Descartes, though I find his answer silly, deserves substantial accolades for asking the question "What can we be certain of?". Asking the question is usually the hard part.
Hume deserves the title of greatest epistemologist ever, for the first decent answering of the question. His answer, of course, is nothing. The fundamental line in Hume is that belief is EITHER about imaginary items (Math) or not certain.
Many many unsatisfactory responses followed.
And then Bayes asked a better question: How certain should I be of a given belief? And then he answered it. The math is simple.
So what does that lead us to...
It means that we can mathematically decide not only what beliefs to have (based on experience), but also how certain we should be of those beliefs. If we pretend rationality, then belief is no longer a question of preference, but rather of evidence + math. I believe what I want defines irrationality.
There are other paths. One can believe as one wishes, without pretending that the beliefs are subject to reason. Reason is merely one epistemic foundation. And I'm not entirely convinced that it's the best one for most people most of the time. It's certainly the hardest. And for an awful lot of folks...it gives less good answers than social epistemology does 9 times out of 10.
One can trust another person...and accept their conclusions, as they are better than one's own. This, like general social epistemology, will give better results than reason for 50+% of the people.
One can not believe much of anything...and wear beliefs as clothes. Beliefs, fundamentally, are properly about getting what you want/need, not about truth. This is highly under-rated as a method.
If one chains oneself to rationality, one has no room to believe in those things one wants...but rather one is forced to conditional belief in many things. If one believes from some other option, those are highly adaptive strategies as well. One's beliefs, then, are constrained elsewhere.
Happy Easter
Subscribe to:
Post Comments (Atom)
Posts
32 comments:
"The fundamental line in Hume is that belief is EITHER about imaginary items (Math) or not certain."
So it turns out I had a second reason for asking about a link/reference to Hume.
As per Ilkka, I'm not willing to actually read Hume. Immensely inefficient.
I want to verify that statement. Mainly because I've found the class of things for which it is false, and I don't want to go around saying I've got Hume beat without knowing for doubleplus sure that he actually said that.
Kant thought so too...
He started 200 years of Continental philosophy on that wrong foundation.
Kant broke things into
analytic/synthetic
a priori/a posteriori
His giant move to attack Hume was to assert the existence of the "synthetic a priori".
Contrarily, either it is true by definition (MATH), or else it is probably true by observation (Science).
La Wik has a pretty good handle on lots of Hume, and IIRC, it would be in his Enquiry, but I haven't found a quote yet.
I wonder if you have read
Socrates and Hume by Peter Kreeft?
Kreeft seeks to expose problems in Hume. I have not read any of Kreeft's books though I intend to but surely all philosophers including the greatest have their detractors.
If you are basing so much on Hume, then his possible and existing detractions must interest you.
Gyan,
Thanks for the recommendation. I'm looking into it already. At the same time, I spent about 2 decades trying to find a sane response to Hume...and while I've found a few...only Bayes firmly answers his actual questions. And it's not a pretty answer either. It starts with: "You're mostly right, but..."
Kant tries mightily, and fails as mightily.
Reid dismisses well, but faded under the logical weight.
Rand answered correctly about Hume's squirminess, but still fails to address the core question.
1 question: How do you know what you think you know?
Hume's answer: You don't.
Gyan,
FWIW, I began as a Randian objectivist, and have slowly, over 2 decades, become resolutely Bayesian.
My creedo:
If you have beliefs (that are disputed by other reasonable people) that you hold at beyond p>0.9, you have escaped sanity.
Hume probably deserves a Chestertonian answer.
To escape from the logic of skepticism, one needs a bigger space.
Have you read Orthodoxy by Chesterton?
It does not discuss Hume or Kant but has fine discussion of logic and how it could be life-denying and life-affirming.
Gyan,
And that is precisely the issue with Hume. As far as I've been able to tell, there are no responses inside of rationality that address his issues.
He who dreams of both certainty and reason triumphant is doomed by Hume, and then further by Bayes. Pick at most one.
