I'm neither an extreme sceptic, nor an extreme realist, and for that reason, I'm going to push in one direction some of the time, and in the other in other places.
Gordon's Thesis
Our knowledge of the truth is fundamentally uncertain because of epistemic circularity caused by the Problem of the Criterion.
More than that..for instance, semantic vagueness has nothing to do with the problem of the criterion. Uncertainty can creep into the map, (eg semantic confusion); the territory , (eg. quantum uncertainty); and the the relationship between them , (eg. the problem of the criterion). But the situation is not hopeless because these problems affect our knowledge making efforts to different extents.
We manage fundamental uncertainty by making pragmatic assumptions that lead us to believe in the truth of claims that help us achieve our goals.
Consequently, the truth that can be known is not independent of us, but rather dependent on that for which we care.
Ok, maybe, but one needs to avoid the basic irrationalism of thinking one can make things true just by believing them.
If you maintain truth and usefulness as separate concepts, and if you stick to a correspondence theory of truth , then it is quite difficult to show that truth depends on what we care about. It is more than usually uncertain claim.
That truth is fundamentally uncertain and grounded in care has far-reaching implications for many of the world's hardest-to-solve problems.
It's also important that uncertainty is very unevenly distributed.
And that attitudes about it are very tribal.
If you belong to the continental/postmodern , you are (apart from being unlikely to be reading this) probably leaning too far in the direction of fundamental uncertainty and modest epistemology.
If you are a rationalist or scientism-ist you are probably not leaning far enough, and for you it will be more therapeutic.
If you are a rationalist or scientism-ist , you are likely to protest that we obviously do have knowledge, because we can demonstrably predict things and achieve practical results. And you'd be right, up to a point.. but only up to a point , because there are different kinds of truth and knowledge that suffer from fundamental uncertainty to differing extents.
If you are a theist or mystic , you are likely to believe in quite different kinds of epistemic justification to scientists and philosophers, which is another if the things ,coat from overall uncertainty , that makes it hard for people to agree. There's a qualitative problem and well as a way quantitative one
It's hard to be therapeutic to all the people all the time. But "there is fundamental uncertainty" needs to be balanced out with "but not to the same extent everywhere". (Fortunately, since strong all embracing scepticism is self defeating).
Chapter 1. How do we know what knowledge is?
But naive epistemology breaks down when there's disagreement"
Even more so when it's disagreement about epistemology!
Circular dependencies stymies reductionism, which may well explain why philosophy makes slow progress compared to science.
"Just use reductionism" doesn't work in philosophy , because its not clear what lies at the bottom of the stack, or if anything does. Logic, epistemology and ontology have all been held to be First Philosophy at different times. Logic, epistemology and ontology also seem to interact. Correct ontology depends on correct epistemology..but what minds are capable of knowing depends on ontology. Logic possibly depends on ontology too, since quantum mechanics arguable challenges traditional bivalent logic. One of the reasons philosophy is difficult is that it has a circular structure.
Well, how would we prove that logic and direct observation are true means of assessing truth? We'd have to show they're true using the self-same means of assessing truth: logic and direct observation. This sets up an infinite regress of justifications, like two mirrors endlessly reflecting back on each other. The only way we can avoid this endless loop of self-referential logic is by making one or more assumptions without justification and then using those assumptions to ground the rest of our reasoning.
On the bright side, criteria dont have to be grounds or foundational -- one doesn't necessarily need even more foundational assumptions to ground ones foundational assumptions. This where the pragmatic twist comes in useful. There's an argument that treating empirical evidence as probably true allows one to make progress , whilst refusing to blocks progress. Meaning that empiricism doesn't give you necessary truth, but does give you something to work on.
Observation and deduction are famously plausible sources if truth -- the two dogmata -- but they are not sufficient to t to do science .. at least realistic science, science that goes beyond making predictions. Deduction tells you what observations imply, but not what is causing them .. the ontology, the noumenal reality , the underlying mechanism, the thing that scientific realism adds to scientific instrumentalism.
Induction doesn't give you much of an explanatory hypothesis, beyond the simple claim that there is some sort of causal stability that makes the same observed patterns likely to repeat.
