperclips). These circuits get information by observing the world and then generating a signal to control the circuit. The process of generating the signal is necessarily lossy, and it’s because we’re going to compress information when generating the signal. As a result, the circuit makes a “choice” about what to signal on, the signal is thus teleological because the signal depends on what purpose the signal has (purpose is an idea in cybernetics I won’t expand here, but basically it’s what the circuit is designed to do), so all the knowledge (information inside the circuits) we have is both uncertain due to compression (and maybe something something quantum mechanics observation, but I don’t think we need to go there) and teleological because circuits impose teleology through their design.
I’ll admit, this is all sort of a weird way to explain it because it hinges on this unusual cybernetic model of the world, so I try to explain myself in other ways that are more familiar to folks. Maybe that’s the wrong approach, but I think it’s reasonable to come at this thing outside in (which is what I do in my book) by addressing the surface level uncertainty first and then peeling back the layers to find the fundamental uncertainty. This post is weird because it’s a bit in the middle.
(I’ll also note there’s an even deeper version of how I think about fundamental uncertainty and teleology today, but I’m not sure I can put it in words because it comes from meditation experience and I can only talk about it in mysterious ways. I think trying to talk about that way of thinking about it is probably more confusing than helpful.)
IMHO the cybernetic picture isn’t weird or uncommon; naively, I expect it to get less pushback.
I think this is what I most want to push back on. My own sense is that you are confused about this. On my understanding, you seem to simultaneously believe that the core foundationalist assumptions make sense, and also believe an impossibility argument which shows them to be inconsistent. This doesn’t make sense to me.
My formalization here is only one possible way to understand the infinite-regress problem (although I think it does a good job of capturing the essence of it) -- but, in this formalization, the contradiction is really direct, which makes it seem pretty silly.
I also think the contradictory axioms do capture two intuitions which, like, beginning philosophy majors might endorse.
So I think the infinite regress problem should be explained to beginning philosophers as a warning against these naive assumptions about justification. (And indeed, this is how I was taught.)
But that’s what it is to me. It seems to be something else for you. Like a paradox. You write of proving the impossibility of solution, rather than resolving the problem. You write that we should “hold the problem”. Like, in some sense it is still a problem even after it has been solved.
(Perhaps the seeming contradiction is merely due to the way the criterion problem conflates multiple problems; EG, the naturalistic question of where knowledge comes from is still a live question after the justification-infinite-regress problem has been resolved.)
It makes sense to me to try to spell out the consequences of the infinite-regress problem; they may be complex and non-obvious. But for me this should involve questioning the naive assumptions of justification, and figuring out what it is they were trying to do. On my analysis, a reasonable place to go from there is the tiling agents problem and Vingean reflection. This is a more sophisticated picture of the problems rational agents run into when philosophizing about themselves, because it admits that you don’t need to already know how you know in order to know—you are already whatever sort of machine you are, and the default outcome is that you keep running as you’ve run. You’re not at risk of all your knowledge evaporating if you can’t justify it. However, there is a big problem of self-trust, and how you can achieve things over time by cooperating with yourself. And there’s an even bigger problem if you do have the ability to self-modify; then your knowledge actually might evaporate if you can’t justify it.
But this problem is to a large degree solved by the lack of a probabilistic Lob’s theorem. This means we are in a much better situation with respect to self-trust, vingean reflection, and the tiling agents problem—so long as we can accept probabilistic fallibility rather than needing perfect self-trust.
So in effect, the answer to (a newer and significantly more sophisticated version of) the recursive justification problem is the same as the answer to skepticism which you agreed was right and proper—namely, embracing probabilism rather than requiring certainty.
I expect you’ll agree with this conclusion; the part where I think we have some sort of fuzzy disagreement is the part where you still say things like “I see (2) as important because knowledge does, in fact, ground out in some sort of foundation, and we try to make sure that grounding is correct by requiring justifications for our grounding.” I think there are some different possible ways to spell out what you meant here, but I don’t think the best ones connect with (2) very much, since (2) is based on some silly-in-hindsight intuitions about justification.
Hmm, I’ll have to think about this.
So I think maybe there’s something goin on here where I’m taking too much for granted that formal systems are the best way to figure out what to believe about the world to get an accurate picture of it, and so long as you’re figuring out what to believe using some formal system then you’re forced to ground it in some assumed foundation that is not itself justified. This seems important because it means there’s a lot of interesting stuff going on with those assumptions and how they get chosen such that they cause the rest of the system to produce beliefs that track reality.