On the other hand, the Randian proto-answer of re-understanding certainty to mean something sane is a great maneuver, which allows us to retain our common language...but we are nonetheless stuck abandoning our historical understanding of certainty.
2 solutions: Bail on reason. Bail on certainty. I haven't read Chesterton...but his solution looks a lot like: abandon the stringent demands of rationality, for something that is more suited to the Human Animal. I prefer to give up silly notions of knowledge, and to rejoice in the flux of uncertainty.
Regarding Descartes, the first question is his place in the Conversation.
Why he was asking the questions he asked?
I thought Descartes was asking his questions in order to defeat skepticism, and demonstrate the existence of God.
And as with most reasoning......he started with a conclusion, and then tried to justify it, rather than starting with information, and trying to reach a conclusion.
Hume, contrarily, seems to have started with a question, and attempted to answer it.
Are you sure Descartes was answering Skepticism?. Was there a lot in his time?.
I thought Hume was the father of Skepticism.
I had a comment regarding your credo but it vanished.
Your credo depends upon the definition of the category of reasonable persons.
So if believers in miracles are unreasonable by definition, then whatever be the evidence, it will not budge you.
Then what kind of empiricism is it?. It looks pretty dogmatic to me.
Gyan,
Humean skepticism re: Miracles is the very simple claim that evidence standards with regards to Miracles should be the same as the evidence standards we use for anything else.
If you see something you don't recognize in the sky, you are insane if you call it a UFO first. You should guess it's 15 other things, including a hallucination, misremembering, or outright deception, before you call it a UFO. All Hume says is that miracles are the same thing.
Frankly hallucination (esp mass hallucination) is a very cheap answer. One may call anything unwelcome to be hallucination but it does not explain much, if anything. It is not as if we have a nice, standard theory of hallucinations.
I dont know if you are aware of the Chestertonian answer: if I believe in a peasant when he claims to witness a murder, then I should believe in him when he claims to witness a ghost or a miracle.
Gyan,
That ignores Bayes Law completely.
I have a prior probability of things happening...
I know that rocks fall DOWN, not up, at 99.9999%.
I know that Trustworthy Tom is almost always truthful at 99.9%
If Tom says he saw a rock fall up...the odds are 1000x higher that Tom is messing with me, than that the rock was going up.
If Tom says that he saw Surly Sam kill a man...and I think Surly Sam has a 90% chance of not killing anyone...but that gives me a 99% odds that Sam killed someone, because Tom is so trustworthy.
It's a really simple math proposition. And you don't have any other choices of what to believe, while being non-insane. To do otherwise, is roughly equivalent to believing that the square root of 25 is 6. Takes a little math, but the answer is not negotiable.
You have to start by believing in Miracles for you to rationally believe that a Miracle happened. It's not intolerance, or pigheadedness that gets the result, either. It's math.
Miracles are ,by definition, improbable events since a miracle is a unaccountable violation of a n apparent regularity.
Your numerical illustration was informative but the frequent recourse to hallucinations, esp mass hallucinations does not cut it and only shows that Humean view is dogmatically opposed to miracles.
Could you care to assign a number to a mass hallucination?
If a miracle was to happen once in the whole lifetime of the universe, a Bayesian would not believe it, but they would be wrong to disbelieve it nevertheless.
So I I submit that the correct position is
1) That we observe certain regularities in the universe.
2) There is nothing to rule out violations of these observed regularities.
That is, have no dogma against miracles.
I use hallucination as an example, because EVEN hallucination is known to happen...measureable, specifiable as occurring 1 time in N. Even the low probability is still a fixed number.
more interestingly, you say:
If a miracle were to happen once in the whole lifetime of the universe, a Bayesian would not believe it, but they would be wrong to disbelieve it.
This is the key point...and you're mistaken. They would be absolutely correct to disbelieve it. No decision they could possibly make with any amount of additional information would have been better. Sure, they happen to be wrong...but it was the best possible decision in those circumstances.
Life is like poker...Sometimes you have a pair of Queens, and you know your opponent has a pair. Pushing is your only good choice when bet into...even though sometimes he has the aces. Under limited information, the best, smartest decision is not always the one that ends up corresponding to underlying reality.
