The Problem Of The Criterion

I keep finding cause to discuss the problem of the criterion, and then I figured I'd try my paw at writing up a mail explaining it. I don't have a groovy runway record on writing clear explanations, but I'll exercise my all-time and include lots of links you can follow to make upwardly for any inadequacy on my part.

Motivation

Before we go to the trouble itself, permit's talk most why it matters.

Allow'due south say yous want to know something. Doesn't really matter what. Peradventure you only desire to know something seemingly benign, similar what is a sandwich?

At first this might seem pretty easy: you know a sandwich when y'all come across it! But just to be sure you ask a agglomeration of people what they think a sandwich is and if detail things are sandwiches are not.

Uh oh...

Yous've run headlong into the classic trouble of how to carve upwards reality into categories and assign those categories to words. I'll skip over this function because it'southward been addressed to death already.

And then now you've come out the other side accepting that "sandwich", similar near all categories, has nebulous boundaries, and that at that place is no true sandwich of which you can speak.

Fine, simply being non i easily deterred, y'all come up upward with a very precise, that is a mathematically and physically precise, definition of a sandwich-similar object you call an FPS, a Finely Precise Sandwich. Now you lot want to know whether or not something is an FPS.

You check the conditions and it all seems expert. You have an FPS. Only wait! How exercise you know each condition is true? Mayhap i of your weather is that the FPS is exactly three inches tall. How do you know that it'southward really 3 inches alpine?

Oh, you lot used a ruler? How do you know the ruler is accurately measuring 3 inches? And furthermore, how do y'all know your eyes and brain tin can be trusted to read the ruler correctly to assess the elevation of the would-be FPS? For that matter, how do you even know what an inch is, and why was 3 inches the true height of an FPS anyway?

Heck, what does it fifty-fifty mean to "know" that something is "true"?

If you go on reducing like this, you'll somewhen hit lesser and run into this question: how do you know that something is true? At present you've encountered the problem of the criterion!

The Problem

How exercise yous know something is true? To know if something is true, you have some method, a criterion, by which you appraise its veracity. But how do yous know this criterion is itself truthful? Oh, you lot have a method, i.east. a criterion, by which you assess its veracity.

Oops! Infinite recursion detected!

The problem of the benchmark is normally attested to originate with the aboriginal Greek philosopher Pyrrho. Since then interest in it has come up in and out of favor. Roderick Chisholm, the mod philosopher who rejuvenated interest in the problem, ofttimes phrases information technology as a pair of questions:

What exercise we know? What is the extent of our knowledge?
How are we to decide whether we know? What are the criteria of knowledge?

This creates a kind of epistemic circularity where we go round in circles trying to justify our knowledge with our cognition yet are never able to grab concur of something that itself need not be justified by something already known. If you're familiar with the symbol grounding problem, information technology's substantially the same problem, generalized (cf. SEP on metaphysical grounding).

To that signal, the problem of the benchmark is really more a category of related epistemological problems that accept slightly different forms depending on how they manifest the cadre problem of grounding justification knowledge. Some other problems that are simply the trouble of the benchmark in disguise:

the problem of induction (SEP)
finding the universal prior (LW)
the problem of perception (and specifically epistemological bug of perception)
causality
no gratuitous luncheon theorems
metaethical uncertainty (not only normative uncertainty, only doubtfulness about what norms are even like)

Really whatsoever kind of fundamental incertitude about what is truthful, what is known, what tin be trusted, etc. is rooted in the problem of the criterion. Thus how we respond to it impacts near any endeavor that seeks to make utilise of data.

Okay, and then we have this problem. What to do nigh it?

Possible Responses

Showtime upwards, in that location are no known solutions to the trouble of the criterion, and it appears, properly speaking, unsolvable. Ultimately all responses either beg the question or unask, dissolve, or reframe the question in whole or in office. All the same, we tin can learn something from considering all the ways nosotros might address it.

Chisholm argues that there are only three possible responses to the problem: particularism, methodism, and skepticism. I'd instead say at that place are only three ways to attempt to solve the problem of the criterion, and other responses are possible if we give up finding a proper solution. As mentioned, all these attempts at solutions ultimately beg the question and so none actually resolve the trouble—hence why information technology's argued that the problem is unsolvable—only they are popular responses and deserve consideration to meliorate empathize why the trouble of the criterion is so pernicious.

