The Sin of Underconfidence

There are three great be­set­ting sins of ra­tio­nal­ists in par­tic­u­lar, and the third of these is un­der­con­fi­dence. Michael Vas­sar reg­u­larly ac­cuses me of this sin, which makes him unique among the en­tire pop­u­la­tion of the Earth.

But he’s ac­tu­ally quite right to worry, and I worry too, and any adept ra­tio­nal­ist will prob­a­bly spend a fair amount of time wory­ing about it. When sub­jects know about a bias or are warned about a bias, over­cor­rec­tion is not un­heard of as an ex­per­i­men­tal re­sult. That’s what makes a lot of cog­ni­tive sub­tasks so trou­ble­some—you know you’re bi­ased but you’re not sure how much, and you don’t know if you’re cor­rect­ing enough—and so per­haps you ought to cor­rect a lit­tle more, and then a lit­tle more, but is that enough? Or have you, per­haps, far over­shot? Are you now per­haps worse off than if you hadn’t tried any cor­rec­tion?

You con­tem­plate the mat­ter, feel­ing more and more lost, and the very task of es­ti­ma­tion be­gins to feel in­creas­ingly fu­tile...

And when it comes to the par­tic­u­lar ques­tions of con­fi­dence, over­con­fi­dence, and un­der­con­fi­dence—be­ing in­ter­preted now in the broader sense, not just cal­ibrated con­fi­dence in­ter­vals—then there is a nat­u­ral ten­dency to cast over­con­fi­dence as the sin of pride, out of that other list which never warned against the im­proper use of hu­mil­ity or the abuse of doubt. To place your­self too high—to over­reach your proper place—to think too much of your­self—to put your­self for­ward—to put down your fel­lows by im­plicit com­par­i­son—and the con­se­quences of hu­mil­i­a­tion and be­ing cast down, per­haps pub­li­cly—are these not loathe­some and fear­some things?

To be too mod­est—seems lighter by com­par­i­son; it wouldn’t be so hu­mil­i­at­ing to be called on it pub­li­cly, in­deed, find­ing out that you’re bet­ter than you imag­ined might come as a warm sur­prise; and to put your­self down, and oth­ers im­plic­itly above, has a pos­i­tive tinge of nice­ness about it, it’s the sort of thing that Gan­dalf would do.

So if you have learned a thou­sand ways that hu­mans fall into er­ror and read a hun­dred ex­per­i­men­tal re­sults in which anony­mous sub­jects are hu­mil­i­ated of their over­con­fi­dence—heck, even if you’ve just read a cou­ple of dozen—and you don’t know ex­actly how over­con­fi­dent you are—then yes, you might gen­uinely be in dan­ger of nudg­ing your­self a step too far down.

I have no perfect for­mula to give you that will coun­ter­act this. But I have an item or two of ad­vice.

What is the dan­ger of un­der­con­fi­dence?

Pass­ing up op­por­tu­ni­ties. Not do­ing things you could have done, but didn’t try (hard enough).

So here’s a first item of ad­vice: If there’s a way to find out how good you are, the thing to do is test it. A hy­poth­e­sis af­fords test­ing; hy­pothe­ses about your own abil­ities like­wise. Once upon a time it seemed to me that I ought to be able to win at the AI-Box Ex­per­i­ment; and it seemed like a very doubt­ful and hubris­tic thought; so I tested it. Then later it seemed to me that I might be able to win even with large sums of money at stake, and I tested that, but I only won 1 time out of 3. So that was the limit of my abil­ity at that time, and it was not nec­es­sary to ar­gue my­self up­ward or down­ward, be­cause I could just test it.

One of the chief ways that smart peo­ple end up stupid, is by get­ting so used to win­ning that they stick to places where they know they can win—mean­ing that they never stretch their abil­ities, they never try any­thing difficult.

It is said that this is linked to defin­ing your­self in terms of your “in­tel­li­gence” rather than “effort”, be­cause then win­ning eas­ily is a sign of your “in­tel­li­gence”, where failing on a hard prob­lem could have been in­ter­preted in terms of a good effort.

