Sunday, May 26, 2013

Some discussion of burdens of proof (based on a dialogue on Debunking Christianity)


The thread is here. 
That gets down to some very basic issues in epistemology. I did my doctoral work at a highly secular university philosophy program, but the epistemologists and probability theorist that I worked with, who certainly were not religious people, thought that classical foundationalism was a very problematic doctrine, and that the idea that a limited range of beliefs belonged in the "core" while other beliefs had to be proved by evidence, is in fact an unsupportable position. 
http://www.unc.edu/~theis/phil...
As a result, they were in general skeptical of the claim that one particular position as opposed to another had "THE burden of proof." To say that the burden of proof lies on one side or another that we know always, what beliefs can be accepted without proof and which ones need to be demonstrated, and that project doesn't look to be achievable. Descartes, for example, said that he would doubt everything and believe only what could be proved, and most people think his project didn't work. He started by doubting sense experience and then had to appeal the theological arguments to defend his belief in an external world. Hume's empiricism left him in a position where he had to "justify" the idea that the future will resemble the past simply by appealing to custom and habit. In other words, we really don't have justification for it. In other words, Hume avoids having to justify the belief that the future will resemble the past by claiming that belief this belief doesn't have the burden of proof.
From this one could conclude that you can show that just about any belief is unjustified simply by putting a heavy enough burden of proof on it. If you could only justifiably believe in the external world if you could prove that you aren't a brain in a vat, that might prove difficult.
So, for example, as I learned Bayesian theory, a popular theory was that prior probabilities were subjective, and that people who had different one could in theory eventually come to a consensus by adjusting their probabilities as evidence came in, but the idea of a "proper starting point" or "correct priors" was considered misguided. One of my teachers (again a religious nonbeliever) said that "you are justified in believing what you already believe, unless you have good reason to change your beliefs." I remember asking him about Descartes method of doubt, and in response he mentioned an ancient Greek skeptic who sat on the marketplace wagging his index finger because he couldn't believe anything. In other words, what I learned from the study of epistemology led me to the conclusion that fixing the burden of proof is pretty difficult, and that it is hard to discover a "proper" position for the burden of proof. There are relative burdens of proof that different individuals have for certain claims, but a "correct" location for the burden of proof seems to me difficult to justify.
So, for example, when I first encountered the Outsider Test for Faith, it looked to me as if it was another case of implying classical foundationalism, or perhaps, applying classical foundationalism to religious belief in a way that it is not applied to other types of beliefs, and some of my early responses to the OTF came from this perspective.
If you think the key to refuting religious belief is to inculcate a proper epistemology, which results in a proper location of the burden of proof, then I am likely to be pretty skeptical of that enterprise, and my skepticism comes not from my religion, but rather from widely held views in epistemology that I got from secular philosophy teachers. I'm not saying that these epistemologists couldn't be wrong, but it might take a little work to convince me that they are wrong. 

  • 24 comments:

    Doubting Marcus said...

    I agree that the burden of proof is often a muddled concept and sometimes wielded as a hatchet against whatever belief the speaker currently doesn't hold.

    However, the idea that there are no "correct priors," strikes me as demonstrably false. Priors are basically what evidence has looked like in the past for theories like yours.

    Yes, the available background evidence for each individual is different but you can't just insert whatever value you like and call it a justified prior. On the contrary, if what counts as background evidence is not agreed upon you can settle the dispute by arguing for your position by applying Bayes' theorem while treating that disputed background information as new evidence.

    If this wasn't the case there could be no accumulation of evidence because people could just assert whatever priors they like thus ignoring everything we've learned in the past.

    Crude said...

    However, the idea that there are no "correct priors," strikes me as demonstrably false. Priors are basically what evidence has looked like in the past for theories like yours.

    You're going to need priors to evaluate the evidence to begin with, or even say what does and doesn't count as evidence.

    Crude said...

    And just to comment on the other thread, I'll point out two common absurdities that always seem to come up when talking about religion and evidence.

    1) The claim that if someone thinks Christianity is right, they're committed to the claim that every other religion is completely wrong. But that's absurd - I can easily hold quite a lot of (say) judaism to be correct, but ultimately incomplete. There's a similar situation for Islam.

    2) The claim that someone accepting Christianity must either believe other religions have no evidence in their favor, or at the very least that their being correct cannot be a live possibility. Again, that just seems absurd. I'm willing to say it's a live possibility that Islam is correct after all, or maybe Mormonism - I could well be wrong. I'm able to grant that those views have some evidence or arguments in their favor, even if I think the evidence ultimately favors Catholicism.

    Someone there implied that if a Christian considers these other religions to be live options rather than dismissed out of hand, it should lead to a life of constant anxiety. I think that says far more about them than it does about religion or reason.

    BeingItself said...

    I think VR is essentially correct here.

    What I have noticed is this: skeptics tend to be more conservative in what they take to be basic beliefs, whereas believers tend to be more promiscuous.