But it sounds like you think this is a confused view. I admit I feel a bit differently about things than the way they’re presented in the post, but not entirely, as my comment reflects. I think when I wrote the post I really couldn’t see my way around how to know things without being able to formally justify them. Now I better understand there can be knowing without belief that knowledge is justified or true, but I do think there’s something to the idea that I am still too much captured by my own ontology and not fully seeing it as I see through it, and this causes me to get a bit mixed up in places.
The bit about self-trust is really interesting. To get a little personal and psychological, I think the big difference between me ~4 years ago and me now is that ~4 years ago I learned to trust myself in some big important way that goes deeper than surface level self-trust. On the surface this looks like trusting my feelings and experiences, but to a certain extent anyone can do this if you just give them feedback that says trust those things more than other things. But I’m talking about something a level deeper where I trust the process by which those things get generated to do what they do and that they’re doing their best (in the sense that the universe is deterministic and we’re all always doing our best/worst because it could have been no other way).
If I’m honest I’ve really struggled to understand what’s going on with Lob’s theorem and why it seems like a big deal to everyone. But it seems like from what you’re saying it’s related to the issue that got me started on things, which is how do you deal with the problem of verifying that another agent (which could be yourself at another time) shares your values and is thus in some way aligned with you. If that’s the case maybe I already grasp the fundamentals of this obstacle Lob seems to create and have just failed to realize that’s what folks were talking about when they talk about Lob with respect to aligning AI.
@abramdemski Wanted to say thanks again for engaging with my posts and pointing me towards looking again at Lob. It’s weird: now that I’ve taken so time to understand it, it’s just what in my mind was already the thing going on with Godel, just I wasn’t doing a great job of separating out what Godel proves and what the implications are. As presented on its own, Lob didn’t seem that interesting to me so I kept bouncing off it as something worth looking at, but now I realize it’s just the same thing I learned from GEB’s presentation of Peano arithmetic and Godel when I read it 20+ years ago.
When I go back to make revisions to the book, I’ll have to reconsider including Godel and Lob somehow in the text. I didn’t because I felt like it was a bit complicated and I didn’t really need to dig into it since I think there’s already a bit too many cases where people use Godel to overreach and draw conclusions that aren’t true, but it’s another way to explain these ideas. I just have to think about if Godel and Lob are necessary: that is, do I need to appeal to them to make my key points, or are these things that are better left as additional topics I can point folks at but not key to understanding the intuitions I want them to develop.
I’ve heard Lob remarked that he would never have published if he realized earlier how close his theorem was to just Godel’s second incompleteness theorem; but I can’t seem to entirely agree with Lob there. It does seem like a valuable statement of its own.
I agree, Godel is dangerously over-used, so the key question is whether it’s necessary here. Other formal analogs of your point include Tarski’s undefinability, and the realizablility / grain-of-truth problem. There are many ways to gesture towards a sense of “fundamental uncertainty”, so the question is: what statement of the thing do you want to make most central, and how do you want to argue/illustrate that statement?
I think pragmatism has the problems that the OP points out: it’s necessarily a partial solution , because not everything has a clear winning condition.
Embracing circularity is different from pragmatism. Indeed, without the classifier you mention , it’s completely impractical!
This is why I said “the one which you classify as pragmatism”—Gordon Worley calls it pragmatist because Eliezer’s notion of rationality is ultimately about winning; but that is quite far from being a good summary of Eliezer’s specific argument in Where Recursive Justification Hits Bottom.
I don’t think Eliezer’s specific argument has this problem, since his advice has little to do with winning conditions. His advice is that when considering very abstract questions such as the foundations of epistemology, we have to consider these questions in the same way we consider any other questions, IE using the full strength of our epistemology.
I don’t think there is just one definition of rationality, or that EY has one solution to epistemology.
Winning based rationality doesn’t have to coincide with truth seeking rationality, and there are robust cases where it doesn’t.
The rock bottom argument in no way justifies our epistemic practices, it just says we might as well stick to them.
Using winning based rationality alone entails giving up on every problem that isn’t susceptible to it, including (a)theism, MWI , etc.
Embracing circularity is different to both of the above, and no answer to the the problem of the criterion, since coherence is a criterion.