To get at the ontology, the noumenal reality , the underlying mechanism, you need conjecture as well. Conjecture, or hypothesis, us the process if guessing a possible mechanism. The need for conjecture usproblem, because it is much less truth-apt, in itself, than deduction and observation.
Conjecture in itself is pretty much making things up. In the context of science, conjectures can at least be tested. In the hypothetico-deductive model, the scientist 1. Conjectures a mechanism, 2. deduces expected results from it, and 3. tests for the presence expected results and absence of unexpected ones.
So conjecture in the context of science not just making things up ... but it's not honing in on a singular truth either. In particular , the method of conjecture-deduction-and-testing can leave you with more than one theory that has passed the empirical tests.
Thats one of the reasons "just use science to solve everything" doesn't work. Another is that some things just aren't empirical at all. Not that they are ghostly, but you can't see things like meaning and goodness.
Also, it hasn't worked. People keep reinventing This Sort of Thing, the idea of using science and logic to solve everything every few decades, and it always fizzles out.
Mechanised epistemology doesn't help at all -- because some sort of criterion, or fundamentally assumption about how knowledge works, has to be built into the machine.
Bayesianism has both problems, simultaneously. It is dependent on evidence, so it has all the problems of empiricism; and it has the problems of rationalism because it starts with priors. The argument for Bayesianism is that even agents with wildly differing priors can eventually agree, given sufficient evidence. But that is an argument about ideal Bayesians: in reality , the amount of evidence available might be too limited to allow convergence. The ability of realistic Bayesians to formulate hypotheses is also limited. Alice puts most of her credence on the one hypothesis that seems best supported to her, out of the hypotheses she has heard of, or thought up, but Bob might have a better hypothesis that's not in her set.
Do ordinary people need to understand the scientific method?
The epistemological problems that ordinary people face are to do with finding the causes or interpretation of a fairly fixed and known set of facts. People are currently trying to understand world events, the set of facts, in terms of conspiracy theories about behind-the-scenes actors. Exhortions about empiricism aren't much use to them. Ordinary deductive logic isn't , either, because what they are trying to do is abduction. What would be of more use is knowing about confirmation bias, driving too much, falsifiability ... and that not everything happens for a reason.
The first thing I'll ask is that you always remember, on every page, that everything adds up to normality. What do I mean? It's a phrase I've borrowed from science fiction author Greg Egan, and it means that things continue to be just as they were even when you come to have a new understanding of them"
Why do people keep repeating that? It's not the case that "everything* is the same when you switch to a different explanation. It is the case that everything looks the same when you switch to an empirically acceptable explanation, because that's what empirical acceptability is -- what is observed is predicted by them, and what is not observed is not. But that doesn't happen automatically, it's not given to us free by the universe, it's the result of our epistemic processes -- we reject explanations that don't "save appearances". And appearances, observations, empirical data aren't "everything", either. It's perfectly possible for beliefs about ethics or decision theory to rest on on beliefs about reality.
Chapter 2. How do we know what words mean?
In Chapter 2 we saw that all definitions are ultimately grounded in inference from experience, and this makes it hard to precisely define what words mean
We initially learn language from examples and usage, but we progress to bring able to read dictionary definitions. We can also coin jargon terms that are clearer than natural language, so that semantics is one of the more fixable issues.
Chapter 3. Why can't we all agree on what's good?
The question could have been cast in terms of what's virtuous or whats right. Casting it in terms of the good biases the discussion in terms of consequentialism. Having said that, the answer is:-
The virtuous/good/right are not clearly visible, so raw empiricism is of no use.
Pure logic and maths don't answer any ethical questions because standard formalisations don't have any ethical axioms. Axioms can be added -- there is a logico-mathematical approach to ethics -- but it isn't clear which ones, eg. the Kantian or Millsian.
Ethics at least potentially rests on ontology, so it's as least as uncertain, and ontoloy is a lot more uncertain than prediction.
Individuals aren't starting from a blank slate , they are starting from what their culture tells them , which is very variable. (No, there's no single coherent "human value").
The fact that moral agreement is difficult isn't itself mysterious, so long as you don't hold to naive Scientism. Naive Scientism tells you "everything" is soluble by science...so ethics is as well." But science doesn't claim to be able to pronounce on matters of value ... so naive Scientism goes beyond science.
So not everything is equally uncertain, again.