3. Good decisionmaking has no room for fuzziness like "do not rule out...". Nonsense and baloney. Assign prior probabilities and compute. Base your prior probabilities on what you've seen.
If 99% of the all men believe in miracles then what happens to your calculation?
I would futher submit that beliefs are neither algorithmically determined nor determinable. Eg you have believe that murder is wrong or a 3 yr old child may believe that mother must give out cookies fairly to him and his brother. Which algorithm yields these beliefs?
Bayesianism
is useful in hard sciences where you may put numbers to relevant things but over-generalizing it is not useful. Eg
" Surly Sam has a 90% chance of not killing anyone." means precisely what? Is it an empirical statement? An expression of expectation? Whatever it is, it is not quantifiable,
1. When all folks believe in Miracles, messy things happen to my calculation.
It's one of two reasons why I assert that there's an above-infinitessimal probability for the existence of some sort of super-human entity.
On the other hand, I still argue that it's not sane to assign a positive probability...from rational observation of the physical world.
Assigning a non-zero probability to the notion that I could know ANYTHING (good vs. evil, for instance) about said potential super-being or set of super-beings, living or dead, is positively bat-shit insane.
2. Surly Sam has a 90% chance of not killing anyone means precisely: Given EVERYTHING I know, I expect that 9 times out of 10 in circumstances similar to the ones we are in, Sam will not kill anyone...while 1 time out of 10 in a circumstance like this, Sam would. Empirical, testable, quantifiable, scientific, but not tested.
Furthermore...analysis of humans' actual behavior says we really are running Bayesian algorithms for what we believe, most of the time. Do you expect it to rain tomorrow?
I'm betting about 10:1 against, because it's almost May and I'm in Los Angeles today...and it doesn't rain in LA in May more than 1 day in 10.
Bayesianism is the only sane approach to believing ANYTHING. It is hard to practice, and we all get it wrong a lot. But the notion of certainty is laughable, and the notion that anything could happen is ROFL-able. Bayes or pretending. Pick.
I hope you acknowledge that a lot of non-insane people believe in miracles. Even Newton, in all probability.
Also why is your prior not getting affected by arguments such as Argument from Order and Argument from Reason?
Regarding Sam Surly, one can lay bets but it would be a stretch to claim that
"Empirical, testable, quantifiable, scientific"
The word "empirical" does not distinguish experience and experiment but you have added the scientific and quantifiable too.
I submit that no such experiment are doable even in principle to a man since man is not a machine or even an animal.
Gyan,
But as a poker-player, you do (more accurately...I did) precisely that kind of calculation on a regular (hourly) basis. And then you evaluate it scientifically.
It's clearly a empirical scientific approach to the world appropriate to humans.
Poker may be algorithmically reducible; murder is not.
"Poker may be algorithmically reducible; murder is not."
Can you support this? I mean...we know that on average, a low-IQ, low-Conscientiousness, low-Impulse Control (all measurable) person is at least 10x, but probably less than 1000x less likely than someone high on all 3 measures to kill. The murder rate by young urban men is better than 10x (100x?) that of old rural women. Murders are committed mostly among folks who know each other, and very often for romantic reasons (competition or rejection). Closeness to other folks increases the odds of murder, but blood-relatedness decreases it. Testosterone level is (a) correlated with murder, and accurately predictable from voices.
AND we know the average density of murders, as public statistics. We know FAR FAR more than you suggest...though in probabilistic terms.
I thought we were talking about a particular man, Surly Sam, and not a mass of low-IQs.
Indeed we are. But if there are (only) 10 red marbles and 20 blue marbles in a basket...and I'm talking about the next marble to be randomly removed...1 time in 3, it's red.