I think there are a bunch of related problems being labeled as the criterion problem.
One of them is the question: what is the criterion?
The possible answers to this question are (of course) proposed criteria.
For this version of the criterion problem, your objection seems absurd. Of course a proposed answer to the criterion problem is going to be a criterion! That doesn’t disqualify an answer! Rather, that is a necessary characteristic of an answer! We can’t disqualify proposed criteria simply for being criteria!
Another one of them is the infinite-recursion problem Gordon mentions (also known as the regress argument). This argument rests on the foundationalist intuitions that all beliefs need to be justified, and infinite chains of justification (such as circular justification) don’t count.
Coherence seems to also qualify as a possible answer to this version of the criterion problem, addressing the problem by pointing to which of the contradictory assumptions we should give up.
I think rejecting these assumptions is a legitimate sort of answer to the problem, since the criterion problem isn’t just supposed to be a problem within foundationalist theories—it’s not like rejecting those assumptions misses the point of the problem. Rather, the regress argument is an argument made to all thinking beings, for whom the foundationalist assumptions are tacitly supposed to be compelling. So pointing out which assumptions are not, in fact, compelling seems like a good type of reply.
So overall, I don’t buy your argument that coherence is “no answer” to the criterion problem. It seems like a reasonable sort of answer to pursue, for at least two of the central forms of the criterion problem.
Another is: does it work? Another is : does anything work?
Circularity isn’t just a criterion, it’s one of the unholy trinity of criteria …particularism, regress and Circularity .. that are know to have problems. The problem with Circularity is that it justifies too much, not too little.
Coherence is either circularity, in which case it has the same problems; or something else, in which case it needs to be spelt out.
I don’t think EY has spelt out a fourth criterion that works. Where Regress Stops is particularism.
What do you mean by coherence?
What do you mean by coherence?
My take on “justifies too much” is that coherentism shouldn’t just say coherence is the criterion, in the sense of anything coherent is allowed to fly. Instead, the (subjectively) correct approach is (something along the lines of) coherence with the rest of my beliefs.
Coherence is one of the six response types cited in the OP. I was also using ‘circularity’ and ‘coherence’ somewhat synonymously. Sorry for the confusion.
When I took undergrad philosophy courses, “coherentism” and “foundationalism” were two frequently-used categories representing opposite approaches to justifiation/knowledge. Foundationalism is the idea that everything should be built up from some kind of foundation (so here we could classify it as particularism, probably). Coherentism is basically the negation of that, but is generally assumed to replace building-up-from-foundations with some notion of coherence. Neither approach is very well-specified; they’re supposed to be general clusters of approaches, as opposed to full views.
More specifically, with respect to my previous statement:
Here, we can interpret coherentism specifically as the acceptance of circular logic as valid (the inference “A, therefore A” is a valid one, for example—note that ‘validity’ is about the argument from premise to conclusion, and does not imply accepting the conclusion). This resolves the infinite regress problem by accepting regress, whereas particularism instead rejects the assumption that every statement needs a justification.
I think the details of these views matter a whole lot. Which kind of particularism? How is it justified? Does the motivation make sense? Or which coherentism, etc.
The arguments about the Big Three are vague and weak due to the lack of detail. Chisholm makes his attacks on each approach, which sound convincing as-stated, but which seem much less relevant when you’re dealing with a fully-fleshed-out view with specific arguments in its favor.
(But note that for Chisholm, the three are: particularism, methodism, skepticism. So his categorization is different from yours.)
You classify Eliezer as particularism, I see it as a type of circular/coherentist view, Gordon Worley sees need to invent a “pragmatism” category. The real answer is that these categories are vague and Eliezer’s view is what it is, regardless of how you label it.
Also note that I prefer to claim Eliezer’s view is essentially correct without claiming that it solves the problem of the criterion, since I find the criterion problem to be worryingly vague.
That’s how coherence usually works. If no new belief is accepted unless consistent with existing beliefs, then you don’t get as much quodlibet as under pure circular justification, but you don’t get convergence on a single truth either.
Having re-read Where Recursive Justfication Hits Bottom, I think EY says different things in different places. If he is saying that our Rock Bottom assumptions are actually valid, that’s particularism. If he is saying that we are stuck with them, however bad they are it’s not particularism, and not a solution to epistemology.
