When I was young, I read pop­u­lar physics books such as Richard Feyn­man’s QED: The Strange The­ory of Light and Mat­ter. I knew that light was waves, sound was waves, mat­ter was waves. I took pride in my sci­en­tific liter­acy, when I was nine years old.

When I was older, and I be­gan to read the Feyn­man Lec­tures on Physics, I ran across a gem called “the wave equa­tion.” I could fol­low the equa­tion’s deriva­tion, but, look­ing back, I couldn’t see its truth at a glance. So I thought about the wave equa­tion for three days, on and off, un­til I saw that it was em­bar­rass­ingly ob­vi­ous. And when I fi­nally un­der­stood, I re­al­ized that the whole time I had ac­cepted the hon­est as­surance of physi­cists that light was waves, sound was waves, mat­ter was waves, I had not had the vaguest idea of what the word “wave” meant to a physi­cist.

There is an in­stinc­tive ten­dency to think that if a physi­cist says “light is made of waves,” and the teacher says “What is light made of?” and the stu­dent says “Waves!”, then the stu­dent has made a true state­ment. That’s only fair, right? We ac­cept “waves” as a cor­rect an­swer from the physi­cist; wouldn’t it be un­fair to re­ject it from the stu­dent? Surely, the an­swer “Waves!” is ei­ther true or false, right?

Which is one more bad habit to un­learn from school. Words do not have in­trin­sic defi­ni­tions. If I hear the syl­la­bles “bea-ver” and think of a large ro­dent, that is a fact about my own state of mind, not a fact about the syl­la­bles “bea-ver.” The se­quence of syl­la­bles “made of waves” (or “be­cause of heat con­duc­tion”) is not a hy­poth­e­sis ; it is a pat­tern of vibra­tions trav­el­ing through the air, or ink on pa­per. It can as­so­ci­ate to a hy­poth­e­sis in some­one’s mind, but it is not, of it­self, right or wrong. But in school, the teacher hands you a gold star for say­ing “made of waves,” which must be the cor­rect an­swer be­cause the teacher heard a physi­cist emit the same sound-vibra­tions. Since ver­bal be­hav­ior (spo­ken or writ­ten) is what gets the gold star, stu­dents be­gin to think that ver­bal be­hav­ior has a truth-value. After all, ei­ther light is made of waves, or it isn’t, right?

And this leads into an even worse habit. Sup­pose the teacher asks you why the far side of a metal plate feels warmer than the side next to the ra­di­a­tor. If you say “I don’t know,” you have no chance of get­ting a gold star—it won’t even count as class par­ti­ci­pa­tion. But, dur­ing the cur­rent semester, this teacher has used the phrases “be­cause of heat con­vec­tion,” “be­cause of heat con­duc­tion,” and “be­cause of ra­di­ant heat.” One of these is prob­a­bly what the teacher wants. You say, “Eh, maybe be­cause of heat con­duc­tion?”

This is not a hy­poth­e­sis about the metal plate. This is not even a proper be­lief. It is an at­tempt to guess the teacher’s pass­word.

Even vi­su­al­iz­ing the sym­bols of the diffu­sion equa­tion (the math gov­ern­ing heat con­duc­tion) doesn’t mean you’ve formed a hy­poth­e­sis about the metal plate. This is not school; we are not test­ing your mem­ory to see if you can write down the diffu­sion equa­tion. This is Bayescraft; we are scor­ing your an­ti­ci­pa­tions of ex­pe­rience. If you use the diffu­sion equa­tion, by mea­sur­ing a few points with a ther­mome­ter and then try­ing to pre­dict what the ther­mome­ter will say on the next mea­sure­ment, then it is definitely con­nected to ex­pe­rience. Even if the stu­dent just vi­su­al­izes some­thing flow­ing, and there­fore holds a match near the cooler side of the plate to try to mea­sure where the heat goes, then this men­tal image of flow­ing-ness con­nects to ex­pe­rience; it con­trols an­ti­ci­pa­tion.

If you aren’t us­ing the diffu­sion equa­tion—putting in num­bers and get­ting out re­sults that con­trol your an­ti­ci­pa­tion of par­tic­u­lar ex­pe­riences—then the con­nec­tion be­tween map and ter­ri­tory is sev­ered as though by a knife. What re­mains is not a be­lief, but a ver­bal be­hav­ior.

In the school sys­tem, it’s all about ver­bal be­hav­ior, whether writ­ten on pa­per or spo­ken aloud. Ver­bal be­hav­ior gets you a gold star or a failing grade. Part of un­learn­ing this bad habit is be­com­ing con­sciously aware of the differ­ence be­tween an ex­pla­na­tion and a pass­word.

Does this seem too harsh? When you’re faced by a con­fus­ing metal plate, can’t “heat con­duc­tion?” be a first step to­ward find­ing the an­swer? Maybe, but only if you don’t fall into the trap of think­ing that you are look­ing for a pass­word. What if there is no teacher to tell you that you failed? Then you may think that “Light is wakalixes” is a good ex­pla­na­tion, that “wakalixes” is the cor­rect pass­word. It hap­pened to me when I was nine years old—not be­cause I was stupid, but be­cause this is what hap­pens by de­fault. This is how hu­man be­ings think, un­less they are trained not to fall into the trap. Hu­man­ity stayed stuck in holes like this for thou­sands of years.

Maybe, if we drill stu­dents that words don’t count, only an­ti­ci­pa­tion-con­trol­lers, the stu­dent will not get stuck on “Heat con­duc­tion? No? Maybe heat con­vec­tion? That’s not it ei­ther?” Maybe then, think­ing the phrase “heat con­duc­tion” will lead onto a gen­uinely helpful path, like:

• “Heat con­duc­tion?”

• But that’s only a phrase—what does it mean?

• The diffu­sion equa­tion?

• But those are only sym­bols—how do I ap­ply them?

• What does ap­ply­ing the diffu­sion equa­tion lead me to an­ti­ci­pate?

• It sure doesn’t lead me to an­ti­ci­pate that the side of a metal plate farther away from a ra­di­a­tor would feel warmer.

• I no­tice that I am con­fused. Maybe the near side just feels cooler, be­cause it’s made of more in­su­la­tive ma­te­rial and trans­fers less heat to my hand? I’ll try mea­sur­ing the tem­per­a­ture . . .

• Okay, that wasn’t it. Can I try to ver­ify whether the diffu­sion equa­tion holds true of this metal plate, at all? Is heat flow­ing the way it usu­ally does, or is some­thing else go­ing on?

• I could hold a match to the plate and try to mea­sure how heat spreads over time . . .

If we are not strict about “Eh, maybe be­cause of heat con­duc­tion?” be­ing a fake ex­pla­na­tion, the stu­dent will very prob­a­bly get stuck on some wakalixes-pass­word. This hap­pens by de­fault: it hap­pened to the whole hu­man species for thou­sands of years.

• “Then you may think that “Light is ar­gle­bar­gle” is a good ex­pla­na­tion, that “ar­gle­bar­gle” is the cor­rect pass­word. It hap­pened to me when I was nine years old—not be­cause I was stupid, but be­cause this is what hap­pens by de­fault. This is how hu­man be­ings think, un­less they are trained not to fall into the trap. Hu­man­ity stayed stuck in holes like this for thou­sands of years.”

Okay, but there’s one in­no­cent in­ter­pre­ta­tion even here. Peo­ple learn lan­guage, and when we learn lan­guage we copy the ver­bal be­hav­ior of other peo­ple. Maybe “ar­gle­bar­gle” is a syn­onym for light in some lan­guage, or maybe it’s a su­per­cat­e­gory of light (a cat­e­gory that in­cludes light among other things). Maybe the teacher is still in the pro­cess of ex­plain­ing to us what ar­gle­bar­gle means and the first step is to say that light is ar­gle­bar­gle—later on the teacher will tell us what else is ar­gle­bar­gle so that we will grad­u­ally build a good con­cept of it but ini­tially we need to re­tain the point that light is ar­gle­bar­gle while not yet know­ing what ar­gle­bar­gle is, be­cause this is a step in learn­ing what ar­gle­bar­gle is. In that case, we’re learn­ing new lan­guage when we learn that “light is ar­gle­bar­gle”. That’s in­no­cent, it’s not a mis­take.

This sug­gests that the er­ror may not be learn­ing the teacher’s pass­words per se, as such, but learn­ing the teacher’s pass­words when we should ideally be learn­ing some­thing else. The con­text mat­ters.

But the stu­dent is the stu­dent and there­fore ig­no­rant by as­sump­tion, so it may in many cases be too much to ex­pect the stu­dent to know when it is time to learn the pass­words and when it is time to learn some­thing else. If the stu­dent ex­pe­riences aca­demic suc­cess only from learn­ing the pass­words, then it may be that the stu­dent is not at fault, it’s the cur­ricu­lum that is at fault—the teacher.

So the right recom­men­da­tion may not be to tell the stu­dent to stop learn­ing pass­words. Pass­words are a le­gi­t­i­mate thing to learn, some­times, and the stu­dent, be­ing a stu­dent, doesn’t know ahead of time which times. The right recom­men­da­tion may be to ad­just the cur­ricu­lum so that only the right kind of learn­ing yields aca­demic suc­cess.

That’s prob­a­bly not easy.

• This re­minds me of my own ex­pe­rience as a stu­dent who loved chem­istry. We were told a se­ries of use­ful un­truths about what mat­ter is as we went through the sys­tem.

Molecules and atoms were like billiard balls.

No, that was an ap­prox­i­ma­tion—atoms are made of nu­clei and elec­trons which can be vi­su­al­ised as lit­tle plane­tary sys­tems.

No, that was an ap­prox­i­ma­tion—elec­trons, pro­tons, neu­trons are more use­fully con­sid­ered as prob­a­bil­ity func­tions.

I didn’t do sci­ence at uni­ver­sity level, so I never got to the next level, but quan­tum the­ory was wait­ing for me there.

I did start an elec­tronic en­g­ineer­ing course, and there we learned an­other use­ful half-truth—the equa­tions that de­scribe the be­havi­our of a tran­sis­tor. Only they don’t. They de­scribe a man­age­able func­tion which is some­thing like the be­havi­our of a tran­sis­tor—the real-world be­havi­our is non-lin­ear and dis­con­tin­u­ous (truly hor­rible—I didn’t finish the course...).

All of these use­ful un­truths are like pass­words—they al­low us to re­li­ably ac­com­plish things in the world, but they do not give us real power over or un­der­stand­ing of the do­main they ad­dress. Nev­er­the­less, it would be hard to do with­out them.

• Given how much got ac­com­plished with prior mod­els of the atom, I wouldn’t say these are nec­es­sar­ily good ex­am­ples of pass­words. They also weren’t ap­prox­i­ma­tions so much as older mod­els. It’s sort of like learn­ing the geo­cen­tric model first, and then later up­dat­ing to the he­lio­cen­tric model, and then fi­nally learn­ing that the sun ac­tu­ally re­volves around the cen­ter of the uni­verse as well.

I’m hon­estly a bit puz­zled as to why we in­sist on teach­ing so many older mod­els in sci­ence, with­out ap­pro­pri­ately la­bel­ing them. Per­haps the math is eas­ier to learn, and per­haps it’s just much eas­ier to teach the mod­els you grew up with origi­nally.