    Also, I see believers make this kind of move: since I cannot justify induction (or whatever), they take that as a license for them to believe all kinds of nonsense without justification.

    joesmarts said...

    I see burden of proof as a social responsibility in a given discussion. That is, participants in a discussion have an obligation to take a position and support that position insofar as required by the other participants involved. Discussion is to be seen as a cooperation toward a particular goal. My idea is mostly birthed from the "cooperative principle" as outlined by Paul Grice in the Gricean Maxims.

    And, based on my experience, the New Atheists more often than not want to shun their responsibility to the discussion. They'd rather hide behind the clock of invisibility, er "lack of belief," than present a position and support it.

    joesmarts said...

    Er, that should be cloak.

    Doubting Marcus said...

    Crude,

    "You're going to need priors to evaluate the evidence to begin with, or even say what does and doesn't count as evidence."

    If you have no background information that makes one theory more likely than and others, you uniformly distribute the priors. That's basic probability theory. Are you disputing this?

    And I'm not sure if you meant you need initial (perhaps "foundational") priors to evaluate your very first claim, which you build future evaluations on, or that you need priors to evaluate what counts as evidence. If you meant the former, I'd again say you begin with uniform priors and use what happens in practice to set future priors. If you meant the latter, if you can't tell what counts as evidence for your theory it has nothing to do with the priors. It's murkiness, or possibly self-contradiction, in your theory.

    Crude said...

    If you have no background information that makes one theory more likely than and others, you uniformly distribute the priors. That's basic probability theory. Are you disputing this?

    Alright: give me an example of the probabilities you start with when you have absolutely no priors. Justify the probabilities you give, or explain why they need no justification.

    Papalinton said...

    joesmarts


    Does your group social responsibility obligation standing as an explanation include Jonestown or Branch Davidian? If not why not?

    joesmarts said...

    @Papalinton,

    I have no idea what you mean by "group social responsibility obligation standing as an explanation."

    Zach said...

    you are justified in believing what you already believe, unless you have good reason to change your beliefs.

    In practice the above quote is a recipe for intellectual laziness and inertia. The intellectually curious, truth-centered have a higher obligation to interrogate the rationale for the beliefs that we learned at mama's knee.

    Otherwise we end up (again, in practice, if not the spirit of the above quote) with people who think slavery is just fine because that's what they've always known and they see no good reason to question it (surprise surprise living on a plantation in the South they see no good reasons to question slavery).

    Descartes got it wrong in some ways, we now know, but he did not get it wrong in his admonition to doubt. Advocating doubt is different from advocating foundationalism. We must be careful not to conflate the two

    Doubting Marcus said...

    Crude,

    "give me an example of the probabilities you start with when you have absolutely no priors. Justify the probabilities you give, or explain why they need no justification."

    You can't have "absolutely no priors." My point was when you have a total absence of information to separate any particular theory from others you must distribute the priors equally over all possible options. As from your perspective any outcome is just as likely as any other.

    For example: Suppose you, like me, don't watch rugby but someone tells you there is a four team playoff, featuring team A, team B, team C, and team D, which will crown one champion. From your perspective, you have literally no reason to choose one team over any other as the eventual victor, your prior probability that any given team will win should be 25%. You might object that odds are that they aren't four evenly matched teams, and you might be right, but you have no information that would justify giving more of a chance to a particular team so this fact shouldn't make you randomly choose, say, team C as more likely to win than the even 1 in 4 odds.

    Note, however, that an expert in rugby, may be under no such obligation to evenly distribute her priors as she could have relevant discriminatory information to separate the ability of the teams. Given her background knowledge she could be justified in choosing, say, team D as having a 66% to win before the playoff begins. I say this to enforce the fact that just because priors are subjective that doesn't make them arbitrary.

    Crude said...

    Marcus,

    You can't have "absolutely no priors." My point was when you have a total absence of information to separate any particular theory from others you must distribute the priors equally over all possible options.

    Okay - then what are the priors you absolutely must start with to begin with?

    And why must I distribute the priors equally? How do I even know I'm aware of all the possible options?

    You give the example of 4 Rugby teams and conclude that you must regard each as having a 25% chance of winning. If I know nothing about Rugby, how do I know they can't all lose? Or that two teams can't win? Or that a terrorist event won't happen and the whole competition will be called off? Or any other number of things?

    finney said...

    I don't see the connection between your criticism of foundationalism and the conclusion that there is no discoverable burden of proof. It's much like saying, I do not know how many strands of hair are required to make a beard, and therefore there is no sense of speaking of beards.

    Doubting Marcus said...

    Crude,

    "If I know nothing about Rugby, how do I know they can't all lose?"

    Because it's an example where I explicitly said otherwise. See: "which will crown one champion" above.