He also offers a kind of circular defense of induction, which I don’t think amounts to full fledged circularity,because you need some empirical data to kick things off. The Circular Justifcation of Induction isn’t entirely bad, but it’s yet another partial solution, because it’s limited to the empirical and quantifiable. If you explicitly limit everyone to the empirical and quantifiable, that’s logical.positivism, but EY say he’s isn’t a logical positivist.
What are his views, and what is it correct about? Having read them, I still dont know.
It matters both ways. The solution needs to be clear, if there is one.
“Usually” being the key here. To me, the most interesting coherence theories are broadly bayesian in character.
I’m not sure what position you’re trying to take or what argument you’re trying to make here—do you think there’s a correct theory which does have the property of convergence on a single truth? Do you think convergence on a single truth is a critical feature of a successful theory?
I don’t think it’s possible to converge on the truth in all cases, since information is limited—EG, we can’t decide all undecidable mathematical statements, even with infinite time to think. Because this is a fundamental limit, though, it doesn’t seem like a viable charge against an individual theory.
I don’t think he’s saying either of those things exactly. He is saying that we can question our rock bottom assumptions in the same way that we question other things. We are not stuck with them because we can change our mind about them based on this deliberation. However, this deliberation had better use our best ideas about how to validate or reject ideas (which is, in a sense, circular when what we are analyzing is our best ideas about how to validate or reject ideas). Quoting from Eliezer:
And later in the essay:
See, he’s not saying they are actually valid, and he’s not saying we’re stuck with them. He’s just recommending using your best understanding in the moment, to move forward.
I’m also not sure what “a solution to epistemology” means.
Quoting from Eliezer:
So I think it’s fair to say that Eliezer is trying to answer the regress argument. That’s the specific question he is trying to answer. Observing that his position is “not a solution to epistemology” does little to recommend against it.
I think another subtlety which might differentiate it from what’s ordinarily called circular reasoning is a use/mention distinction. He’s recommending using your best reasoning principles, not assuming them like axioms. This is what I had in mind at the beginning when I said that his essay didn’t hand the reader a procedure to distinguish his recommendation from the bad kind of circular reasoning—a use/mention distinction seems like a plausible analysis of the difference, but he at least doesn’t emphasize it. Instead, he seems to analyze the situation as this-particular-case-of-circular-logic-works-fine:
I take this position to be accurate so far as it goes, but somewhat lacking in providing a firm detector for bad vs good circular logic. Eliezer is clearly aware of this:
Bayesianism is even more explicit about the need for compatibility with existing beliefs, I’m priors.
I don’t think there is a single theory that achieves every desideratum (including minimality of unjustified assumptions). Ie. Epistemology is currently unsolved.
I think convergence is a desideratum. I don’t know of a theory that achieves all desiderata. Theres any number of trivial theories that can converge, but do nothing else.
That’s not the core problem: there are reasons to believe that convergence can’t be achieved, even if everyone has access to the same finite pool of information. The problem of the criterion is one of them… if there is fundamental disagreement about the nature of truth and evidence, then agents that fundamentally differ won’t converge in finite time.
Bayesianism is even more explicit about the need for compatibility with existing beliefs, ie. priors.
I don’t think there is a single theory that achieves every desideratum (including minimality of unjustified assumptions). Ie. Epistemology is currently unsolved.
I think convergence is a desideratum. I don’t know of a theory that achieves all desiderata. Theres any number of trivial theories that can converge, but do nothing else. There’s lots of partial theories as well, but it’s not clear how to make the tradeoffs.
That’s not the core problem: there are reasons to believe that convergence can’t be achieved, even if everyone has access to the same finite pool of information. The problem of the criterion is one of them… if there is fundamental disagreement about the nature of truth and evidence, then agents that fundamentally differ won’t converge in finite time.
Yes...the circularity of the method weighs against convergence in the outcome.
Out best ideas relatively might not be good enough absolutely. In that passage he is sounding like a Popperism, but the Popperism approach is particularly unable to achieve convergence.
Doesn’t imply that what he must use is any good in absolute terms.
I was never arguing against this. I broadly agree. However, I also think it’s a poor problem frame, because “solve epistemology” is quite vague. It seems better to be at least somewhat more precise about what problems one is trying to solve.
Well, just because he says some of the things that Popper would say, doesn’t mean he says all of the things that Popper would say.