Many have tried to provide clear intensional definitions of "good" and "bad", but, alas, those definitions have proven insufficient. Either they're circular, like dictionary definitions that say "good" means "morally right" and "bad" means "morally wrong", or specific in ways that many people disagree with."
Yes, there are two problems there: the problem of providing a non vacuous definition, and the problem of providing a convincing theory. The second problem is much harder.
It would take multiple books to explain everything there is to know about Bayesians"
But it wouldn't do the average person much good at all , because Bayes isn't a complete epistemology. In over a century , Bayes has not achieved more than a niche role in science. In a couple of decades , Yudowsky's Bayesian rationality hasn't solved any notable problems..
Similarly Aumann's Agreement Theorem, has no relevance to real life.
In the same way, we are able to agree about morals to the extent that our notions of "good" and "bad" are compatible enough to allow agreement, and we disagree to the extent they are in fundamental conflict. That we sometimes agree and sometimes don't shapes the landscape of our moral conflicts and determines much about how we try to resolve them."
Or you can agree about meta ethics, and derive your object level ethics from that.
Ethical persuasion happens. It often happens by persuading someone of an abstract principle, and then showing that they should have different object level ethics based on the principle. For instance, you can persuade someone out of sexism on the basis of racism.
Ethical persuasion is an easier problem than ethical convergence
Chapter 4. Why don't we always do what we should?
Grace could restrict herself to evaluating only systems sufficiently less powerful than her current one such that she could safely prove whether or not they're sound. She won't bother to do this, though, because less powerful systems won't help her achieve her goal. Her whole purpose in evaluating formal systems is to find more powerful ones that will make her better at Bayesian reasoning. A less powerful formal system that places additional restrictions on her thinking would be antihelpful. She thus finds herself up against the Löbian obstacle: she can't prove the soundness of a formal system that's equally or more capable than her current system, and any system she can prove to be sound will be weaker and thus less useful for helping her achieve her goal.
This generalises beyond formal systems into informal epistemology. Simply abandoning certainty doesnt solve epistemology, because it doesn't tell you what to do instead. Simply adopting probability doesn't solve all the problems, because probabilistic inference still relies on axioms, and probabilistic empiricism still has a instrumentalism realism divide.
The various approaches that are left in the absence of complete certainty about everything can be seen as abandoning different epistemological desiderata.
Relativism means giving up on a single truth.(As does coherentism)
Pragmatism means giving up in truth in favour of usefulness.
Skepticism means giving up on truth and knowledge.
What are the Desiderata?
Certainty. What is true is necessarily true. A huge issue in early modern philosophy which has been largely abandoned in contemporary Completeness. Everything is either true or false, nothing is neither.
Consistency. Nothing is both true and false..
Convergence. Whatever various assumptions a group of individuals start from, they will end with the same conclusions, given enough evidence and reflection. (Convergence implies non arbitrariness).
But does this explanation really answer Doug's questions? The proofs that establish Bayesian idealness are mathematical arguments. Therefore, to trust that Bayes' Theorem is the correct way to update beliefs, we must also trust that mathematical proofs correctly establish what's true"
They dont, because they depend on axioms. The mathmatical , ie. axiomatic, argument for Bayes rests on the correctness of the axioms. But pragmatic proof, such as betting arguments , is available as well.
Chapter 6. How do we know what we know?
More true our map is. This method works remarkably well, and it's how scientists, engineers, and many others build accurate models of the world that enable them to do their work. And although not necessarily framed as testing, this is also essentially what Bayesians are doing when they update their beliefs: they're checking their map to see if it's accurate, then updating it if they find a mismatch with what they observe.
They are checking their map to see if it's predictive. They can not make direct comparison to reality, any more than anyone else. Updating us better than nothing, but it's not everything.
But how do we know that such testing is a reliable way to verify the truth, and can we be certain that our methods of testing are valid? Alas, we don't and we can't, or at least not with complete certainty. This time, it's the Problem of the Criterion that stands in our way, and, as we'll see, it's the reason that all our knowledge of the truth is fundamentally uncertain."