Now, I am not quite sure this is how an adept ra­tio­nal­ist should think about these things: ra­tio­nal­ity is sys­tem­atized win­ning and try­ing to try seems like a path to failure. I would put it this way: A hy­poth­e­sis af­fords test­ing! If you don’t know whether you’ll win on a hard prob­lem—then challenge your ra­tio­nal­ity to dis­cover your cur­rent level. I don’t usu­ally hold with con­grat­u­lat­ing your­self on hav­ing tried—it seems like a bad men­tal habit to me—but surely not try­ing is even worse. If you have cul­ti­vated a gen­eral habit of con­fronting challenges, and won on at least some of them, then you may, per­haps, think to your­self “I did keep up my habit of con­fronting challenges, and will do so next time as well”. You may also think to your­self “I have gained valuable in­for­ma­tion about my cur­rent level and where I need im­prove­ment”, so long as you prop­erly com­plete the thought, “I shall try not to gain this same valuable in­for­ma­tion again next time”.

If you win ev­ery time, it means you aren’t stretch­ing your­self enough. But you should se­ri­ously try to win ev­ery time. And if you con­sole your­self too much for failure, you lose your win­ning spirit and be­come a scrub.

When I try to imag­ine what a fic­tional mas­ter of the Com­pet­i­tive Con­spir­acy would say about this, it comes out some­thing like: “It’s not okay to lose. But the hurt of los­ing is not some­thing so scary that you should flee the challenge for fear of it. It’s not so scary that you have to care­fully avoid feel­ing it, or re­fuse to ad­mit that you lost and lost hard. Los­ing is sup­posed to hurt. If it didn’t hurt you wouldn’t be a Com­peti­tor. And there’s no Com­peti­tor who never knows the pain of los­ing. Now get out there and win.”

Cul­ti­vate a habit of con­fronting challenges—not the ones that can kill you out­right, per­haps, but per­haps ones that can po­ten­tially hu­mil­i­ate you. I re­cently read of a cer­tain the­ist that he had defeated Christo­pher Hitchens in a de­bate (severely so; this was said by athe­ists). And so I wrote at once to the Blog­ging­heads folks and asked if they could ar­range a de­bate. This seemed like some­one I wanted to test my­self against. Also, it was said by them that Christo­pher Hitchens should have watched the the­ist’s ear­lier de­bates and been pre­pared, so I de­cided not to do that, be­cause I think I should be able to han­dle damn near any­thing on the fly, and I de­sire to learn whether this thought is cor­rect; and I am will­ing to risk pub­lic hu­mil­i­a­tion to find out. Note that this is not self-hand­i­cap­ping in the clas­sic sense—if the de­bate is in­deed ar­ranged (I haven’t yet heard back), and I do not pre­pare, and I fail, then I do lose those stakes of my­self that I have put up; I gain in­for­ma­tion about my limits; I have not given my­self any­thing I con­sider an ex­cuse for los­ing.

Of course this is only a way to think when you re­ally are con­fronting a challenge just to test your­self, and not be­cause you have to win at any cost. In that case you make ev­ery­thing as easy for your­self as pos­si­ble. To do oth­er­wise would be spec­tac­u­lar over­con­fi­dence, even if you’re play­ing tic-tac-toe against a three-year-old.

A sub­tler form of un­der­con­fi­dence is los­ing your for­ward mo­men­tum—amid all the things you re­al­ize that hu­mans are do­ing wrong, that you used to be do­ing wrong, of which you are prob­a­bly still do­ing some wrong. You be­come timid; you ques­tion your­self but don’t an­swer the self-ques­tions and move on; when you hy­poth­e­size your own in­abil­ity you do not put that hy­poth­e­sis to the test.

Per­haps with­out there ever be­ing a wa­ter­shed mo­ment when you de­liber­ately, self-visi­bly de­cide not to try at some par­tic­u­lar test… you just.… slow..… down......