Particularism is an try to resolve the problem of the criterion by picking detail things and declaring them truthful, trusted, etc. by fiat. If you've ever dealt with axioms in a formal system, you've engaged in a particularist solution to the trouble of the criterion. If you're familiar with Wittgenstein'southward notion of hinge propositions, that's an instance of particularism. My impression is that this is widely considered the all-time solution since, although information technology leaves us with some number of unverified things nosotros must trust on faith, in practice it works well enough past merely stopping the infinite regress of justifications at some point past which it doesn't matter (more than on the pragmatism of when to stop presently). To paraphrase Chisholm, the master merit of particularism is that the trouble of the criterion exists yet we know things anyway.

Methodism tries to solve the problem of the criterion by picking the criterion rather than some axioms or swivel propositions. Descartes is probably the best known epistemic methodist. The problem, argues Chisholm and others, is that methodism collapses into particularism where the thing taken on organized religion is the criterion! Therefore nosotros tin can but ignore methodism every bit a special case of particularism.

And and then there's skepticism, arguably the only "correct" position of Chisholm'south three in that it's the only ane that seemingly doesn't require assuming something on faith. Spoiler alert: it does because information technology yet needs some reason to prefer skepticism over the alternatives, thus it still ends up begging the question. Skepticism is also not very useful because even though it might not lead to incorrectly assertive that a false thing is truthful, it does this by not allowing one to believe anything is truthful! It's also not clear that humans are capable of true skepticism since we clearly know things, and information technology seems that perhaps our brains are designed such that we can't help simply know things, even if they are truthful just believed without properly grounded justification. So, alas, pure skepticism doesn't seem workable.

Despite Chisholm'due south claim to those beingness the only possible responses, some other responses be that turn down the premise of the problem in some way. Permit's consider a few.

Coherentist responses decline the idea that truth, cognition, etc. must be grounded and instead seek to notice a style of balancing what is known with how it is known to form a cocky-consequent system. If you're familiar with the method of reflective equilibrium (SEP), that'due south an example of this. Arguably this is what modern scientific discipline actually does, repeatedly gathering evidence and reconsidering the foundations to produce something similar a self-consistent arrangement of noesis, though at the cost of giving upward (total) completeness (LW, SEP).

Another manner of rejecting the premise is to give upwards the thought that the problem matters at all via epistemic relativism. In its strongest grade, this gives up both any notion of objectivity or intersubjectivity and just accepts that knowledge is totally subjective and ultimately unverifiable. In practise this is a kind of attractor position for folks burnt out on there-is-simply-one-ontology scientism who overcorrect too far in the other direction, and although some of the arguments made for relativism are valid, consummate rejection or even heavy devaluing of intersubjectivity makes this position essentially a solipsistic one and thus, like extreme skepticism, non useful for much.

Finally, we come to a special case of particularist responses known as pragmatism, and information technology's this kind of response that Yudkowsky offered. The idea of pragmatism is to say that at that place is some purpose to be served and past serving that purpose we tin can do an end-run effectually the trouble of the criterion by tolerating unjustified noesis and so long as information technology works well enough to achieve some terminate. In Yudkowsky'south case, that end is "winning". We might summarize his response as "exercise the all-time you can in order to win", where "win" here means something similar "alive a maximally fulfilling life". I'd argue this is basically right and in practice what almost people do, even if their best is oft not very good and their idea of winning is adequately limited.

Nonetheless, I find something lacking in pragmatic responses, and in particular in Yudkowsky'due south response, because they too easily turn from pragmatism to motivated stopping of epistemic reduction. If pragmatism becomes a fashion to sweep the problem of the criterion under the rug, then the lens has failed to see its own flaws. More is possible if we can footstep back and hold both the problem of the criterion and pragmatism near it simultaneously. I'll endeavour to sketch out what that means.

Holding the Trouble

At its heart, the problem of the criterion is a problem by virtue of being trapped by its own framing. That is, it'due south a trouble considering nosotros desire to know about the world and empathize information technology and take that knowledge and understanding fit within some coherent ontological system. If we stopped trying to map and model the globe or gave upward on that model being true or predictive or otherwise useful and merely allow it exist there would be no problem.

We're not going to practice that, though, because then we'd be rocks. Instead nosotros are optimization processes, and that requires optimizing for something. That something is what we care almost, our telos, the things we are motivated to do, the stuff we desire, the purposes we take, the things we value, and maybe even the something nosotros have to protect. And instrumental to optimization is building a adept enough map to successfully navigate through the territory to a earth state where the something optimized for is in fact optimized.

So nosotros're going to endeavour to sympathize the earth because that'southward the kind of beings we are. But the existence of the problem of the criterion suggests we've set ourselves an impossible job that we can never fully consummate, and we are forced to endeavor in it because the only other selection is ceasing to exist in the world. Thus nosotros seem inescapably trapped by the tension of wanting to know and not being able to know perfectly.