The Problem of the Criterion is straightforward to state. To know if a claim is true, we need a method for testing if it's true. Suppose we have such a method, known as the criterion of truth. Is the criterion of truth valid"
There are different kinds of knowledge and their confirmation has different difficulty levels. The easiest is pragmatic knowledge. It's almost analytical that if you can do something, you know how to do it. (In a way that's quite different to hoping that predictive accuracy somehow adds up to correspondence-to-reality.) So perhaps the pragmatic approach is the best starting point for understanding how we do know as much as we know -- how it does add up to normality.
Is it circular to use reliability and predictive ability as evidence of themselves? More to the point, is it circular in a bad way?
Circularity is often condemned our of hand, but why? In the context of deductive reasoning circularity implies quodlibet, the ability to prove anything. There are infinite arguments of the form "P, therefore P". Moreover , it would be possible to prove infinitely many contradictions by setting Q=not P.
That's in the context of deduction. The circular justification of induction doesn't have the problem of quodlibet, because when you are doing induction you are not starting from an arbitrary claim, you are starting from evidence.
My conclusion
As with many philosophical problems, the hard problem of epistemology is to find a single theory that fully meets a set of desiderata. Philosophers have essentially discovered that this is impossible. In further bad news, it isn't clear which partial approach is the best. In better news, a partial solution is possible , and better than no solution. Strong scepticism is self defeating, but moderate scepticism, which is the same thing as modest epistemology, is available.
If epistemology is doubtful, everything in philosophy is doubtful. If there are reasons to think epistemology is unsolvable, there are reasons to think philosophy is. The problems with epistemology are particularly acute because it takes an epistemology to decide an epistemological question ... there's no neutral ground to start from.
Even if metaphilosophy can't tell you witch epistemology to switch to, it can tell you to something, which is to switch to modesty, if you haven't already.
My Introduction.
This is a review of @Gordon Seidoh Worley's book on epistemology.
I'm neither an extreme sceptic, nor an extreme realist, and for that reason, I'm going to push in one direction some of the time, and in the other in other places.
Gordon's Thesis
More than that..for instance, semantic vagueness has nothing to do with the problem of the criterion. Uncertainty can creep into the map, (eg semantic confusion); the territory , (eg. quantum uncertainty); and the the relationship between them , (eg. the problem of the criterion). But the situation is not hopeless because these problems affect our knowledge making efforts to different extents.
Ok, maybe, but one needs to avoid the basic irrationalism of thinking one can make things true just by believing them.
If you maintain truth and usefulness as separate concepts, and if you stick to a correspondence theory of truth , then it is quite difficult to show that truth depends on what we care about. It is more than usually uncertain claim.
It's also important that uncertainty is very unevenly distributed.
And that attitudes about it are very tribal.
If you belong to the continental/postmodern , you are (apart from being unlikely to be reading this) probably leaning too far in the direction of fundamental uncertainty and modest epistemology.
If you are a rationalist or scientism-ist you are probably not leaning far enough, and for you it will be more therapeutic.
If you are a rationalist or scientism-ist , you are likely to protest that we obviously do have knowledge, because we can demonstrably predict things and achieve practical results. And you'd be right, up to a point.. but only up to a point , because there are different kinds of truth and knowledge that suffer from fundamental uncertainty to differing extents.
If you are a theist or mystic , you are likely to believe in quite different kinds of epistemic justification to scientists and philosophers, which is another if the things ,coat from overall uncertainty , that makes it hard for people to agree. There's a qualitative problem and well as a way quantitative one
It's hard to be therapeutic to all the people all the time. But "there is fundamental uncertainty" needs to be balanced out with "but not to the same extent everywhere". (Fortunately, since strong all embracing scepticism is self defeating).
Chapter 1. How do we know what knowledge is?
Even more so when it's disagreement about epistemology!
Circular dependencies stymies reductionism, which may well explain why philosophy makes slow progress compared to science.
"Just use reductionism" doesn't work in philosophy , because its not clear what lies at the bottom of the stack, or if anything does. Logic, epistemology and ontology have all been held to be First Philosophy at different times. Logic, epistemology and ontology also seem to interact. Correct ontology depends on correct epistemology..but what minds are capable of knowing depends on ontology. Logic possibly depends on ontology too, since quantum mechanics arguable challenges traditional bivalent logic. One of the reasons philosophy is difficult is that it has a circular structure.