• Ac­tu­ally, when I learned these I learned them all at once, with the “older model” tag at­tached to them, and then I was given a “cur­rent model” that I was told that I wouldn’t un­der­stand yet, and so we worked with the plane­tary sys­tem thing.

That’s progress?

(You are try­ing to sub­mit too fast. try again in 711 mil­lisec­onds. This web­site re­ally val­ues ac­cu­racy.)

• (You are try­ing to sub­mit too fast. try again in 711 mil­lisec­onds. This web­site re­ally val­ues ac­cu­racy.)

• A pass­word is a type of (usu­ally par­tial) ex­ten­sive defi­ni­tion (a list of the mem­bers of a set). What we want to teach is in­ten­sive defi­ni­tions (the defin­ing char­ac­ter­is­tics of sets). An ex­ten­sive defi­ni­tion is not en­tirely use­less as a learn­ing aid, be­cause an stu­dent could, in the­ory, work out the re­lated in­ten­sive defi­ni­tion. Un­for­tu­nately, this is ex­traor­di­nar­ily difficult when the defi­ni­tions re­late to wave dy­nam­ics, for ex­am­ple.

A pass­word is an ex­ten­sive defi­ni­tion be­ing treated like the ob­jec­tive—a float­ing defi­ni­tion, where the in­ten­sive defi­ni­tion is no longer be­ing sought. Ex­ten­sive defi­ni­tions do not re­strict an­ti­ci­pa­tion, and so can only ever be a step to­wards teach­ing in­ten­sive defi­ni­tions. Learn­ing pass­words is wrong by defi­ni­tion—if it’s use­ful, it’s no longer a pass­word.

In the case of a stu­dent learn­ing a for­eign lan­guage, pro­vid­ing ex­ten­sive defi­ni­tions is very use­ful be­cause there is no differ­ence be­tween say­ing “the set {words-for-ap­ple} in­cludes apfel” and “apfel means ap­ple, and the set {words-for-ap­ple} con­tains all words mean­ing ap­ple” - the in­ten­sive and ex­ten­sive defini­tons are pro­vided to­gether, as­sum­ing the stu­dent knows what what an ap­ple is.

• 90% of any­thing is crud, in­clud­ing schools and per­haps the schol­arly mo­ti­va­tion of most stu­dents, but the bet­ter schools don’t teach the bet­ter stu­dents by rote learn­ing.

The magic words, as you point out at the bot­tom of your es­say, are helpful for get­ting one’s thoughts into the right part of sci­ence. Most peo­ple would have a train of thought that is not quite as re­flec­tive as what you de­scribed, some­thing a bit more con­fused like:

• “Heat con­duc­tion?”

• Some­thing to do with heat spread­ing out from hot to cold.

• This is strange. The near side should be hot­ter.

• Well, maybe the plate’s made of some weird ma­te­rial or some­thing.

The teacher is pos­ing a trick ques­tion, by re­quiring thought out­side physics and in the area of teacherly psy­chol­ogy. This is a good way of re­mind­ing the stu­dents that all ar­eas of sci­ence are con­nected. But good sci­ence works by cre­at­ing mod­els which are sim­plifi­ca­tions of the world. In a physics class, the model fo­cuses on physics and not hu­man psy­chol­ogy. Too many trick ques­tions, and stu­dents will never learn the laws of ther­mo­dy­nam­ics, as they spend their brain­power try­ing to out­wit the teacher.

• Well, maybe the plate’s made of some weird ma­te­rial or some­thing.

My first re­sponse would be “meta­ma­te­ri­als”. Then it would be an ex­tremely ex­cited feel­ing, be­cause the teacher just Vio­lated Ther­mo­dy­nam­ics™. Then it would be con­fu­sion, and I’d stick up my hand and say “the hot air was blown to­wards the far side” or some­thing.

We don’t tend to ques­tion the premise when the Trusted Author­ity Figure is ask­ing us ques­tions, be­cause the chance of that is very un­likely. The chance of it ac­tu­ally be­ing heat con­duc­tion in some way is higher than the teacher fak­ing the situ­a­tion. The chance of me be­ing com­pletely wrong about Physics is higher than the teacher ly­ing, and “heat con­duc­tion” is my eas­iest way out, whilst sav­ing face.

• You touch into a topic that is all too com­mon in the first half, and that is the prob­lem of defi­ni­tions. It is not un­usual to find peo­ple hav­ing an ar­gu­ment over some­thing, with­out first do­ing a clear defi­ni­tion of the ques­tion.

For ex­am­ple, what is in­tel­li­gence, and what is con­science? There are lots of dis­cus­sions about the pos­si­bil­ity or im­pos­si­bil­ity to cre­ate ar­tifi­cial ver­sions of these, with­out first hav­ing a com­mon defi­ni­tion. Such an ar­gu­ment is al­most a waste of time, ex­cept for the situ­a­tion where it may lead to bet­ter un­der­stand­ing of defi­ni­tion.

Another fa­mous ex­am­ple is the ques­tion of the mean­ing of life. Define life first, and I think the ques­tion will be eas­ier.

Yet an­other ex­am­ple is the Chi­nese Room (a hu­man not un­der­stand­ing Chi­nese will take a chi­nese ques­tion and use a set of writ­ten rules to pro­duce an an­swer in chi­nese). The ques­tion is then, can this con­struc­tion be con­sid­ered to un­der­stand chi­nese? This is a very hot dis­cus­sion, but I see no at­tempt to first define what you mean with “un­der­stand chi­nese”.

I can’t help think­ing about the the com­puter that pro­duced the an­swer 42, be­cause the ques­tion wasn’t ex­act enough. While quite funny the first time I read it, peo­ple still ask ques­tions that way.

• Ex­cel­lent post Eliezer.

• Bravo.

• I think of poli­tics here: try­ing to guess the elec­torate’s pass­word. An elected offi­cial is re­warded for giv­ing the an­swer that will get him the most votes. If that hap­pens to be a well-con­ceived policy de­ci­sion, that is a happy co­in­ci­dence.

The hard part about teach­ing stu­dents not to guess the pass­word is teach­ing teach­ers not to ac­cept pass­word guesses, and how to dis­t­in­guish ver­bal be­hav­ior from thought.

• Un­less my mem­ory de­cives me very badly, Richard Feyn­man has made a closely re­lated point in the open­ing chap­ter of “Surely you’re jok­ing Mr.Feyn­man” (when his father told him that things move be­cause of the sun, while he learns in school that things move be­cause of “en­ergy”).

• There’s a lovely bit in Egan’s Di­as­pora show­ing the view­point char­ac­ter un­der­stand­ing a con­cept from physics by ap­ply­ing it in var­i­ous con­texts.

More gen­er­ally, I don’t know if much is known about how peo­ple get from in­put to un­der­stand­ing.

Pos­si­bly of in­ter­est: Math­se­man­tics, which grew out of a pro­ject to find em­ploy­ees who un­der­stood what num­bers mean. The book (about a ques­tion­aire for the pur­pose) is very in­ter­est­ing, the ar­ti­cles listed mostly look minor ex­cept for the one about grokdu­el­ling (you win if you un­der­stand the other side bet­ter), and they’re look­ing for re­search ideas.

• Okay, but there’s one in­no­cent in­ter­pre­ta­tion even here. Peo­ple learn lan­guage, and when we learn lan­guage we copy the ver­bal be­hav­ior of other peo­ple.

This is not in­no­cent! Just be­cause ev­ery­one does it, doesn’t make it okay. You can’t trust your in­stincts! Hu­man­ity stayed stuck in a thirty-thou­sand-year trap be­cause of this—be­cause we took ev­ery­thing con­fus­ing, and found we could imi­tate ver­bal be­hav­iors about it just as well as we could imi­tate ver­bal be­hav­iors about any­thing else.

Too many trick ques­tions, and stu­dents will never learn the laws of ther­mo­dy­nam­ics, as they spend their brain­power try­ing to out­wit the teacher.

Need they have such limited brain­power? Next time, use a spe­cial ma­te­rial with the in­su­la­tor nearer the ra­di­a­tor.

You touch into a topic that is all too com­mon in the first half, and that is the prob­lem of defi­ni­tions.

What do you mean by “first half”? But in any case, yes.

Richard Feyn­man has made a closely re­lated point in the open­ing chap­ter of “Surely you’re jok­ing Mr.Feyn­man”

Good point. I changed “ar­gle­bar­gle” to “wakalixes” in the text.

• “Okay, but there’s one in­no­cent in­ter­pre­ta­tion even here. Peo­ple learn lan­guage, and when we learn lan­guage we copy the ver­bal be­hav­ior of other peo­ple.

This is not in­no­cent! Just be­cause ev­ery­one does it, doesn’t make it okay. You can’t trust your in­stincts! ….”

We have to be care­ful about the no­tion of “can’t trust your in­stincts.” There is a fun­da­men­tal pro­cess of lan­guage aqui­si­tion I think we’d be hard pressed to deny. This is similar to learn­ing about ther­mo­dy­nam­ics. The first word we know is “mommy” (of­ten, +/​-). At this stage, and for a long time, we don’t un­der­stand what it re­ally means, this con­cept of “mother” (and many of us never will). How is that differ­ent than “con­duc­tion”? You have to start some­where, and nec­es­sar­ily small.

On the other hand, the no­tion in teach­ing that you have suc­ceeded if they know the pass­words is in­sidious and lazy. The big pass­word fetish nowa­days is stan­dard­ized tests: if the stu­dents test well, then you’ve suc­ceeded in teach­ing them well. It’s pass­words with no fol­low-up, no syn­the­sis.

• “Need they have such limited brain­power?” -per­haps. I’m in­ter­ested in our en­ergy/​en­vi­ron­men­tally con­strained limits, past, pre­sent, and fu­ture, on our abil­ity to model re­al­ity, etc. We do ap­par­ently have limited brain­power, and sloppy routes to the ac­cu­mu­la­tion of knowl­edge. It’s worth ex­am­in­ing the most effi­cient paths (even if they may have ir­ra­tional el­e­ments) to the ac­cu­mu­la­tion of best mod­els of re­al­ity.

• On the other hand, if tech­ers make first-graders in­tu­itively pre­dict what 2+2 equals, they’ll never get around to mak­ing them an­ti­ci­pate who it was that dis­cov­ered Amer­ica. There are a lot of pass­words. Plus, the kids just won’t work that hard.

• “This is not in­no­cent! Just be­cause ev­ery­one does it, doesn’t make it okay.”

That’s not what I said. I said it is okay, di­rectly. Not be­cause ev­ery­one does it, but be­cause it is a le­gi­t­i­mate as­pect of some kinds of valuable learn­ing. But there’s prob­a­bly a mis­com­mu­ni­ca­tion here about ex­actly what “it” refers to. My point was to try to draw a dis­tinc­tion be­tween a cou­ple of things which you seem to have missed in your re­ply.