It doesn’t seem worth­while any more, to go on try­ing to fix one thing when there are a dozen other things that will still be wrong...

There’s not enough hope of triumph to in­spire you to try hard...

When you con­sider do­ing any new thing, a dozen ques­tions about your abil­ity at once leap into your mind, and it does not oc­cur to you that you could an­swer the ques­tions by test­ing your­self...

And hav­ing read so much wis­dom of hu­man flaws, it seems that the course of wis­dom is ever doubt­ing (never re­solv­ing doubts), ever the hu­mil­ity of re­fusal (never the hu­mil­ity of prepa­ra­tion), and just gen­er­ally, that it is wise to say worse and worse things about hu­man abil­ities, to pass into feel-good feel-bad cyn­i­cism.

And so my last piece of ad­vice is an­other per­spec­tive from which to view the prob­lem—by which to judge any po­ten­tial habit of thought you might adopt—and that is to ask:

Does this way of think­ing make me stronger, or weaker? Really truly?

I have pre­vi­ously spo­ken of the dan­ger of rea­son­able­ness—the rea­son­able-sound­ing ar­gu­ment that we should two-box on New­comb’s prob­lem, the rea­son­able-sound­ing ar­gu­ment that we can’t know any­thing due to the prob­lem of in­duc­tion, the rea­son­able-sound­ing ar­gu­ment that we will be bet­ter off on av­er­age if we always adopt the ma­jor­ity be­lief, and other such im­ped­i­ments to the Way. “Does it win?” is one ques­tion you could ask to get an al­ter­nate per­spec­tive. Another, slightly differ­ent per­spec­tive is to ask, “Does this way of think­ing make me stronger, or weaker?” Does con­stantly re­mind­ing your­self to doubt ev­ery­thing make you stronger, or weaker? Does never re­solv­ing or de­creas­ing those doubts make you stronger, or weaker? Does un­der­go­ing a de­liber­ate crisis of faith in the face of un­cer­tainty make you stronger, or weaker? Does an­swer­ing ev­ery ob­jec­tion with a hum­ble con­fes­sion of you fal­li­bil­ity make you stronger, or weaker?

Are your cur­rent at­tempts to com­pen­sate for pos­si­ble over­con­fi­dence mak­ing you stronger, or weaker? Hint: If you are tak­ing more pre­cau­tions, more scrupu­lously try­ing to test your­self, ask­ing friends for ad­vice, work­ing your way up to big things in­cre­men­tally, or still failing some­times but less of­ten then you used to, you are prob­a­bly get­ting stronger. If you are never failing, avoid­ing challenges, and feel­ing gen­er­ally hope­less and dis­pir­ited, you are prob­a­bly get­ting weaker.

I learned the first form of this rule at a very early age, when I was prac­tic­ing for a cer­tain math test, and found that my score was go­ing down with each prac­tice test I took, and no­ticed go­ing over the an­swer sheet that I had been pen­cilling in the cor­rect an­swers and eras­ing them. So I said to my­self, “All right, this time I’m go­ing to use the Force and act on in­stinct”, and my score shot up to above what it had been in the be­gin­ning, and on the real test it was higher still. So that was how I learned that doubt­ing your­self does not always make you stronger—es­pe­cially if it in­terferes with your abil­ity to be moved by good in­for­ma­tion, such as your math in­tu­itions. (But I did need the test to tell me this!)

Un­der­con­fi­dence is not a unique sin of ra­tio­nal­ists alone. But it is a par­tic­u­lar dan­ger into which the at­tempt to be ra­tio­nal can lead you. And it is a stop­ping mis­take—an er­ror which pre­vents you from gain­ing that fur­ther ex­pe­rience which would cor­rect the er­ror.

Be­cause un­der­con­fi­dence ac­tu­ally does seem quite com­mon among as­piring ra­tio­nal­ists who I meet—though rather less com­mon among ra­tio­nal­ists who have be­come fa­mous role mod­els)—I would in­deed name it third among the three be­set­ting sins of ra­tio­nal­ists.