As best I tin can tell, the only style out is up, equally in upwards to a college frame of thinking that can concur the problem of the criterion rather than be held by it. That is, to exist able to simultaneously acknowledge that the problem exists, accept that you must both accost it and by virtue of addressing it your ontology cannot be both universal and everywhere coherent, and also get out yourself enough space betwixt yourself and the map that you can look up from it and discover the territory just equally it is.

That's why I call up pragmatism falls brusque on its own: it'south role of the response, but not the whole response. With but a partial response nosotros'll continually find ourselves lost and confused when nosotros need to serve a different purpose, when we modify what we care most, or when we encounter a part of the globe that can't be made to fit within our existing understanding. Instead, we demand to deal with the key dubiety in our knowledge as best nosotros tin while non forgetting we're limited to doing our best while falling short of achieving perfection because we are bounded beings who are embedded in the world. Doing and so cultivates a deep form of epistemic humility that not only acknowledges that we might be mistaken, but that our very notion of what it means to be mistaken or correct is itself non fully knowable fifty-fifty every bit we go on with living all the same.

NB: Holding the trouble of the criterion is difficult.

Before nosotros get farther, I want to acknowledge that actually holding the problem of the criterion every bit I describe in this section is hard. Information technology'south difficult because successfully doing what I describe is not a purely academic or intellectual pursuit: it requires going out into the world, really doing things that require you to grapple with the trouble of the criterion, frequently failing, and then learning from that failure. And fifty-fifty afterwards all that there'due south a good chance you'll still make mistakes all the time and get dislocated about primal points that you understand in theory. I know I certainly do!

Information technology'due south also not an all at in one case process to learn to concur the problem. You tin can detect comments and posts of mine over the past few years showing a slowly building improve understanding of how to respond to the problem of the criterion, so don't become too downwardly on yourself if you read the above, aspire to hold the problem of the criterion in the way I describe, and yet find it nigh incommunicable. Be gentle and just keep trying.

And keep exploring! I'm not convinced I've presented some terminal, ultimate response to the problem, and so I expect to learn more and have new things to say in time. What I present is just equally far every bit I've gotten in wrangling with information technology.

Implications

Having reached the depths of the problem of the criterion and found a way to answer, let's consider some places where it touches on the projects of our lives.

A straightforward issue of holding the problem of criterion and adopting pragmatism near information technology is that all knowledge becomes ultimately teleological knowledge. That is, there is always some purpose, motivation, or human business organization backside what nosotros know and how nosotros know information technology considering that's the mechanism we're using to fix the gratis variable in our ontology and cull amongst particular hinge propositions to assume, even and peculiarly if those propositions are implicit. Said another style, no knowledge is better or worse without beginning choosing the purpose by which cognition tin can be evaluated.

This is not a license to believe whatever we want, though. For most of us, most of the time, our purpose should probably be to believe that which all-time predicts what we notice about the world, i.e. nosotros should believe what is true. The key insight is to non get dislocated and retrieve that a norm in favor of truth is the same thing as truth existence the merely possible way to confront reality. The ground of understanding is not solid, so to speak, and if we don't cull to make it house when needed it volition crumble away below our anxiety.

Thus, the problem of the criterion is also, as I hope will be clear, a killer of logical positivism. I point this out considering logical positivism is perennially appealing, is baked into the style science is taught in schools, and has a way of sneaking back in when one isn't looking. The problem of the criterion is non the but problem with logical positivism, but it's a big one. In the same way, the problem of the criterion is also a problem for direct realism about various things because the problem implies that there is a gap between what is perceived and the thing being perceived and suggests the best we can promise for is indirect realism, assuming you recollect realism virtually something is the correct approach at all.

Finally, all this suggests that dealing with knowledge and truth, whether every bit humans, AIs, or other beings, is complicated. Thus why we meet a certain amount of natural complexity in trying to build formal systems to grapple with noesis. Real solutions tend to expect like a combination of coherentism and pragmatism, and we see this in logical induction, Bayesian inference, and attempts to implement Solomonoff induction. This obviously has implications for building aligned AI that are nevertheless being explored.

Epitomize

This mail service covered a lot of footing, simply I retrieve it can exist summarized thusly:

There is a problem in grounding coherent noesis because to know something is to already know how to know something.
This circularity creates the trouble of the criterion.
The problem of the criterion cannot be adequately solved, but it can be addressed with pragmatism.
Pragmatism is non a complete response because no item pragmatism is privileged without respect to some purpose.
A complete response requires both holding the trouble as unsolvable and getting on with knowing things anyway in a country of perpetual epistemic humility.
The implications of necessarily grounding cognition in purpose are vast because the trouble of the benchmark lurks behind all cognition-based activeness.