• Con­stant, just be­cause some­thing is part of how hu­mans learn lan­guage, does not make it okay. We know that some­thing is ar­gle­bar­gle with­out know­ing what ar­gle­bar­gle is, but, this is not la­beled as “a hint about some­one else’s men­tal as­so­ci­a­tions to a word”, it is la­beled as know­ing that some­thing is ar­gle­bar­gle. That’s the er­ror, right there: a float­ing, non-an­ti­ci­pa­tion-con­trol­ling be­lief that feels like or­di­nary knowl­edge, and is not la­beled as a hint about some­one else’s word-as­so­ci­a­tions. Just be­cause this is part of how hu­man minds learn lan­guage, still doesn’t make it okay. The hu­man mind has a crappy de­sign. That’s what blogs like Over­com­ing Bias are for.

• “Con­stant, just be­cause some­thing is part of how hu­mans learn lan­guage, does not make it okay.”

Learn­ing the pass­words is okay in the con­text of learn­ing lan­guage (I think). Take it away and you im­pede the abil­ity to learn lan­guage. More gen­er­ally, I think that learn­ing empty slo­gans isn’t wrong, it’s merely in­com­plete. Part of learn­ing is un­in­tel­li­gent ap­ing, where we first learn to go through some of the mo­tions with­out nec­es­sar­ily un­der­stand­ing them. You have to start some­where. Un­der­stand­ing is com­plex and so nec­es­sar­ily re­quires pas­sage through stages which do not each one con­sti­tute un­der­stand­ing, just as an office build­ing as it is be­ing built passes through many stages where what it is is just a use­less struc­ture that serves no im­me­di­ate pur­pose. What is bad is not learn­ing the pass­words, it’s failing to learn more be­yond the pass­words. Similarly, what is bad is not lay­ing a foun­da­tion un­der a build­ing, what is bad is aban­don­ing the con­struc­tion of the build­ing af­ter the foun­da­tion is laid but be­fore the build­ing is com­plete.

“a float­ing, non-an­ti­ci­pa­tion-con­trol­ling be­lief that feels like or­di­nary knowl­edge, and is not la­beled as a hint about some­one else’s word-as­so­ci­a­tions”

It is al­most nec­es­sar­ily the case that peo­ple who do not un­der­stand yet (e.g., stu­dents who are in the pro­cess of learn­ing), do not yet have the un­der­stand­ing that it takes to un­der­stand that they do not un­der­stand yet. Know­ing that you did not know is more eas­ily done in hind­sight once you know, than while you still do not know.

So the bur­den is prob­a­bly bet­ter placed on the teacher to en­sure that the stu­dents’ ed­u­ca­tion is not left in­com­plete.

I my­self find it hard to see what the gaps are in my knowl­edge un­less that knowl­edge is tested. So I think a good place to start in fix­ing what­ever prob­lem it is that you see, is in the test­ing of the stu­dents.

• Eliezer,

In shop class if the pieces didn’t fit to­gether, weren’t sanded down smooth, or the con­trap­tion didn’t work, you flunked the course.

In chem­istry lab, if you didn’t mea­sure the pH right, same prob­lem.

In physics if your mea­sure­ments of waves or ac­cel­er­a­tion down the in­clined plane were wrong down went your grade.

Guess we must have gone to differ­ent schools.

John

• (Thread necro­mancy cour­tesy of TeMPOraL’s com­ment.)

in­clined plane

Here’s Feyn­man crit­i­ciz­ing the Brazilian ed­u­ca­tional sys­tem (in the late 1940s and early 1950s), but I get the im­pres­sion from his writ­ing that he thought this was a wide­spread prob­lem that was par­tic­u­larly bad in Brazil. (See for in­stance the stuff about Amer­i­can text­books later “Surely You’re Jok­ing”.)

Then I held up the el­e­men­tary physics text­book they were us­ing. “There are no ex­per­i­men­tal re­sults men­tioned any­where in this book, ex­cept in one place where there is a ball, rol­ling down an in­clined plane, in which it says how far the ball got af­ter one sec­ond, two sec­onds, three sec­onds, and so on. The num­bers have ‘er­rors’ in them—that is, if you look at them, you think you’re look­ing at ex­per­i­men­tal re­sults, be­cause the num­bers are a lit­tle above, or a lit­tle be­low, the the­o­ret­i­cal val­ues. The book even talks about hav­ing to cor­rect the ex­per­i­men­tal er­rors—very fine. The trou­ble is, when you calcu­late the value of the ac­cel­er­a­tion con­stant from these val­ues, you get the right an­swer. But a ball rol­ling down an in­clined plane, if it is ac­tu­ally done, has an in­er­tia to get it to turn, and will, if you do the ex­per­i­ment, pro­duce five-sev­enths of the right an­swer, be­cause of the ex­tra en­ergy needed to go into the ro­ta­tion of the ball. There­fore this sin­gle ex­am­ple of ex­per­i­men­tal ‘re­sults’ is ob­tained from a fake ex­per­i­ment. No­body had rol­led such a ball, or they would never have got­ten those re­sults!

-- “Surely You’re Jok­ing, Mr. Feyn­man!”, Richard Feyn­man (p. 217)

• Page 212-213 is even more on point.

After a lot of in­ves­ti­ga­tion, I fi­nally figured out that the stu­dents had mem­o­rized ev­ery­thing, but they didn’t know what any­thing meant. … So you see, they could pass the ex­am­i­na­tions, and “learn” all this stuff, and not know any­thing at all, ex­cept what they had mem­o­rized.

He gives sev­eral ex­am­ples re­lat­ing to the physics of light and torque. The stu­dents give great defi­ni­tions but have no idea how to ap­ply the terms to any real ob­jects, even when the ob­jects are pointed out to them.

• Not to men­tion all the 20th-cen­tury text­books men­tion­ing the tongue map thing...

• I’ve never seen or heard of such a school, at least not in my coun­try. Maybe vo­ca­tional schools grade like that, but in high schools I know, there’s no fit­ting to­get­ger, sand­ing, or mea­sur­ing any­thing. It’s just mem­o­riz­ing the­ory and solv­ing ex­er­cises.

• In both 10th grade chem­istry and AP chem we had some de­gree of grad­ing based on how close our val­ues were to the cor­rect val­ues. I’m not ac­tu­ally sure this helps that much in prac­tice though be­cause I’m pretty sure that some kids fudged their data.

• I’m pretty sure that some kids fudged their data

Well, that means that at least they knew what data they would have ex­pected to get if they had done the ex­per­i­ment right, which takes more un­der­stand­ing than just mem­o­riz­ing the teacher’s pass­words.

We had this situ­a­tion on CS stud­ies in nu­mer­i­cal meth­ods class and in metrol­ogy class. In both cases, most of the stu­dents fudged the data in the re­ports and/​or just plainly copied stuff from what the pre­vi­ous year did.

• To ex­pand on my point, I think there is not enough test­ing in schools, and what test­ing there is is too as­so­ci­ated with grades. Sup­pose that you are teach­ing a child to ride a bike, but sup­pose that each time he falls you give him a bad grade. At the end of it he knows how to ride a bike re­ally well, and any hon­est as­sess­ment of him should be, “great, he learned well, he’s done”. Un­for­tu­nately, that is not what hap­pens in school. In­stead, all his stum­bles on the way to learn­ing are sum­ma­rized and then put into the re­port card that his par­ents see.

Once the child has learned, why should it mat­ter how many times he failed in the pro­cess of learn­ing? What should mat­ter is whether he knows now.

Be­cause chil­dren are keen not to get poor grades, they are wor­ried about never stum­bling even once in any of the tests of their knowl­edge, be­cause they know that each stum­ble will make it into their per­ma­nent record. In or­der to make it at least pos­si­ble for stu­dents never to stum­ble, the tests must be de­signed so that with enough prepa­ra­tion stu­dents will not stum­ble. This limits what can be tested.

The test of get­ting on a bike and try­ing to ride will in­vari­ably re­sult in stum­ble af­ter stum­ble un­til the child even­tu­ally learns to ride. This kind of test is there­fore nec­es­sar­ily ex­cluded from the tests that the school will give chil­dren. And since such tests are key to learn­ing im­por­tant as­pects of knowl­edge, schools will fail to teach those as­pects of knowl­edge.

Imag­ine a bike rid­ing course in which chil­dren are taught the the­ory of bike rid­ing but not ac­tu­ally placed on a bike un­til the end of the semester. In fact, to en­sure that at least some stu­dents will pass the test, even the fi­nal exam can­not place them on a bike, but will in­stead need to be a form where they re­gur­gi­tate the bike rid­ing the­ory they learned.

I am ex­ag­ger­at­ing of course, but I do see this as a pro­nounced ten­dency in learn­ing.

• W Ed­wards Dem­ing pointed out in the 1950s, in his books about qual­ity man­age­ment, the folly of com­bin­ing mea­sur­ing for the pur­pose of im­prove­ment with mea­sure­ment for the pur­pose of re­mu­ner­at­ing peo­ple. If you do this, the whole mea­sure­ment pro­cess is cor­rupted—“you get what you mea­sure”. Al­most in­vari­ably how­ever the two forms of mea­sure­ment are com­bined be­cause it seems more effi­cient. As I speak, am­bu­lances are wait­ing for many hours idle out­side my lo­cal hos­pi­tal for a spare bed for their pa­tients, be­cause if the hos­pi­tal tells them to go to an­other hos­pi­tal the hos­pi­tal gets de­mer­its for failing to have a bed available. If they make them wait, no such de­mer­its are given.

Later, psy­chol­o­gists found that when you ex­ter­nally re­ward and pun­ish peo­ple for do­ing things, any other in­trin­sic re­wards from the ac­tivity tend to be ex­tin­guished. Thus (in part) the con­trast be­tween the 5 year old who is brim­ming with en­thu­si­asm for learn­ing and the re­sent­ful 14 year old who does as lit­tle at pos­si­ble at school.

Maria Montes­sori found that if you place chil­dren con­sis­tently at the sweet spot of learn­ing, where they have to make some effort but it is not dis­cour­ag­ingly difficult, they re­main en­thu­si­as­tic and learn rapidly. Micro-tests oc­cur all the time in Montes­sori class­rooms to as­sess progress. Mainly the tests in­volve a self-as­sess­ment that this ac­tivity is bor­ing now so I will move to the next one.

• Some of the crit­ics of Eliezer’s point are fal­ling in the trap of jus­tify­ing a func­tional sys­tem rather than look­ing how to op­ti­mize it. I think Con­stant’s point that we of­ten re­gur­gi­tate on the path to un­der­stand­ing. But still, I think pri­macy should be given to figur­ing out how to op­ti­mize these learn­ing pro­cesses, rather than jus­tify the func­tional el­e­ments of the learn­ing pro­cesses we cur­rently have.

• “I think pri­macy should be given to figur­ing out how to op­ti­mize these learn­ing pro­cesses, rather than jus­tify the func­tional el­e­ments of the learn­ing pro­cesses we cur­rently have.”