If people on average commit 6.3 murders per 100,000 people per year, and Surly Sam fits risk factors that put him at 1000x the normal risk (20yo young urban male with low-iq, high testosterone, low impulse control, low conscientiousness, with other young urban males regularly trying to "steal his girl")
...it's not a stretch to argue that there's a 1 in 10 chance that he tries to off the guy sometime over a year. With folks who aren't Sam, the numbers are just lower.
there's a 1 in 10 chance that he tries to off the guy sometime over a year.
Yes, but what does that mean?
I mean I can understand probability in context of physics experiments where you have repeatability. There you get hard numbers.
But, as applied to an individual man, your statement is mystical.
The significance of this number is
1) You can lay a bet on Surly Sam, and you can calculate your chances of winning a bet.
2) You can say that being around Surly Sam is safer than being around someone else or not.
Thats'all. Leaving aside the bet, the information you gain is person A is safer than person B. That A is 2.5 times safer than B is a meaningless thing to say.
On the technical side:
You are reducing the man Surly Sam to his IQ, his testosterone, impulse control, conscientiousness (whatever that may be).
You show a lot of confidence, surprising in one so skeptical and distrusting of authority that such things can be measured in a sensible way.
Gyan,
Why is the communication failing so hard here?
When I say there's a one in 10 chance that Sam will do something...how is that different from saying: there's a 1 in 10 chance that the other guy in my heads-up-poker game will reraise with junk.
Either he does or doesn't, but given what I know right now...I can make likelihood statements just fine. Just like I make likelihood statements about rocks floating in the air (pretty low).
A poker hand could be computable. Human behavior is not.
You have to trust multiple theories of human behavior (a very inexact science)
even to get at what factors are relevant and what is not. You may say IQ is important but church-going is not.
Perhaps if you knew Surly better, even if all his risk factor, his chances of murdering could be negligible even.
Methinks you misunderstand poker.
The guy is holding something. My probability computations in Poker are a computation of what I don't know. If I "know" he has a pair, and compute a 2/13 chance that he can beat my queens... He Either has Kings or he doesn't. Probability is talking about MY lack of knowledge...not about the state of the world...and yet, it's fabulously good, when trained, at determining what to do. When I say he's got a 2/13 chance of having aces or kings...then I'm saying that in 2/13 of situations like this one (however "likeness" is relevant), he'll have aces or kings. If I watch 1000 hands like this one...then I'll have some idea of how good my 2/13 guess is.
Human behavior on an individual basis is very hard...
Human behavior on a statistical basis is less so.
AFAIK, church attendance is not a great proxy for crime-free-ness. However, statistically, IQ is a very strong proxy for being crime free.
Fact is...I might be wrong. But I might be wrong about the poker game too. Being wrong in your predictions, and having a mistaken method are two different things.
You are right. I dont know poker though I play bridge-like games.
But the point is that a card-game is computable. There exist objective algorithms to compute probabilities.
Given rules for a card game, you can compute them even if you may never have played it.
Whereas for human (even animal) behavior, no objective algorithms exist or can exist.
All that exists is guesswork, informed or misinformed .
So the probabilities in this case seem to be in a different class.
I agree that they seem to...but that's not what's going on. I don't play bridge, but I occasionally read the bridge discussion in the daily paper. When you finish your bidding in bridge, you have a guess of what everyone is holding. If you're careful, you could write down what the probabilities are that the other folks hold various hands. But fundamentally...probability is not a claim about what's out there. It's a claim about the limits to our own knowledge.
When Bob opens with 4 No Trump, you probably know that that means something. I personally don't. You're probability distribution for his range of possible hands is different than my probability distribution. However, there is also the actuality of what hand he has. Unfortunately, neither you nor I has access to that hand...we have access to our information, and we can ONLY operate on the probability distributions we have...we don't get to peek at his hand.
You have better guesses than I do in bridge. I have better guesses (I'm assuming) than you do in Poker. And both our guesses are more accurate in cards than in estimating Sam's chances of killing folks. But as always... probability is our estimation of our level of knowledge. The situation is not appreciably different between poker and human-pinball.
I accept that probability is our estimation of our level of knowledge
But humans are not pinballs and algorithms do not and can not exist to predict what they may and may not do.
Post a Comment