    "And why must I distribute the priors equally? How do I even know I'm aware of all the possible options? ...Or that a terrorist event won't happen and the whole competition will be called off? Or any other number of things?"

    These are legitimate considerations but extremely unlikely events don't substantially change your calculations. There's always a non-zero probability that "aliens will invade and interrupt the playoff" but the prior for such a claim is so small, on the order of billions to one against, the effect it will have on your calculations will be effectively zero. Even all such improbable theories combined have a prior with a sum so low as to not substantially effect the equation, allowing us to reasonably treat all incredibly unlikely hypotheses as having a 0 percent prior and all very, very likely hypotheses as 1. In theory then the probability space has to be divided over infinitely many competing hypotheses but we can safely discard all but a few finite number of them, in my example that would mean only the four discrete team options. This does mean that in practice all Bayesian calculations are approximations but you don't need precision to the nth decimal place in order to have an effective epistemology.

    Of course if you have some specific evidence that a terrorist attack at such an event is likely that changes this rule, but in general when uniformly distributing your priors the level of contention of what number exactly to set consequent probabilities dwarfs any effect of absurdly improbable hypotheses by many orders of magnitude.

    Crude said...

    Marcus,

    Because it's an example where I explicitly said otherwise. See: "which will crown one champion" above.

    I was assuming you don't know anything about rugby. How much knowledge do you get to assume to make these calculations? And how do you justify assuming one thing over another?

    These are legitimate considerations but extremely unlikely events don't substantially change your calculations.

    How do you know they're extremely unlikely? I thought in the absence of knowledge you had to give these questions equal footing. What were the odds the US would win no medals in the 1980 Olympics?

    Doubting Marcus said...

    Crude,

    "How do you know they're extremely unlikely? I thought in the absence of knowledge you had to give these questions equal footing."

    Because you know other things about reality. You don't need to know anything about rugby to conclude aliens interrupting the match is extremely unlikely. However, because you seem to be missing the point of the example, the same idea could be made with a four sided die.

    If you were rolling a four-sided die it's possible the die could spontaneously explode or a vanish but your background knowledge of how often such objects spontaneously combust or disappear means you know, in the absence of specific evidence to indicate otherwise, that such events are unlikely. In practice then you don't have to account for the one in 100 quadrillion or so chance the die will explode when you roll it, and likewise with other such improbable events. You can safely distribute your priors evenly over the four possible outcomes that have more than a negligible chance (a one, two, three or four coming up) knowing that all other outcomes have such a low prior that they can be discarded in practice.

    "I was assuming you don't know anything about rugby. How much knowledge do you get to assume to make these calculations? And how do you justify assuming one thing over another?"

    Because I told you that you know nothing about rugby but that you are informed of the facts I explicitly listed. That's what makes it an example, a theoretical state of knowledge about a given situation. I would hope it's not considered a stretch to agree to go along with the details of such a thought experiment. Had you objected that you couldn't be informed of such facts so my example was impossible or too difficult to imagine, I would disagree but I'd see that as fair game. As it stands now though it seems you are rejecting the idea of thought experiments.

    Crude said...

    Marcus,

    If you were rolling a four-sided die it's possible the die could spontaneously explode or a vanish but your background knowledge of how often such objects spontaneously combust or disappear means you know, in the absence of specific evidence to indicate otherwise, that such events are unlikely.

    Sure, if I have knowledge of such and such things, then I can feed them into the equation. But how do I get that knowledge? I'm going to need priors to begin with. All the way back to the start where I have some given priors and probabilities. How to fix those is a question I'm interested in.

    As it stands now though it seems you are rejecting the idea of thought experiments.

    What I'm skeptical of is the incredible utility of 'probability theory' to reasonably settle metaphysical or philosophical questions, and the certainty that such-and-such ultimate initial priors, with such-and-such related probabilities, are clear and warranted. It's a different story when various priors and probabilities are just plain granted right out of the gates.

    Zach said...

    Assigning equal priors to the elements in your sample space can be a useful heuristic, especially when there is no natural ordering defined over the set. But the uniform distribution isn't magic (though it has maximum entropy, which some people treat as magical).

    Zach said...

    Those who advocate Bayesian approaches should peruse (and note I mean peruse, not lightly skim) this guy's work. There is a bit of a Cult of Bayes sweeping through the "skeptical" community right now, and as usual they have no fraking idea what they are talking about for the most part. Richard Carrier is their God. If that isn't scary...

    Victor Reppert said...

    We may need another showdown with Tim McGrew. That one didn't go so well for Carrier last time.

    Papalinton said...

    joesmarts
    "@Papalinton,
    I have no idea what you mean by "group social responsibility obligation standing as an explanation.""



    That's fine. Others will understand what i am getting at.

    joesmarts said...

    @Papalinton,

    You said:
    That's fine. Others will understand what i am getting at.

    What exactly are you getting at? Please enlighten me.

    Zach said...

    The McGrews schooled Carrier and will do it again.