Part of the prob­lem may be that a lot of the ed­u­ca­tional sys­tem sim­ply has no proper crite­rion of suc­cess. What, af­ter all, is a suc­cess­ful class in (say) physics? The teacher’s suc­cess is mea­sured by the stu­dents’ suc­cess, but the stu­dents’ suc­cess is mea­sured by their perfor­mance on the ex­ams, and the ex­ams are writ­ten by the teacher, so the teacher is in­di­rectly eval­u­at­ing him­self. Which is not a proper crite­rion of suc­cess. Even if the ed­u­ca­tional sys­tem as a whole writes a na­tional exam to test the teacher, the fun­da­men­tal prob­lem re­mains, be­cause the test is writ­ten by the same or­ga­ni­za­tion that is ul­ti­mately test­ing it­self.

What is a teacher sup­posed to teach? How is this de­cided? I think there is such a large gap be­tween learn­ing and life that in effect a lot of learn­ing goes untested, or tested badly (e.g., tested by the very or­ga­ni­za­tion whose perfor­mance is be­ing tested).

This is pre­sum­ably much less of a prob­lem in vo­ca­tional train­ing. The value of a vo­ca­tional school is tested in the work­place, and it is quickly dis­cov­ered whether a par­tic­u­lar school churns out worth­less grad­u­ates.

• Uh, Eliezer, when you were 9, had you seen a wave? Did you have a sense that a wave was differ­ent in sev­eral ways from other kinds of things, even from other kinds of fluid be­hav­iors—differ­ent from a cur­rent, say?

Now, when you were 9, had you seen an ar­gle­bar­gle? Did you have a sense of how ar­gle­bar­gles differed from other things? (If so, how?)

And are there ways in which the char­ac­ter­is­tics you did rec­og­nize in waves, when you were 9, also do in fact ap­ply to light? On the other hand, are there ways in which the char­ac­ter­is­tics you rec­og­nize in ar­gle­bar­gles ap­ply to light?

I would ven­ture to say that you did learn some­thing mean­ingful by re­peat­ing your teacher’s ex­pla­na­tion that light is “made of waves”. You are overly dis­count­ing what counts as ex­per­i­men­tal ev­i­dence of the real-world phe­nom­ena which words re­fer to. You had a lot of ex­per­i­men­tal ev­i­dence of waves at age 9. You’d seen them, for a start. When your teacher told you that light is made of waves, he was helping you to un­der­stand some­thing about light by ex­plain­ing it in terms of some­thing you had real-world ex­pe­rience of.

Similarly, when my 3-year-old asks whether stars are burn­ing but planets are not, I don’t tell him that ac­tu­ally stars aren’t “burn­ing” as such. I say yes, and what he said counts as learn­ing, not “guess­ing my pass­words”—even though he doesn’t fully un­der­stand yet what I would un­der­stand by the word “burn­ing”.

• Disagree­ing with the origi­nal post is a bit like dis­agree­ing with the state­ment that there’s too much crime. What I find re­mark­able, though, is how use­ful pass­word knowl­edge is. I mean you might think it would have zero use, but it doesn’t. For in­stance, all knowl­edge of ge­og­ra­phy is pass­word knowl­edge, un­less you have ac­tu­ally been to the place. Yet, most peo­ple will ar­gue that knowl­edge of ge­og­ra­phy is a good thing be­cause it makes you form lit­tle men­tal boxes that are use­ful for or­ga­niz­ing fu­ture knowl­edge (e.g. there are some peo­ple from a place called China, and they’re prob­a­bly very differ­ent from peo­ple from a place called Botswana).

• Disagree­ing with the origi­nal post is a bit like dis­agree­ing with the state­ment that there’s too much crime. What I find re­mark­able, though, is how use­ful pass­word knowl­edge is. I mean you might think it would have zero use, but it doesn’t. For in­stance, all knowl­edge of ge­og­ra­phy is pass­word knowl­edge, un­less you have ac­tu­ally been to the place. Yet, most peo­ple will ar­gue that knowl­edge of ge­og­ra­phy is a good thing be­cause it makes you form lit­tle men­tal boxes that are use­ful for or­ga­niz­ing fu­ture knowl­edge (e.g. there are some peo­ple from a place called China, and they’re prob­a­bly very differ­ent from peo­ple from a place called Botswana).

• When you are pre­sented with a new con­cept, the first step is to “me­chan­i­cally” learn it. At that point, you are able to solve only ques­tions that closely match what you were taught.

The next step is to re­ally un­der­stand the mean­ing of the con­cept, in a deeper sence. In school, this is usu­ally achieved by pro­vid­ing ex­cer­cises that are pro­gres­sively harder and harder. Harder in this case means that the ques­tions di­verges more from the learnt ma­te­rial and more and more re­quires deeper un­der­stand­ing of the con­cept.

If you only go so far as learn­ing the “me­chan­i­cal” part of the con­cept, you usu­ally will for­get about it rather quickly. If you, on the other hand, re­ally un­der­stand the mean­ing be­hind the con­cept, the knowl­edge will stick much longer. And not only that, it is okey to for­get about the me­chan­i­cal steps and only keep the ba­sic un­der­stand­ing as you can quickly and eas­ily look up the me­chan­i­cal part when you rec­og­nize a prob­lem.

I think this is all a nor­mal learn­ing pro­cess, prac­tised ev­ery­day in school.

• Some of this is be­cuase school is more about test­ing than about teach­ing. It is eas­ier to test for the words.

• “Imag­ine a bike rid­ing course in which chil­dren are taught the the­ory of bike rid­ing but not ac­tu­ally placed on a bike un­til the end of the semester. In fact, to en­sure that at least some stu­dents will pass the test, even the fi­nal exam can­not place them on a bike, but will in­stead need to be a form where they re­gur­gi­tate the bike rid­ing the­ory they learned.”

You think that pas­sage is a joke, but it’s an ex­act de­scrip­tion of my high school’s driver’s ed­u­ca­tion unit.

• For ex­am­ple, what is in­tel­li­gence, and what is con­science?

Um, do you mean “con­scious­ness”?

• Great es­say! As some­one who has taught physics off and on for decades, I couldn’t agree more!

See the dis­cus­sion of Cog­ni­tive In­struc­tion, Math Model­ing at the ASU Model­ing Physics In­struc­tion site, http://​​mod­el­ing.asu.edu/​​CIMM.html, for an ap­proach to math teach­ing founded on this in­sight. Ver­bal be­hav­ior and sym­bol con­struc­tion, which is nearly all of what is taught in math­e­mat­ics classes af­ter 4th grade (ex­cept­ing Geom­e­try) do not rep­re­sent con­cep­tual un­der­stand­ing, and in­deed aren’t re­ally math­e­mat­ics. No won­der stu­dents hate it and most can’t re­ally learn it!

• But, brooks­foe. When you say “Yes”, you could then ex­pand, re­fine and add new in­for­ma­tion to your child.

You could say, “Yes, but it’s a very spe­cial kind of burn­ing. It’s called fu­sion. It’s what hap­pens when lots of pres­sure makes a gas called hy­dro­gen turn into an­other gas called he­lium, giv­ing off lots of heat and light!”

• I’ve had this sort of con­ver­sa­tion—I’m go­ing to guess you’re not a par­ent Arc­turus—your re­sponse is full but fails to ex­plain at 3 year old level what a gas is, what pres­sure is, what hy­dro­gen and he­lium are, what “heat and light” means in this con­text. Ba­si­cally all the im­por­tant con­cepts are un­defined. Are you go­ing to pro­ceed and ex­plain ev­ery­thing you know about the uni­verse?

Some con­cepts have to be held in abeyance pend­ing for­mu­la­tion of more ba­sic con­cepts and de­vel­op­ment of meth­ods to “un­der­stand” those con­cepts such as ex­pe­rience of wa­ter waves that helps us to un­der­stand the con­cept of E-M waves via metaphor.

• Just ex­pos­ing the child to words used in their cor­rect con­text is of sig­nifi­cant value. Some of these the child may prompt fur­ther cu­ri­os­ity but even those that are ig­nored have be­gun to be built into a frame­work for un­der­stand­ing.

• yes thats why we read the bible ;P lol

• I re­mem­ber dis­cov­er­ing this se­cret through a few re­lated events. The one that comes fore­front is my brother and I laugh­ing at a mag­a­z­ine ad­vert show­ing a sumo wrestler ski jump­ing. I made the com­ment, “Ha, he’s go­ing to fall like a rock!” My father was there and asked why? Didn’t I know that the weight of the ob­ject doesn’t de­ter­mine its fal­ling speed? Ap­par­ently, when­ever that con­cept was taught to me it didn’t stick and so I wasn’t us­ing it to make pre­dic­tions. But based on the tone of his voice I knew I was sup­posed to know this and, even though I had no clue what he was talk­ing about, I im­me­di­ately re­sponded by say­ing, “Be­cause of drag!”

This was stupid of me. It got my dad off my back and so I had cor­rectly guessed the teacher’s pass­word and the rest of the day I tried to figure out why the weight of an ob­ject wouldn’t make it fall faster. It wasn’t un­til years later that I saw proof in a sci­ence mu­seum ex­per­i­ment and I ac­cepted the the­ory as fact.

The event caused me to no­tice, how­ever, that I had switched the­o­ries to try ex­plain­ing my pre­dic­tion. I made a bad pre­dic­tion, but in­stead of say­ing oops I started des­per­ately grab­bing for ev­i­dence that backed up my re­sult. How evil of my younger self. But now that I no­ticed I was do­ing this I was able to stop it. It also caused me to start see­ing this hap­pen in oth­ers. I am still sur­prised at how of­ten peo­ple do this and never think that they could be do­ing some­thing wrong.

• “If you aren’t us­ing the diffu­sion equa­tion—putting in num­bers and get­ting out re­sults that con­trol your an­ti­ci­pa­tion of par­tic­u­lar ex­pe­riences - then the con­nec­tion be­tween map and ter­ri­tory is sev­ered as though by a knife. What re­mains is not a be­lief, but a ver­bal be­hav­ior.”

I’m sur­prised noone’s com­mented on this be­fore, but I think this is overly re­stric­tive. If I’m fa­mil­iar with the pro­cess of diffu­sion, I in­tu­itively know what’s go­ing to hap­pen, with­out ac­tu­ally plug­ging in num­bers. I don’t need to do math to reach the right con­clu­sion, I can just sort of vi­su­al­ize what’s hap­pen­ing. I think you progress from me­chan­i­cally plug­ging in num­bers to gen­er­al­iz­ing the be­hav­ior into a pat­tern that you can ap­ply, like peo­ple can catch a ball in the air with­out ac­tu­ally calcu­lat­ing a parabola. That doesn’t mean that there is no con­nec­tion be­tween map and ter­ri­tory. Of course, you have to take some care to make sure you’re ap­ply­ing the right pat­tern, but that’s no differ­ent than know­ing to ap­ply the right equa­tion.

• It seems to me that pass­words and place­hold­ers oc­cupy about the same space, and prob­a­bly look fairly similar when they’re first be­ing taught. Know­ing that light, sound, and mat­ter are all waves tells me they have some­thing in com­mon. It means once I learn what a wave is, I ought to be able to pre­dict some be­hav­iors of all three based on this new in­for­ma­tion. If I also know that some things are not waves, I’ll have a de­cent foun­da­tion for when to ap­ply the wave equa­tion, once I learn it.

