Actuarial Outpost
Go Back   Actuarial Outpost > Blogs > sweetiepie
FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions

Fill in a brief DW Simpson Registration Form
to be contacted when new jobs meet your criteria.

Rate this Entry

bayesian humility

Posted 01-23-2009 at 08:59 PM by sweetiepie
Updated 01-24-2009 at 12:40 AM by sweetiepie

It was freshman year, and I had never been kissed, when Laura asked me, "would you ever cheat on your wife?"

I replied "I don't know."


"It's not like I'd cheat on her now-- I mean-- I don't know what I'll be like later."

"So you really think you might?!"

"Well of course I feel like I would never cheat on my wife, it's absurd, but who doesn't feel like it's absurd to cheat on a future spouse? But then they go ahead and do it anyway. So it's like there's no way of knowing." This, of course, was not the explanation she was looking for.

It would be another 4 years or so before I realized that my notions of love, sex, and monogamy are very different than the average *******, and will always be different. Which is good. I feel a little less determined to become something rotten. But the general method, of using priors to reduce the impact of seemingly crucial certainty, has become more and more important to me.

When people argue over a predicate, they on average have a 50/50 chance of being wrong. Of course some people are more right than others, but the average is 50/50. Of course, nobody ever thinks they're wrong 50% of the time even though a good number of people must be, and generally speaking, there's no consensus on who the usually-wrong people are. Everyone just takes it for granted that their confidence is unique. Nobody ever says "We'll we're both sure of ourselves-- I believe I get it and you don't, and you believe that you get it and I don't, so, really, there's a 50% chance I'm wrong here." No. People don't say that. They say "Lets agree to disagree" because they're full of shit.

This goes double for any kind of controversial situation. Morals-- economics-- religion-- art-- recent science-- anything like that. Whatever your position is, there are people who are smarter than you, wiser than you, more concerned than you, less biased than you, that reach opposite conclusion. Doesn't that scare you? It should.

This-- this notion of priors, is also where I derive most of my morals. I don't (can't) really know what's good and what's bad, and so, more often than not, when it comes to dealing with someone, I feel like I have to adopt their value system. Except I can't just use it by itself either-- because peoples values are pretty flawed logically, if not emotionally, so I not only have to adopt their beliefs-- but I have to build on them? It's weird... I guess. But it makes sense. Because whatever I feel by myself is probably wrong, and although whatever they feel is probably wrong too, at least it makes sense to them, or would, if they understood it? Sigh.

It's a matter of selflessness anyway. Not love exactly, although it has the same consequences. It's humility. It means recognizing that you're not special. Not in your beliefs. Not in your desires. Not in your logic. Not in what you deserve. Not in being right.

What I mean to say is this: Nothing about you is unique, unless you have reason to believe, and not in the way that everyone has a reason to believe, you are.
Posted in Uncategorized
Views 29028 Comments 5
« Prev     Main     Next »
Total Comments 5


  1. Old Comment
    Marid Audran's Avatar
    Whatever your position is, there are people who are smarter than you, wiser than you, more concerned than you, less biased than you, that reach opposite conclusion.
    This has occurred to me, too, and I'm never entirely sure how to handle it.

    What is the connection with bayesian reasoning and priors? FYI: I'm still entry-level...have not learned that topic yet. If there's no short answer to my question, then never mind
    Posted 01-28-2009 at 09:39 AM by Marid Audran Marid Audran is offline
  2. Old Comment
    I don't know. I'm not even entry level.

    I will, at least say, that the first scenario is a matter of basic (if messy) math.
    S=you feel sure you will not cheat on your future spouse.
    C=you cheat on your spouse.

    and after substituting--
    and after distributing--
    P(C)=p(C|S)+p(~S)*(p(C|~S)-p(C|S)) *

    and p(C|~S)-p(C|S) is going to be positive, so the function is maximized for a large p(~S)

    And some very rough although theoretically measurable inequalities:

    and of course p(C|~S)<1

    P(C)=p(C|S)+p(~S)*(p(C|~S)-p(C|S)) *
    .4 > p(C|S)+.1(1-p(C|S))
    .3 > .9p(C|S)
    1/3 > P(C|S)
    Posted 01-28-2009 at 01:06 PM by sweetiepie sweetiepie is offline
    Updated 01-28-2009 at 03:05 PM by sweetiepie (bunch of errors)
  3. Old Comment
    Marid Audran's Avatar

    "Not even entry level"? Do you mean, not yet in an actuarial job? If so, then I'm in the same boat. I was using the phrase strictly in reference to my level of knowledge.

    If you mean something else by that phrase, then I wonder what brings you to
    Posted 01-28-2009 at 01:31 PM by Marid Audran Marid Audran is offline
  4. Old Comment
    oh. yeah. i have entry level knowledge too then. i just meant i don't have a job.
    Posted 01-28-2009 at 01:56 PM by sweetiepie sweetiepie is offline
  5. Old Comment
    If you're interested in learning more about why people disagree from a Bayesian standpoint (and the claim is that people lie about their priors), there is some great work being done by a professor out of George Mason University (Robin Hanson) Check this out
    Posted 09-20-2011 at 11:19 PM by mason_anton mason_anton is offline
    Updated 09-20-2011 at 11:20 PM by mason_anton (I messed up the link)

All times are GMT -4. The time now is 05:34 PM.

Powered by vBulletin®
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.
*PLEASE NOTE: Posts are not checked for accuracy, and do not
represent the views of the Actuarial Outpost or its sponsors.
Page generated in 0.20112 seconds with 19 queries