If you’ve got an argument that a desideratum can’t be achieved, don’t you want to take a step back and think about what’s achievable? In the quoted section above, it seems like I offer one argument that convergence isn’t achievable, and you pile on more, but you still stick to a position something like, we should throw out theories that don’t achieve it?
That’s why I’m using it as an example of coherentism.
IDK what the problem is here, but it seems to me like there’s some really weird disconnect happening in this conversation, which keeps coming back in full force despite our respective attempts to clarify things to each other.
But you also said that:-
Correct about what? That he has solved epistemology, or that epistemology is unsolved, or what to do in the absence of a solution? Remember , the standard rationalist claim is that epistemology is solved by Bayes. That’s the claim that people like Gordon and David Chapman are arguing against. If you say you agree with Yudkowsky, that is what people are going to assume you mean.
I just told you told you what that means “a single theory that achieves every desideratum (including minimality of unjustified assumptions)”
So Yudkowsky is essentially correct about ….something… but not necessarily about the thing this discussion is about.
He says different things in different places, as I said. He says different things in different places, so hes unclear.
I don’t think all desiderata are achievable by one theory. That’s my precise reason for thinking that epistemology is unsolved.
I didn’t say that. I haven’t even got into the subject of what to do given the failure of epistemology to meet all its objectives.
What I was talking about specifically was the inability of Bayes to achieve convergence. You seem to disagree, because you were talking about “agreement Bayes”.
I already addressed this in a previous comment:
I am trying to be specific about my claims. I feel like you are trying to pin very general and indefensible claims on me. I am not endorsing everything Eliezer says; I am endorsing the specific essay.
Is there some full list of desiderata which has broad agreement or which you otherwise want to defend??
I feel like any good framing of “solve epistemology” should start by being more precise than “solve epistemology”, wrt what problem or problems are being approached.
Agreement that you didn’t say that. It’s merely my best attempt at interpretation. I was trying to ask a question about what you were trying to say. It seems to me like one thing you have been trying to do in this conversation is dismiss coherentism as a possible answer, on the argument that it doesn’t satisfy some specific criteria, in particular truth-convergence. I’m arguing that truth-convergence should clearly be thrown out as a criterion because it’s impossible to satisfy (although it can be revised into more feasible criteria). On my understanding, you yourself seemed to agree about its infeasibility, although you seemed to think we should focus on different arguments about why it is infeasible (you said that my argument misses the point). But (on my reading) you continued to seem to reject coherentism on the same argument, namely the truth-convergence problem. So I am confused about your view.
To reiterate: then why would they continue to be enshrined as The Desiderata?
I mean, looking back, it was me who initially agreed with the OP [original post], in a top-level comment, that Eliezer’s essay was essentially correct. You questioned that, and I have been defending my position… and now it’s somehow off-topic?
I don’t see where.
Certainty. A huge issue in early modern philosophy which has now been largely abandoned.
Completeness. Everything is either true or false, nothing is neither.
Consistency. Nothing is both true and false.
Convergence. Everyone can agree.
Objectivity: Everyone can agree on something that’s actually true
I dismissed it as an answer that fulfills all criteria, because it doesn’t fulfil Convergence. I didn’t use the word possible—you did in another context, but I couldn’t see what you meant. If nothing fulfils all criteria, then coherentism could be preferable to approaches with other flaws.
Because all of them individually can be achieved if you make trade offs.
Because things don’t stop being desireable when they are unavailable.
That isn’t at all clear.
Lowering the bar to whatever you can jump over ,AKA Texas sharpshooting,isn’t a particularly rational procedure. The absolute standard you are or are not hitting relates to what you can consistently claim on the object level: without the possibility convergence on objective truth, you can’t claim that people with other beliefs are irrational or wrong, (at least so long as they hit some targets).
It’s still important to make relative comparisons even if you can’t hit the absolute standard...but it’s also important to remember your relatively best theory is falling short of absolute standards.
Except for consistency, none of these seem actually desirable as requirements.
Certainty: Don’t need it, I’m fine with probabilities.
Completeness: I don’t need everything to be either true or false, I just need everything to have a justifiable probability.
Consistency: Agree.
Convergence+Objectivity: No Universally Compelling Arguments
That’s also Completeness.
Giving up on convergence has practical consequences. To be Consistent, you need to give up on the stance that your views are a slam dunk, and the other tribe’s views are indefensible.