In­deed, “pass­word” it­self seems to be a pass­word for the con­cept of a place­holder. Now that I’ve puz­zled that out, though, the new ques­tion arises: is it ac­tu­ally use­ful to learn a place­holder, or should we teach the wave equa­tion first, and then go in to “light is a wave, and here is why”, “sound is a wave and here is why”?

Cer­tainly, be­ing up-front about it when we’re us­ing a place­holder or an ap­prox­i­ma­tion would be wise. I must have had an odd ex­pe­rience in school, since there was gen­er­ally at least one other stu­dent that would ask “what does ‘wave’ MEAN?” and then we’d all ei­ther learn what it meant, or that it was a place­holder to be learned about later.

• I agree with you, a year and a half late. In fact, the idea can be ex­tended to EY’s con­cept of “float­ing be­liefs,” webs of code words that are only defined with re­spect to one an­other, and not with re­spect to ev­i­dence. It should be noted that if at any time, a mem­ber of the web is cor­re­lated in some way with ev­i­dence, then so is the en­tire web.

In that sense, it doesn’t seem like wasted effort to main­tain webs of “pass­words,” as long as we’re re­spon­si­ble about up­dat­ing our best guesses about re­al­ity based on only those be­liefs that are ev­i­dence-re­lated. In the long term, given enough mem­ory ca­pac­ity, it should speed our un­der­stand­ing.

• When I was a kid I had this book called “Think­ing Physics”, which was ba­si­cally a book of mul­ti­ple choice physics ques­tions (such as “an elephant and a feather are fal­ling, which one ex­pe­riences more air re­sis­tance ?”, or “Ke­pler and Gal­ileo made telescopes around the same time and Ke­pler’s was adopted widely, why ?”) aimed to point out where our nat­u­ral in­stincts or pre­sup­po­si­tions go against how physics ac­tu­ally work, and ex­plain­ing, well, how physics ac­tu­ally work.

Really, the sim­ple idea that physics are a habit of thought that have to be worked on be­cause our de­faults are in­cor­rect (or, as I re­al­ized much later, are cor­rect only in the spe­cial case of the ev­ery­day life of a so­cial ho­minid) has been helpful to me ever since, and too few peo­ple have it or re­al­ize it’s im­por­tant.

I think it gets to what you’re say­ing : one shouldn’t learn physics (or any­thing for that mat­ter) as a list of facts or meth­ods to ap­ply in the class­room, one should work to in­te­grate them into one’s men­tal model of the world. Which is not as easy as it sounds.

• Wel­come to Less Wrong! Do you also post at Slack­tivist, where Fred dis­sects the Left Be­hind se­ries?

If so, you might ap­pre­ci­ate this: I think the part about in­te­grat­ing knowl­edge into your view of the world gets at the root prob­lem with Left Be­hind. If we as­sume those books rep­re­sent the au­thors’ views in some way, then the au­thors seem to have no con­cept of facts fit­ting to­gether. So for ex­am­ple it doesn’t mat­ter that Je­sus lies about the char­ac­ters’ ac­tions, and their effect on his own, while us­ing mind con­trol to make those char­ac­ters give the “right” re­sponses to his canned speech. It only mat­ters that the words come from the Bible. This makes the bibli­cal “pre­dic­tion” tech­ni­cally true. The Bible it­self is re­li­able be­cause of bibli­cal in­errancy, and for no other rea­son—God is not always hon­est in other con­texts.

• Hi ! Yep, it’s the same me, thanks for the wel­come !

I don’t know if I’d call in­te­grat­ing knowl­edge THE root prob­lem of Left Be­hind, which has many root prob­lems, and a lack of in­te­gra­tion strikes me as too high-level and wide­spread among hu­mans to qual­ify as [i]root[/​i] per se...

But yeah, good illus­tra­tion of the prin­ci­ple :-)

(and thanks for the wel­come link, I’d some­how missed that page)

• Well, I was try­ing to think of a gen­eral rule that L&J could fol­low in or­der to re­pair their wor­ld­view. (Ob­vi­ously we should con­sider fol­low­ing this rule as well, if we can find one.) I came up with, ‘Ask how all the facts, as you see them, fit to­gether.’

We could prob­a­bly find a bet­ter ver­sion. Eliezer sug­gests the rule, ‘Try to find the thought that hurts the most,’ or loosely para­phrased, ‘Ask what the Sort­ing Hat from Meth­ods of Ra­tion­al­ity would tell you.’ But for L&J such an imag­i­nary con­ver­sa­tion would likely end with the phrase, ‘Get be­hind me, Satan!’ They do not ex­pect to have the brains to see how their facts fit to­gether, con­sid­er­ing this the province of God. And yet I feel like they must have some ex­pe­rience with judg­ing be­liefs suc­cess­fully. Surely they can’t have formed a pub­lish­ing em­pire, and sur­vived to the point where they could hire as­sis­tants, solely by do­ing what ‘the proper God-ap­pointed au­thor­i­ties’ told them to do? (Though they do lump air­plane pi­lots and sci­en­tists to­gether as ‘prac­ti­cal au­thor­i­ties.’ And we know that more be­lief in the wis­dom of ‘tra­di­tional au­thor­ity’ cor­re­lates with more gullibil­ity about any claim the au­thor­i­ties in ques­tion have not con­demned—see Chap­ter 3 here. Hmm.)

So I want to say that a per­son­al­ized ver­sion of my rule would have a bet­ter effect than imag­in­ing crit­i­cisms di­rectly. Per­haps we could tell the au­thors to imag­ine in de­tail what they would see if (or when) God shows them how all their facts fit to­gether, and ex­actly how this would al­low them to an­swer any and all ob­jec­tions. This seems con­nected with the act of imag­in­ing a par­adise you would ac­tu­ally pre­fer to Reedspacer’s Lower Bound as a long-term home. Both in­volve the rule, ‘Ask how it would ac­tu­ally work given what you know about peo­ple/​your­self.’

You might think that au­thors who wrote about the king­dom Je­sus will es­tab­lish on Earth wouldn’t need to hear these rules. You’d be wrong. :-)

• “Surely they can’t have formed a pub­lish­ing em­pire, and sur­vived to the point where they could hire as­sis­tants, solely by do­ing what ‘the proper God-ap­pointed au­thor­i­ties’ told them to do? ”

Dunno; I wouldn’t un­der­es­ti­mate to what ex­tent plain in­stinct can make one be­have in a ra­tio­nal-like man­ner even though one’s be­liefs aren’t ra­tio­nal. How those in­stincts are ra­tio­nal­ized post-hoc, if they’re ra­tio­nal­ized at all, isn’t that rele­vant.

“Per­haps we could tell the au­thors to imag­ine in de­tail what they would see if (or when) God shows them how all their facts fit to­gether, and ex­actly how this would al­low them to an­swer any and all ob­jec­tions.”

I would agree with Eliezer’s rule more than with yours here. For one thing, the is­sue isn’t so much that L&J aren’t fol­low­ing the right ra­tio­nal­ity rules; I sus­pect they don’t want to fol­low the right ra­tio­nal­ity rules. I don’t know if they haven’t re­al­ized they have to fol­low them to be right, or don’t care that much about be­ing right (or to be more ac­cu­rate, they’re suffi­ciently mar­ried to their cur­rent wor­ld­view that they don’t even want to con­sider it might be wrong), but I’m pretty sure if some­one sug­gested they fol­low ei­ther your rule or Eliezer’s they’d just stare blankly.

So there’s that. But if we as­sume we man­aged to get them to listen to what we say, then I think Eliezer’s rule would work much bet­ter, be­cause it’s much harder to mi­suse. “Ask your­self how things would ac­tu­ally work” is prone to ra­tio­nal­iza­tion, I can just pic­ture the sen­tence get­ting trans­lated by some brain mod­ule as “pic­ture your cur­rent model. Nice, in­nit ?”.

Or, put an­other way, I think that the part of the brain that ac­tu­ally ex­am­ines one’s be­liefs, and the part of the brain that gives you the warm glow of self-satis­fac­tion from be­ing right, are not the same part of the brain. Your ques­tion will get in­ter­cepted by the warm glow part of the brain.. Eliezer’s ques­tion… will not. In fact it looks speci­fi­cally de­signed to avoid it.

In par­tic­u­lar, if “try to find the thought that hurts you the most” would elicit “get be­hind me Satan”, I’m not con­vinced that “try and work out how your wor­ld­view would ac­tu­ally work” wouldn’t have the same re­sults. Satan is the Great De­ceiver af­ter all. How easy would it be to as­sume, once you meet the first con­tra­dic­tion, that Satan is cloud­ing your thoughts...

• Although this is an old ar­ti­cle I came to it from the The­ory of Knowl­edge ar­ti­cle (link be­low). I’m com­ment­ing be­cause this crys­tal­lizes my ob­jec­tions to a re­peated theme at LW: that ir­ra­tional­ity comes from un­ques­tioned cached thoughts, and that mod­ern ed­u­ca­tion sys­tems ex­ac­er­bate this ten­dency. In other words, I’m ques­tion­ing whether pass­word-guess­ing and mem­o­riza­tion in ed­u­ca­tion are ac­tu­ally avoid­able, even at the high­est lev­els of op­ti­miza­tion, and whether this isn’t in fact the re­sult of the ex­pan­sion of knowl­edge and the limits of hu­man cog­ni­tion.

I’m not even go­ing to ar­gue whether those are true state­ments and whether they are bad. But my own back­ground has raised ques­tions about what the solu­tion might be. My ideal­is­tic as­sump­tions about how peo­ple should learn used to track the pre­dom­i­nant ones on LW much more closely. I’m cur­rently in a tech­ni­cal grad­u­ate pro­gram for a pro­fes­sion that is differ­ent from most other types of grad­u­ate pro­grams be­cause in this field, the first few years are spent mem­o­riz­ing huge reams of in­for­ma­tion and tak­ing mul­ti­ple choice tests (have a guess as to what I’m study­ing?) My fel­low stu­dents and I would gen­uinely love to re­ally un­der­stand from a crit­i­cal per­spec­tive all the in­for­ma­tion that is get­ting shov­eled into our brains. But do­ing that, rather than mem­o­riz­ing lists of word cor­re­spon­dences, takes time that we just don’t have. There­fore in re­al­ity our tests are re­ally just ex­er­cises in figur­ing out which bar to press so we get the food pel­let. (I like that anal­ogy bet­ter than “guess­ing pass­words”.) And per­haps most scan­dalous, once we’re “in the sys­tem”, we stu­dents rec­og­nize the need for it to be this way even though it’s frus­trat­ing.

Scan­dal! Uneth­i­cal? Stupid? Cyn­i­cal, at least? Maybe yes to all. But the prag­matic re­al­ity is that there’s a trade-off. There’s such a mas­sive amount of ma­te­rial that no hu­man who ever lived would be able to crit­i­cally di­gest it all un­less the train­ing lasted far longer than it does, and while the sys­tem is cer­tainly not op­ti­mized, the body of knowl­edge it­self is so rife with de­tail not pre­dictable from first prin­ci­ples that un­til the limits on cog­ni­tion fun­da­men­tally change, this will con­tinue to be the case.