On the bright side, criteria dont have to be grounds or foundational -- one doesn't necessarily need even more foundational assumptions to ground ones foundational assumptions. This where the pragmatic twist comes in useful. There's an argument that treating empirical evidence as probably true allows one to make progress , whilst refusing to blocks progress. Meaning that empiricism doesn't give you necessary truth, but does give you something to work on.
Observation and deduction are famously plausible sources if truth -- the two dogmata -- but they are not sufficient to t to do science .. at least realistic science, science that goes beyond making predictions. Deduction tells you what observations imply, but not what is causing them .. the ontology, the noumenal reality , the underlying mechanism, the thing that scientific realism adds to scientific instrumentalism.
Induction doesn't give you much of an explanatory hypothesis, beyond the simple claim that there is some sort of causal stability that makes the same observed patterns likely to repeat.
To get at the ontology, the noumenal reality , the underlying mechanism, you need conjecture as well. Conjecture, or hypothesis, us the process if guessing a possible mechanism. The need for conjecture usproblem, because it is much less truth-apt, in itself, than deduction and observation.
Conjecture in itself is pretty much making things up. In the context of science, conjectures can at least be tested. In the hypothetico-deductive model, the scientist 1. Conjectures a mechanism, 2. deduces expected results from it, and 3. tests for the presence expected results and absence of unexpected ones.
So conjecture in the context of science not just making things up ... but it's not honing in on a singular truth either. In particular , the method of conjecture-deduction-and-testing can leave you with more than one theory that has passed the empirical tests.
Thats one of the reasons "just use science to solve everything" doesn't work. Another is that some things just aren't empirical at all. Not that they are ghostly, but you can't see things like meaning and goodness.
Also, it hasn't worked. People keep reinventing This Sort of Thing, the idea of using science and logic to solve everything every few decades, and it always fizzles out.
Mechanised epistemology doesn't help at all -- because some sort of criterion, or fundamentally assumption about how knowledge works, has to be built into the machine.
Bayesianism has both problems, simultaneously. It is dependent on evidence, so it has all the problems of empiricism; and it has the problems of rationalism because it starts with priors. The argument for Bayesianism is that even agents with wildly differing priors can eventually agree, given sufficient evidence. But that is an argument about ideal Bayesians: in reality , the amount of evidence available might be too limited to allow convergence. The ability of realistic Bayesians to formulate hypotheses is also limited. Alice puts most of her credence on the one hypothesis that seems best supported to her, out of the hypotheses she has heard of, or thought up, but Bob might have a better hypothesis that's not in her set.
Do ordinary people need to understand the scientific method?
The epistemological problems that ordinary people face are to do with finding the causes or interpretation of a fairly fixed and known set of facts. People are currently trying to understand world events, the set of facts, in terms of conspiracy theories about behind-the-scenes actors. Exhortions about empiricism aren't much use to them. Ordinary deductive logic isn't , either, because what they are trying to do is abduction. What would be of more use is knowing about confirmation bias, driving too much, falsifiability ... and that not everything happens for a reason.
Why do people keep repeating that? It's not the case that "everything* is the same when you switch to a different explanation. It is the case that everything looks the same when you switch to an empirically acceptable explanation, because that's what empirical acceptability is -- what is observed is predicted by them, and what is not observed is not. But that doesn't happen automatically, it's not given to us free by the universe, it's the result of our epistemic processes -- we reject explanations that don't "save appearances". And appearances, observations, empirical data aren't "everything", either. It's perfectly possible for beliefs about ethics or decision theory to rest on on beliefs about reality.
Chapter 2. How do we know what words mean?
We initially learn language from examples and usage, but we progress to bring able to read dictionary definitions. We can also coin jargon terms that are clearer than natural language, so that semantics is one of the more fixable issues.
Chapter 3. Why can't we all agree on what's good?
The question could have been cast in terms of what's virtuous or whats right. Casting it in terms of the good biases the discussion in terms of consequentialism. Having said that, the answer is:-
The virtuous/good/right are not clearly visible, so raw empiricism is of no use.
Pure logic and maths don't answer any ethical questions because standard formalisations don't have any ethical axioms. Axioms can be added -- there is a logico-mathematical approach to ethics -- but it isn't clear which ones, eg. the Kantian or Millsian.
