Stephen Landsburg, of More Sex is Safer Sex, has a new book out. I haven't read it yet. However, Alex Tabarrok did, and liked it. And it was also reviewed at Slashdot. I've been meaning to comment for days on something he reportedly said in the book.
"If you're objecting to a logical argument, try asking yourself exactly which line in that argument you're objecting to. If you can't identify the locus of your disagreement, you're probably just blathering."
This is a beautiful statement. I've agreed with it for probably 25 years. And it's almost certainly 100% wrong.
By wrong, I mean...the idea works only on a very narrow, and almost never useful, category of questions. AND he's basically wrong in what you should do as a rational agent, though he would be advantaged if you did that.
ISSUE #1:
Most interesting discussions between 2 intelligent people who disagree are not of the form Landsburg seems to be talking about:
Landsburg: W is true,
Landsburg: X is True,
Landsburg: Y is true,
Landsburg: If W,X,Y are true, then Z is true
Landsburg: Therefore Z.
Aretae: I agree with W, X, Y, but not with Z.
Landsburg: But you must agree with Z, if you accept W,X,Y
Rather, most disagreements are:
Landsburg: We should approach Problem A from the WXY perspective
Landsburg: W is true,
Landsburg: X is True,
Landsburg: Y is true,
Landsburg: (If W,X,Y then Z) is true
Landsburg: Therefore Z.
Landsburg: (If Z, then pursue B) is also true
Landsburg: Therefore we should pursue policy B.
Aretae: I agree with WXY, tentatively accept Z, but not Z-->B, and not B.
Landsburg: But you must agree with B, if you accept W,X,Y
In most interesting arguments I see, the disagreement is about which perspective is most useful to approach the problem from...not about the details of the argument. However, it is also true that for most disagreements, there is no formal statement of which approach to use, nor is the assumption Z-->B ever expressed. Therefore, people like Mr. Landsburg are stuck saying that you have accepted the premises but not targetted the point of disagreement, when in reality the point of disagreement is external to the stated premises.
Issue #2:
Verbal fluency of an arguer and truthiness are not highly correlated. I happen to be better at both mining data and at verbal contortions than several relatives of mine. Unfortunately for me, this does not mean that I am always right when arguing with them. Frequently, instead, it means that I'm playing with words in a way that isn't fair, but the unfairness is not obvious to me, and my wordiness is often impenetrable to my conversation partner.
Issue #3:
Landsburg is playing (as per the whole reported philosophy of the book), as an arch-rationalist philosopher. In short, that's bogus. Bayes rules the real world. By Bayes, if I present a clear, concise, simple, ironclad, bulletproof, artistically elegant argument for a position, and you know that not only do you currently disagree, but so too does everyone whose opinion you respect, your proper response is to decrease your odds that you are right from 99.9% down to a measley 98%. In reality, on questions of theory, you should decrease your certainty from about 80%, where your opinions should naturally cap out on contentious questions, down to 75%.
Issue #4:
Most smart new opinions are wrong. Just because it sounds good, and you (and I, and Landsburg) can't poke holes in it, doesn't make it true. Most new ideas fail when tried. Most "good" ideas suck. Most business plans that look good bomb. Most of any genius's new ideas will be wrong. The BEST method known to man of finding the truth is to believe what everyone else believes. It's only weirdos who are willing to be wrong a lot, or who are too stupid to know they'll be wrong a lot, who do anything else...and that applies even to folks like Bill Gates or Warren Buffet, or who got lucky and caught the RIGHT idea (ideas are like flus).
To confound the topic:
On the whole...Landsburg's idea is coming from a long line of philosophers, dating from the Pythagoreans 5 centuries before Christ. My line of argument began properly with Reverend Bayes, of the late 1700s, and was slowly developed since. On that count, he might ought to be taken more seriously than me. On the other hand, my (Bayesian) approach is a successful predictor, and undergirds a great deal of modern science. His method is poor at predicting. Also, his approach is a great deal simpler on most dimensions than is mine, and Occam might tend to prefer his.
Subscribe to:
Post Comments (Atom)
Posts
5 comments:
"The BEST method known to man of finding the truth is to believe what everyone else believes."
Almost no one believes this, and the more expert in epistemology the less likely they are to believe it --> I don't believe it.
So if you believe it you must have used some non-BEST method to arrive at the conclusion... why not use the best method to reject it?
(note: I am open to the idea that many/most/all people act like they believe the BEST method much/most of the time...except that you would need to replace everyone with those locally around me, peer group etc)
Wow, this post and the IP one are awesome! A very productive Aretae day!
I do question your use of "simple". (Eliezer has a great piece on ~computational simplicity, on how to better implement Occam's Razor (on its real meaning). It sounds like you haven't seen it: http://lesswrong.com/lw/jp/occams_razor/ .)
And, that very ORazor-type-thing might strike near the heart of what's bothering Rob here. Your "BEST method" is from a context (set of PERSPECTIVEs as you very well express) of available heuristics in which somehow (unspecified, but kinda makes sense to me) observing others' beliefs and copying them is cognitively very cheap/simple...and so there's nothing else that really competes with that "BEST method" on its own very simple terms (at least in cases where most everyone else is cheaply observed to agree, and at least a bunch of those people have some plausible methods of otherwise having a clue...).
Rob's point is, unsurprisingly, tremendously powerful, and rhetorically elegant.
First, it raises a point I need to write more about...since most independent ideas are wrong, can one believe any of even one's own independent ideas. I can't answer that question right now. I should file that (I believe that many of my contrarian positions are right) under my most absurd belief.
Second...I now need to offer a meager defense.
1. I believe it's fairly demonstrated that mostly, what people DO is believe what they THINK everyone else believes.
2. I believe that Everyone believes that what someone does more accurately reflects the person's beliefs than what they say...notwithstanding that when this logic is being used about the speaker, they claim that exemptions apply.
3. By experiment groups that are diverse get better answers than groups of specialists. Groups of epistemologists are not thus the relevant group to sample. When the relevant group grows to all those who think extensively on the topic "what to believe"...their collective agreement is: believe what the local high status group believes. This reduces to: believe what everyone believes when you include natural human tribal tendencies.
4. Even if you don't like 3, then at least you can't apply any group prior to what epistemologists believe, as there is no consensus. And the method itself fails on epistemology and any other topic with no large majority opinion
5. Most of the above is just confusion. BEST was poorly defined in the original post. A clearer claim is: Assume that the consensus carries a 80-90% prior distribution in favor of the (large) majority opinion. Be stingy in your probabilistic updates to your prior distribution. This avoids entirely the major thrust of the critique.
Andrew,
I almost cited computational simplicity (from OvercomingBias) in the post. I figured it was less complex not to cite it. I seem to have calculated wrong.
well said on computationally simple...but I was hoping to make a second point as well. I also wish to use the open source credo...with enough eyes, all bugs are shallow. The odds of your thinking arriving at the wrong answer due to a personal error in thinking are almost universally MUCH larger than the odds of the consensus being wrong.
The praise is also appreciated.
The very fact we can communicate intelligibly means we agree on a huge majority of basic facts about the world. A consensus, if you will.
But we don't even notice these things - the only kinds of ideas where we explicitly raise the possibility of the consensus being right are a few ideas puttering around on the edges.
So I'm saying you're equivocating. The consensus that has six-nines accuracy and the 'consensus opinions' that actually get talked about are different.
My experience (!= logic) is that these consensuses are sketchy. My guess is that because Landsburg's idea is so old, so traditional, logical argument is preferred, and so "it's the consensus" is (generally) the last refuge of a debater with nothing more concrete to offer.
Post a Comment