So, re­gard­ing the strat­egy stu­dents can adopt, the two ends of the spec­trum are: 1) You can just mem­o­rize de­tails long enough to re­gur­gi­tate them in the lever-press­ing ex­per­i­ments, and un­der­stand noth­ing crit­i­cally; or 2) You can in­sist on try­ing to crit­i­cally un­der­stand ev­ery­thing, and you will cer­tainly fall be­hind and fail. You. Yes, you, read­ing this, be­cause you’re hu­man, and this ap­plies to all hu­mans. (“Not me! I’m spe­cial! I still care about my fel­low man, the hu­man spirit, etc.” Sorry, can you tell I’ve had this con­ver­sa­tion be­fore?)

Un­for­tu­nately, given the pace the ma­te­rial is pre­sented and tested, you’ll end up much closer to #1 than #2. Even if you dis­agree with my as­sess­ment that a body of knowl­edge can be such that large amounts of mem­o­riza­tion are nec­es­sary, this still raises the om­nip­re­sent prob­lem of how to max­i­mize what­ever it is you want to max­i­mize while sur­rounded by ir­ra­tional hu­mans (who are run­ning the schools, and who be­lieve that mem­o­riza­tion is nec­es­sary). If you want to do the thing you’re be­ing trained for, you need the piece of pa­per. To get that pa­per, you have to pass the tests. To pass the tests you need to just mem­o­rize and not think. I guess I could just de­clare the whole en­ter­prise un­eth­i­cal and for­get about this pro­fes­sion and move out in the woods some­where. So if you can tell me how to max­i­mize ca­reer satis­fac­tion and in­come out in a nice for­est away from ev­ery­one in­stead of mem­o­riz­ing reams of B.S., I’m all ears. Se­ri­ously.

As knowl­edge ac­cu­mu­lates, this prob­lem is only go­ing to get worse, and ex­tend to more fields. Cer­tainly ed­u­ca­tion has not yet been op­ti­mized but all fields are not math and physics, and there’s enough un­pre­dictable de­tail in the world that as knowl­edge ac­cu­mu­lates, so will the mem­o­riza­tion re­quired to learn that knowl­edge. And if we can’t always use crit­i­cal think­ing at ev­ery step—which we can’t—then pass­word-guess­ing is bet­ter than noth­ing.

• that ir­ra­tional­ity comes from un­ques­tioned cached thoughts, and that mod­ern ed­u­ca­tion sys­tems ex­ac­er­bate this ten­dency.

I don’t think LW claims that’s the only place ir­ra­tional­ity comes from. There’s the var­i­ous bi­ases, an in­abil­ity to up­date, akra­sia, and so on...

I’m cur­rently in a tech­ni­cal grad­u­ate pro­gram for a pro­fes­sion that is differ­ent from most other types of grad­u­ate pro­grams be­cause in this field, the first few years are spent mem­o­riz­ing huge reams of in­for­ma­tion and tak­ing mul­ti­ple choice tests (have a guess as to what I’m study­ing?)

Ac­tu­ar­ial sci­ence, I imag­ine. My sec­ond guess is phar­macy, but that seems less tech­ni­cal than ac­tu­ar­ial sci­ence.

And per­haps most scan­dalous, once we’re “in the sys­tem”, we stu­dents rec­og­nize the need for it to be this way even though it’s frus­trat­ing.

Can you ex­plain speci­fi­cally “the need for it to be this way”? Would a per­son learn­ing these things via SRS or a com­pa­rable long-term mem­o­riza­tion sys­tem not wipe the floor with the cram­mers in real life? If some­one who ac­tu­ally knows the ma­te­rial isn’t at an ad­van­tage, then why do you need to know these things in the first place.

I ask be­cause I’m lean­ing to­ward “rec­og­niz­ing in hind­sight the rea­son be­hind the struc­ture of the sys­tem” is a bias in its own right, though I can’t say its been an­a­lyzed any­where, and I don’t have enough ev­i­dence to defini­tively say one way or an­other.

I’m a math grad stu­dent, and I also have a larger-than-fea­si­bly-pos­si­ble, un­wieldy mess of in­for­ma­tion I need to learn, but I don’t get the mercy of a mul­ti­ple-choice exam.

So, re­gard­ing the strat­egy stu­dents can adopt, the two ends of the spec­trum are: 1) You can just mem­o­rize de­tails long enough to re­gur­gi­tate them in the lever-press­ing ex­per­i­ments, and un­der­stand noth­ing crit­i­cally; or 2) You can in­sist on try­ing to crit­i­cally un­der­stand ev­ery­thing, and you will cer­tainly fall be­hind and fail.

False di­chotomy.

If you want to do the thing you’re be­ing trained for, you need the piece of pa­per. To get that pa­per, you have to pass the tests. To pass the tests you need to just mem­o­rize and not think.

Step two to three is lack­ing in jus­tifi­ca­tion.

I guess I could just de­clare the whole en­ter­prise un­eth­i­cal and for­get about this pro­fes­sion and move out in the woods some­where. So if you can tell me how to max­i­mize ca­reer satis­fac­tion and in­come out in a nice for­est away from ev­ery­one in­stead of mem­o­riz­ing reams of B.S., I’m all ears. Se­ri­ously.

What? If you truly feel the pro­fes­sion is un­eth­i­cal (and why you would feel this way is not quite clear), pivot into an­other pro­fes­sion. It’s not like ev­ery­one has to be a physi­cist or a con­struc­tion worker. There are plenty of pro­fes­sions in the world (though per­haps not as many jobs as there used to be, I sup­pose).

As knowl­edge ac­cu­mu­lates, this prob­lem is only go­ing to get worse, and ex­tend to more fields.

That’s why we have SRS, nootrop­ics, ex­pert sys­tems and (ul­ti­mately, some­day) FAI.

Cer­tainly ed­u­ca­tion has not yet been op­ti­mized but all fields are not math and physics, and there’s enough un­pre­dictable de­tail in the world that as knowl­edge ac­cu­mu­lates, so will the mem­o­riza­tion re­quired to learn that knowl­edge.

Heh! Math has plenty of “un­pre­dictable de­tail.” Heh.

And if we can’t always use crit­i­cal think­ing at ev­ery step—which we can’t—then pass­word-guess­ing is bet­ter than noth­ing.

There’s a differ­ence be­tween pass­word-guess­ing and mem­o­riza­tion that I think you’ve ig­nored in this… well, for lack of a bet­ter word, rant. There’s noth­ing wrong with mem­o­riz­ing facts; there’s ev­ery­thing wrong with mem­o­riz­ing the an­swers to ques­tions.

For in­stance, at some point a phar­ma­cist needs to mem­o­rize that, say, grapefruit juice is con­traindi­cated for some kinds of high blood pres­sure medicine (or was it choles­terol medicine? I don’t know, I’m not a phar­ma­cist.) If we ask them, “What do I need to know about high blood pres­sure medicine?”, the an­swer “Don’t take it with grapefruit juice,” isn’t a fake an­swer.

What would be a fake an­swer is if we were tak­ing a differ­ent class of HBP medicine that didn’t in­ter­act with grapefruit juice, but the phar­ma­cist said “Don’t take it with grapefruit juice,” any­way on the prin­ci­ple that grapefruit juice in­ter­acts with some HBP medicines.

That’s prob­a­bly un­clear, but I’m sure some­one else will clar­ify the situ­a­tion bet­ter.

• So, re­gard­ing the strat­egy stu­dents can adopt, the two ends of the spec­trum are: 1) You can just mem­o­rize de­tails long enough to re­gur­gi­tate them in the lever-press­ing ex­per­i­ments, and un­der­stand noth­ing crit­i­cally; or 2) You can in­sist on try­ing to crit­i­cally un­der­stand ev­ery­thing, and you will cer­tainly fall be­hind and fail.

False di­chotomy.

Noth­ing de­scribed as “two ends of the spec­trum” is ever a false di­chotomy. Or a true di­chotomy, for that mat­ter.

• While they do some­times treat this as two ends of a spec­trum, their ul­ti­mate conclusion

To pass the tests you need to just mem­o­rize and not think.

Relies on the ear­lier (per­haps un­stated, or im­plic­itly stated) premise that 1) and 2) are mu­tu­ally ex­clu­sive.

• If you be­lieve that the pas­sage in ques­tion is a false di­chotomy then you ei­ther do not un­der­stand what a di­chotomy is or you parsed the quoted text in­cor­rectly. The two ends of the spec­trum are mu­tu­ally ex­clu­sive (and look some­what ex­treme) but for them to be di­choto­mous the en­tire rest of the spec­trum—all the mod­er­ate parts—must be ex­cluded.

Even if the text is some­how wrong or stupid it is not a false di­chotomy.

• Would you rather I called it a fake spec­trum?

EDIT: I re­ally don’t un­der­stand the is­sue. In or­der to ap­ply ex­cluded mid­dle to reach the con­clu­sion, they have to both hold that the two ends of the spec­trum are di­choto­mous, and that the rest of the spec­trum ei­ther doesn’t ex­ist, or is in­con­se­quen­tial. So it’s a spec­trum in a name only, which as far as I can see is as good as a false di­chotomy.

• Would you rather I called it a fake spec­trum?

That is one of many things you could say that isn’t just strictly and triv­ially false.

But it isn’t my prefer­ences you are satis­fy­ing. I don’t mind if you say wrong things, I’m just go­ing to re­ply with a cor­rec­tion if you do. (So ‘rather?’ is a strange ques­tion!)

• Let me rephrase, then. “Is there any­thing you would cor­rect in char­ac­ter­iz­ing the afore­men­tioned as a fake spec­trum?”

Hon­estly, if I have to write so strictly, it’s al­most too much effort.

• I thought the an­swer ‘no’ would have come across, even while I dis­tanced my­self from the per­sonal con­no­ta­tions.

• I don’t think LW claims that’s the only place ir­ra­tional­ity comes from. There’s the var­i­ous bi­ases, an in­abil­ity to up­date, akra­sia, and so on...

Analo­gously,

Happy fam­i­lies are all al­ike; ev­ery un­happy fam­ily is un­happy in its own way.

Leo Tols­toy, Anna Karen­ina, Chap­ter 1, first line

• An al­ter­na­tive to hav­ing stu­dents of a field in­cre­men­tally mem­o­rize its in­for­ma­tion for re­gur­gi­ta­tion on tests only to for­get most of it af­ter the test, would be to store the in­for­ma­tion in a well or­ga­nized database and teach stu­dents how to search the database for in­for­ma­tion rele­vant to the cur­rent prob­lem they face.

I don’t know how a stu­dent in the field could achieve this change.

How­ever, there are fields, such as soft­ware en­g­ineer­ing, where you can work that way if you want, and make lots of money with­out get­ting li­censes or de­grees from es­tab­lished in­sti­tu­tions.

(It is en­couraged to read old posts, and com­ment on them when you have some­thing to say.)

• In the school sys­tem, it’s all about ver­bal be­hav­ior, whether writ­ten on pa­per or spo­ken aloud. Ver­bal be­hav­ior gets you a gold star or a failing grade. Part of un­learn­ing this bad habit is be­com­ing con­sciously aware of the differ­ence be­tween an ex­pla­na­tion and a pass­word.