Ethics at least potentially rests on ontology, so it's as least as uncertain, and ontoloy is a lot more uncertain than prediction.
Individuals aren't starting from a blank slate , they are starting from what their culture tells them , which is very variable. (No, there's no single coherent "human value").
The fact that moral agreement is difficult isn't itself mysterious, so long as you don't hold to naive Scientism. Naive Scientism tells you "everything" is soluble by science...so ethics is as well." But science doesn't claim to be able to pronounce on matters of value ... so naive Scientism goes beyond science.
So not everything is equally uncertain, again.
Yes, there are two problems there: the problem of providing a non vacuous definition, and the problem of providing a convincing theory. The second problem is much harder.
But it wouldn't do the average person much good at all , because Bayes isn't a complete epistemology. In over a century , Bayes has not achieved more than a niche role in science. In a couple of decades , Yudowsky's Bayesian rationality hasn't solved any notable problems..
Similarly Aumann's Agreement Theorem, has no relevance to real life.
Or you can agree about meta ethics, and derive your object level ethics from that.
Ethical persuasion happens. It often happens by persuading someone of an abstract principle, and then showing that they should have different object level ethics based on the principle. For instance, you can persuade someone out of sexism on the basis of racism.
Ethical persuasion is an easier problem than ethical convergence
Chapter 4. Why don't we always do what we should?
This generalises beyond formal systems into informal epistemology. Simply abandoning certainty doesnt solve epistemology, because it doesn't tell you what to do instead. Simply adopting probability doesn't solve all the problems, because probabilistic inference still relies on axioms, and probabilistic empiricism still has a instrumentalism realism divide.
The various approaches that are left in the absence of complete certainty about everything can be seen as abandoning different epistemological desiderata.
Relativism means giving up on a single truth.(As does coherentism)
Pragmatism means giving up in truth in favour of usefulness.
Skepticism means giving up on truth and knowledge.
What are the Desiderata?
Certainty. What is true is necessarily true. A huge issue in early modern philosophy which has been largely abandoned in contemporary Completeness. Everything is either true or false, nothing is neither.
Consistency. Nothing is both true and false..
Convergence. Whatever various assumptions a group of individuals start from, they will end with the same conclusions, given enough evidence and reflection. (Convergence implies non arbitrariness).
They dont, because they depend on axioms. The mathmatical , ie. axiomatic, argument for Bayes rests on the correctness of the axioms. But pragmatic proof, such as betting arguments , is available as well.
Chapter 6. How do we know what we know?
They are checking their map to see if it's predictive. They can not make direct comparison to reality, any more than anyone else. Updating us better than nothing, but it's not everything.
There are different kinds of knowledge and their confirmation has different difficulty levels. The easiest is pragmatic knowledge. It's almost analytical that if you can do something, you know how to do it. (In a way that's quite different to hoping that predictive accuracy somehow adds up to correspondence-to-reality.) So perhaps the pragmatic approach is the best starting point for understanding how we do know as much as we know -- how it does add up to normality.
Is it circular to use reliability and predictive ability as evidence of themselves? More to the point, is it circular in a bad way?
Circularity is often condemned our of hand, but why? In the context of deductive reasoning circularity implies quodlibet, the ability to prove anything. There are infinite arguments of the form "P, therefore P". Moreover , it would be possible to prove infinitely many contradictions by setting Q=not P.
That's in the context of deduction. The circular justification of induction doesn't have the problem of quodlibet, because when you are doing induction you are not starting from an arbitrary claim, you are starting from evidence.
My conclusion
As with many philosophical problems, the hard problem of epistemology is to find a single theory that fully meets a set of desiderata. Philosophers have essentially discovered that this is impossible. In further bad news, it isn't clear which partial approach is the best. In better news, a partial solution is possible , and better than no solution. Strong scepticism is self defeating, but moderate scepticism, which is the same thing as modest epistemology, is available.
If epistemology is doubtful, everything in philosophy is doubtful. If there are reasons to think epistemology is unsolvable, there are reasons to think philosophy is. The problems with epistemology are particularly acute because it takes an epistemology to decide an epistemological question ... there's no neutral ground to start from.
Even if metaphilosophy can't tell you witch epistemology to switch to, it can tell you to something, which is to switch to modesty, if you haven't already.