In other words, I’m ques­tion­ing whether pass­word-guess­ing and mem­o­riza­tion in ed­u­ca­tion are ac­tu­ally avoid­able, even at the high­est lev­els of optimization

So, re­gard­ing the strat­egy stu­dents can adopt, the two ends of the spec­trum are...I guess I could just de­clare the whole en­ter­prise un­eth­i­cal and for­get about this pro­fes­sion and move out in the woods some­where.

This seems like shift­ing the goal­posts. How the sys­tem should be and what to do given that it is what it is are differ­ent ques­tions. The OP ad­dresses the first and you crit­i­cize it as if it ad­dresses the lat­ter.

• “But con­cise is not always pre­cise, and with­out pre­ci­sion, con­ci­sion is just vague at best, and mis­lead­ing at worst. Sev­eral years ago, a stu­dent wanted to con­test the scor­ing of one of his test an­swers in my in­tro­duc­tory psy­chol­ogy course. The test ques­tion was some­thing to the effect of “What is the pri­mary ad­van­tage of an ex­per­i­men­tal study over a cor­re­la­tional study?” and an ex­am­ple suffi­cient an­swer would have been, “Causal con­clu­sions may be drawn from an ex­per­i­ment, but not from a cor­re­la­tional study.”

The stu­dent’s an­swer was, “In an ex­per­i­ment, you ac­tu­ally test some­thing” (the word ‘test’ was un­der­lined twice). When he ques­tioned why his an­swer earned no points, I ex­plained that it failed to dis­t­in­guish the two study types at all, as both cor­re­la­tional and ex­per­i­men­tal stud­ies “test some­thing” (i.e., in in­tro­duc­tory terms, a re­la­tion­ship be­tween vari­ables and cause-and-effect, re­spec­tively). He looked at me earnestly and re­but­ted, “But you said some of the ques­tions could be an­swered in one sen­tence, and I un­der­lined ‘test’ twice.”

After a mo­ment of silence (to con­tain my dis­be­lief at his state­ment), I asked him what it would have meant if he had only un­der­lined “test” once, or not at all, and how was I to know those dis­tinc­tions in mean­ing. He had no an­swer.”

--”Pre­ci­sion First”, L. Kim­berly Epting

• I’ve just bumped into a fun link about guess­ing the teach­ers pass­word. It is only 184 words so here it is:

I do a game called k9 nose­work with my corgi. You take a qtip dipped in cer­tain scents and hide it in boxes in the be­gin­ning then pro­gress­ing to fur­ni­ture and out­side ar­eas. The game is for the dog to find the qtip and iden­tify; there’s even a con­test /​ sport. If you de­cide to do the sport, the han­dler has to read the dog and de­cide if the dog is iden­ti­fy­ing a given bit of car or what­ever as the lo­ca­tion of the goal scent or if the dog is just sniffing. The point of this ram­bling is even the plain train­ers that do k9 nose­work know that the han­dler has to go blind and can’t know where the scent is or the dog will learn to read the han­dler in­stead of learn­ing to use its nose to find the scent. The idea that the po­lice are un­aware of this is lu­dicrous, since this sport grew out of nose train­ing for po­lice dogs, and it’s a ma­jor train­ing ob­sta­cle.

The meat of the com­ment is “the han­dler has to go blind”. It took me a while to work out what is be­ing said here. My un­der­stand­ing:

You take up the sport. You hide the scented qtip your­self. You train your dog to sniff it out. Your dog does good. You go to the com­pe­ti­tion. The judge hides the scented qtip and you don’t know where. Your dog does bad. Ex­pla­na­tion: Your dog never learned to sniff out the scented qtip, he learned to read you and the tells you gave away when he sniffed near where you knew the qtip to be.

Why is this fun? Well, John Holt’s book How Chil­dren Fail has a lovely story about guess­ing the teach­ers password

… since it was clear to me from the way the chil­dren were talk­ing and act­ing that they hadn’t a no­tion of what Nouns, Ad­jec­tives, and Verbs were. Fi­nally one child said, ‘Miss ---, you shouldn’t point to the an­swer each time.’ The teacher was sur­prised, and asked what she meant. The child said, ‘Well, you don’t ex­actly point, but you kind of stand next to the an­swer.’ This was no clearer, since the teacher had been stand­ing still. But af­ter a while, as the class went on, I thought I saw what the girl meant. Since the teacher wrote each word down in its proper column, she was, in a way, get­ting her­self ready to write, point­ing her­self at the place where she would soon be writ­ing. From the an­gle of her body to the black­board the chil­dren picked up a sub­tle cue to the cor­rect an­swer.

Dogs do this too!

The re­la­tion of my com­ment to Eliezer’s point is del­i­cate. It is not cor­rect to de­cen­ter and imag­ine your­self as the teacher; Eliezer is not writ­ing ped­a­gogy. You have to de­cen­ter and imag­ine your­self as the sniffer dog. You train hard, win­ning lots of praise and treats from your owner. You wag your tail hap­pily. Then you go to the com­pe­ti­tion and dis­cover that you haven’t learned the skill that you were sup­posed to learn, at all!

• Bravo, pretty much all I have to say...

• This is no doubt a good way to con­trol the prob­lem, but I’m not grasp­ing whether or not de-em­pha­siz­ing the use of “pass­words” would be the ideal thing to do. After all, lan­guage was in­vented so that peo­ple could effec­tively and eas­ily com­mu­ni­cate their un­der­stand­ing with oth­ers. It hardly helps to NOT drill in the com­mon lan­guage to stu­dents, oth­er­wise we’d be deal­ing with peo­ple us­ing long, com­pli­cated cir­cum­lo­cu­tion to de­scribe what should be a very sim­ple con­cept.

But this ar­ti­cle is cor­rect. Tech­ni­cal knowl­edge: ask­ing HOW some­thing works, and solid­ify­ing crit­i­cal think­ing skills is paramount to good ed­u­ca­tion… but that’s no rea­son to be overtly strict, and do­ing away with ver­bal be­hav­ior. It’s been used and has worked for cen­turies, which doesn’t nec­es­sar­ily mean that it’s use­less.

• In 1990, af­ter seven years of teach­ing at Har­vard, Eric Mazur, now Balkan­ski pro­fes­sor of physics and ap­plied physics, was de­liv­er­ing clear, pol­ished lec­tures and demon­stra­tions and get­ting high stu­dent eval­u­a­tions for his in­tro­duc­tory Physics 11 course, pop­u­lated mainly by premed and en­g­ineer­ing stu­dents who were suc­cess­fully solv­ing com­pli­cated prob­lems. Then he dis­cov­ered that his suc­cess as a teacher “was a com­plete illu­sion, a house of cards.”

The epiphany came via an ar­ti­cle in the Amer­i­can Jour­nal of Physics by Ari­zona State pro­fes­sor David Hestenes. He had de­vised a very sim­ple test, couched in ev­ery­day lan­guage, to check stu­dents’ un­der­stand­ing of one of the most fun­da­men­tal con­cepts of physics—force—and had ad­ministered it to thou­sands of un­der­grad­u­ates in the south­west­ern United States. As­ton­ish­ingly, the test showed that their in­tro­duc­tory courses had taught them “next to noth­ing,” says Mazur: “After a semester of physics, they still held the same mis­con­cep­tions as they had at the be­gin­ning of the term.”

The stu­dents had im­proved at han­dling equa­tions and for­mu­las, he ex­plains, but when it came to un­der­stand­ing “what the real mean­ings of these things are, they ba­si­cally re­verted to Aris­totelian logic—thou­sands of years back.” For ex­am­ple, they could re­cite New­ton’s Third Law and ap­ply it to nu­mer­i­cal prob­lems, but when asked about a real-world event like a col­li­sion be­tween a heavy truck and a light car, many firmly de­clared that the heavy truck ex­erts a larger force. (Ac­tu­ally, an ob­ject’s weight is ir­rele­vant to the force ex­erted.)

Mazur tried the test on his own stu­dents. Right at the start, a warn­ing flag went up when one stu­dent raised her hand and asked, “How should I an­swer these ques­tions—ac­cord­ing to what you taught me, or how I usu­ally think about these things?” To Mazur’s con­ster­na­tion, the sim­ple test of con­cep­tual un­der­stand­ing showed that his stu­dents had not grasped the ba­sic ideas of his physics course: two-thirds of them were mod­ern Aris­totelians. “The stu­dents did well on text­book-style prob­lems,” he ex­plains. “They had a bag of tricks, for­mu­las to ap­ply. But that was solv­ing prob­lems by rote. They floundered on the sim­ple word prob­lems, which de­manded a real un­der­stand­ing of the con­cepts be­hind the for­mu­las.”

http://​​har­vard­magaz­ine.com/​​2012/​​03/​​twilight-of-the-lecture

• This re­minds me of an ex­change from when I was in high­school physics, talk­ing with some­one a year ahead of me who was a fresh­man in col­lege physics.

I’ve tried look­ing for the origi­nal con­ver­sa­tion, but we ap­par­ently talked about Bat­man, as­ter­oids and boats too much for Google to find it.

Ba­si­cally, I posted a prob­lem from my Physics home­work in which we had to de­ter­mine what would hap­pen if Bat­man jumped off a bridge into a pass­ing boat, in which a crim­i­nal was mak­ing an es­cape (or com­mit­ting a crime, or some such thing).

He com­plained that the only prob­lems he got were ran­dom as­ter­oid col­li­sions in the mid­dle of Nowhere, Space.

There was an­other that was de­scrip­tive rather than in­volv­ing solv­ing an equa­tion, ask­ing to ex­plain what was wrong in the situ­a­tion where Su­per­man, hov­er­ing in an alley, reaches down to grab a flee­ing crim­i­nal, with­out mov­ing in any other way. (Con­ser­va­tion of mo­men­tum, and all that). He at­tended a cer­tain pres­ti­geous en­g­ineer­ing uni­ver­sity in the Amer­i­can Mid­west, and I was at my state’s Math and Science school (Ju­niors/​se­niors, sci­ence fair was a big deal tak­ing up a year and then some, and I didn’t have to go through a hor­rible chain of time-con­sum­ing pre­reque­sits to get into most of the classes I wanted to take).

We both bailed on math and physics around the same time, af­ter Calcu­lus 2. And lab re­ports in col­lege physics proved a huge ob­sta­cle (if the over­whelming ugh fields around aca­demic writ­ing weren’t enough, the fact that we were mostly just let­ting a cart roll down a slope, and writ­ing about it in the con­text of the phe­nomenon of the week made my only in­ter­est in point­ing out sources of er­ror rather than treat­ing it as an ac­tual ex­per­i­ment).

I then pro­ceeded to sit in a build­ing for the next five years and fail roughly one class ev­ery other semester for three of those years, and still don’t have the piece of pa­per say­ing I can speak French. Look­ing for the above-men­tioned con­ver­sa­tion did bring up posts where I could ac­tu­ally speak math, though, mak­ing me won­der how things would have gone had I found Over­com­ing Bias /​ LessWrong in 2007 or 2008. (Per­haps I would have been less prone to go off on wild tan­gents com­plain­ing about my life? Hmm...)

• (Re­sponse to old post.)

If you’re pos­tu­lat­ing any­thing like most peo­ple’s ideas of Su­per­man, you’re already say­ing “ig­nore con­ser­va­tion of mo­men­tum”, be­cause in­her­ent in the con­cept is that he flies around in ar­bi­trary ways with­out hav­ing to push on any­thing. In other words, by the way you phrased the prob­lem, you just asked the stu­dent to ig­nore the very thing you’re ask­ing the prob­lem about. This won’t turn out well.

• Eliezer, thank good­ness you ex­ist.

• This is kind of like in the (fan­tas­tic) (chil­dren’s) book The My­se­ri­ous Bene­dict So­ciety. In the book, a bad per­son in­vents a ma­chine that uses TV and ra­dio broad­casts to con­trol peo­ple’s minds. They pro­ject words into ev­ery broad­cast, so that peo­ple are sub­limi­nally af­fected. The good guy builds a ma­chine that al­lows him to hear the words di­rectly. He gets gib­ber­ish. Things like, “poi­son ap­ples, poi­son worms.” He sends in a group of four chil­dren to in­ves­ti­gate. They are to go to the school set up by the bad per­son and at­tempt to gather more knowl­edge about the ma­chine. While at the school, they ex­cel at their stud­ies through, um, var­i­ous means, be­cause they heard that the best stu­dents get hooked up to the ma­chine. While they are be­ing hooked up to the ma­chine, the bad guy asks them what they think of when he says the phrase, “Poi­son ap­ples, poi­son worms.” Im­me­di­ately, the les­son on how cor­rupt the gov­ourn­ment was, the les­son that was en­ti­tled “Poi­son ap­ples, poi­son worms” comes back to them. And it’s like this that the bad guy can con­trol peo­ples’ minds with gib­ber­ish, the words mean some­thing to the chil­dren that he uses. What counts is what’s go­ing on in­side the chil­dren’s minds, not the ac­tual broad­casted words at all.

• I should give me two cents in here, but know that I am new here, and may not be com­pletely ac­cu­rate as of now. I read this over, and a few com­ments, and took into con­sid­er­a­tion most of your views

To me this is a relevent piece. Me be­ing in class I am pay­ing at­ten­tion to the Ed­u­ca­tor be­fore me and won­der­ing why Xe Just gives out the mean­ing. Then it hit me. The teacher was not go­ing into de­tail about [why] its relevent. Im cur­rently in Geom­e­try, and she was talk­ing about ma­jor, and minor arks; how­ever Xe was not go­ing into what could be used in the situ­a­tion that would need arks. Xe is im­plant­ing the pass­word to ev­ery­one as we speak, and its not Xir fault, but the Ed­u­ca­tion Xe is re­quired to give. Xe has no de­ci­sion in do­ing this. Xe has to, its Xir job to tell us the an­swers and for us to Re­peat.

• An in­ter­est­ing lit­tle demon­stra­tion is to pose the “How old is the shep­herd?” word prob­lem /​ rid­dle to kids at school: http://​​robertkaplin­sky.com/​​how-old-is-the-shep­herd/​​

• I no­tice that I am con­fused by this post.

Is the claim that this is a school thing or a life thing? I can see how this be­hav­ior might hap­pen if a stu­dent is more in­ter­ested in get­ting good grades than in ac­tual learn­ing. In such a situ­a­tion “learn­ing the teach­ers pass­word” might be a short cut to get to your ac­tual goals.

If the claim is that this is a life thing, could some one give me some more non-class­room ex­am­ple? Or­ga­nized re­li­gion counts as class­room.

When I fist heard that light is a wave, then I in­ter­preted that sen­tence in my brain an gave it mean­ing. I can’t say for sure that I gave it the cor­rect mean­ing. But I defi­antly know that I did not just save a way the sound pat­tern, as truth. Be­cause I don’t think that way, and I can’t even imag­ine think­ing that way.

I can, on the other hand, imag­ine think­ing: “If I write ‘be­cause of heat con­duc­tion’ on a test, I have a chance of get­ting points.” This is not how I went though school, be­cause I was in­ter­ested in ac­cusal learn­ing, but I can model a stu­dent who thinks this way.

“If I write ‘be­cause of heat con­duc­tion’ on a test, I have a chance of get­ting points.” is an an­ti­ci­pa­tion con­trol­ler.

• Is the claim that this is a school thing or a life thing?

This is a life thing. One pro­gram­ming ex­am­ple might be peo­ple run­ning code they’ve copy-pasted off of Stack­Overflow to see if it solves their prob­lem—they don’t un­der­stand what it will do, but they have a vague hope that it will be the magic in­can­ta­tion that will do what they want it to do.

But even there they may have a sense that pro­gram­ming has some ob­jec­tivity to it. Prob­a­bly a bet­ter ex­am­ple is dys­func­tional or­ga­ni­za­tional dy­nam­ics, where guess­ing what the boss wants you to say serves you bet­ter than try­ing to es­ti­mate what best ac­com­plishes or­ga­ni­za­tional goals.

“If I write ‘be­cause of heat con­duc­tion’ on a test, I have a chance of get­ting points.” is an an­ti­ci­pa­tion con­trol­ler.

Right, but read this sec­tion again:

This is not a hy­poth­e­sis about the metal plate. This is not even a proper be­lief. It is an at­tempt to guess the teacher’s pass­word.

Guess­ing the teacher’s pass­word is ob­vi­ously a hy­poth­e­sis—but it’s a hy­poth­e­sis about the teacher, not the plate.

• I am un­sure if we are dis­agree­ing or not. I think that it is bad if the sys­tem en­courage peo­ple to go for the wrong in­cen­tives. My point is that, I be­lieve that peo­ple know when they are hack­ing the sys­tem. I think that the stu­dents them­selves know that their hy­poth­e­sis is about the teacher and not the plate.

This is a life thing. One pro­gram­ming ex­am­ple might be peo­ple run­ning code they’ve copy-pasted off of Stack­Overflow to see if it solves their prob­lem—they don’t un­der­stand what it will do, but they have a vague hope that it will be the magic in­can­ta­tion that will do what they want it to do.

If my goal is to just make the pro­gram work, then copy-past from Stack­Overflow might be a good idea. As long as I know what I am do­ing, and don’t fool my­self in to think­ing that I un­der­stood what I just copy pasted, I don’t see the prob­lem.

I have done a lit­tle am­a­teur pro­gram­ming and I ad­mit that I have used this method. Of course I would pre­fer to un­der­stand ev­ery­thing, but at one point of an other, I just wanted some lines to do X for me, so that I could get to the part of the code that I was ac­tu­ally in­ter­ested in.

Prob­a­bly a bet­ter ex­am­ple is dys­func­tional or­ga­ni­za­tional dy­nam­ics, where guess­ing what the boss wants you to say serves you bet­ter than try­ing to es­ti­mate what best ac­com­plishes or­ga­ni­za­tional goals.

Yes, that is a good life ex­am­ple. How­ever, in this ex­am­ple I think that it is even more clear that the em­ployee has ac­cu­rate be­liefs about the world. The er­ror is with the sys­tem, not with the em­ployee.

This is not a hy­poth­e­sis about the metal plate. This is not even a proper be­lief. It is an at­tempt to guess the teacher’s pass­word.

Guess­ing the teacher’s pass­word is ob­vi­ously a hy­poth­e­sis—but it’s a hy­poth­e­sis about the teacher, not the plate.

I agree with you, Vaniver, as you say: “it’s a hy­poth­e­sis about the teacher” But I dis­agree with Yud­kowsky. A be­lief about the teacher is a proper be­lief.

Yud­kowsky claims that Guess­ing the teacher’s pass­word is a be­hav­ior that oc­curs be­cause the stu­dent does not un­der­stand their own knowl­edge or lack there of.

I claims that Guess­ing the teacher’s pass­word is an ex­am­ple of per­verse in­stan­ti­a­tion. The stu­dents have cor­rect be­liefs and are do­ing the ra­tio­nal thing, given their in­cen­tive struc­ture. They don’t think that they un­der­stand heat con­duc­tion, and they don’t care, be­cause un­der­stand­ing heat con­duc­tion is not their goal. Their goal is to get ac­cept­able grades with min­i­mum amount of effort.

Us­ing proxi-in­ten­sives works badly on in­tel­li­gent agents, even if they are made out of flesh.

• While you could be right, you’re claiming the stu­dents have a clear and ac­cu­rate model of their own be­liefs. What does that mean? Could they ex­plain the na­ture of tech­ni­cal ex­pla­na­tion?

There’s a cer­tain pop­u­lar se­ries of books which por­trays in­tel­li­gence as a mat­ter of par­rot­ing facts, with­out try­ing to con­nect any two of them—not even the dis­ap­pear­ance of ev­ery child in the world and an event in world poli­tics shortly there­after. Now, you could try to ex­plain this by say­ing the au­thors (plu­ral) are de­liber­ately sel­l­ing their cus­tomers garbage to max­i­mize re­turn-on-in­vest­ment. And you could try to claim that their fans are buy­ing the books just to sig­nal tribal mem­ber­ship. And that ex­pla­na­tion may have some power—but I draw the line at say­ing that they all un­der­stand clearly what the books lack in terms of cred­i­bil­ity.

• This an­swer have taken some time be­cause I wanted to read the link you gave, be­fore writ­ing back. I still have not read it most of it, but I think I have read enough to get your refer­ence.

I can’t com­ment on the books, be­cause I have no idea what se­ries you are talk­ing about. But your tone do sug­gest that you ex­pect me to know what se­ries this is? I am guess­ing that these books are very pop­u­lar in USA, but more or less com­pletely un­heard of in Europe? Prob­a­bly some­thing with at Chris­tian theme?

I know for cer­tain that there ex­ists stu­dents in the world, that uses teach­ers pass­word, or similar tech­niques, fully know­ing of what they do. I am slowly ac­cept­ing that there prob­a­bly also ex­ists peo­ple who think they know stuff when all they have is a state­ment they do not ac­tu­ally un­der­stand. I have cur­rently no good es­ti­mate as to which of these are most com­mon.

As to weather the first type of stu­dents could ex­plain the na­ture of tech­ni­cal ex­pla­na­tion? Why do you mean ex­actly? I am not ab­solutely sure about what Yud­kowsky mean by this con­cept, but that only mean that I am un­cer­tain about his mind, not about my own.

To me, an ex­pla­na­tion does not feel like an ex­pla­na­tion, un­less I un­der­stand all the bits, and I can not re­mem­ber ever think­ing differ­ently. If I would be told for the first time that light is a wave, then I would try to fit my cur­rent best un­der­stand­ing of light with my cur­rent best un­der­stand­ing of wave, to try to figure out what “light is a wave” could pos­si­bly mean, and then I would ask for more in­for­ma­tion, be­cause that is clearly not an ex­pla­na­tion. This must have hap­pened at some time, even though I can’t re­mem­ber the ex­act event. I do know that some­thing in my child­hood trig­gered me to want to know more about wa­ter waves.

For me it is re­ally hard to imag­ine that any­one could con­fuse a teach­ers pass­word with knowl­edge, which makes me bi­ased to­wards other ex­pla­na­tions. So maybe I am wrong. But also, do not un­der­es­ti­mate peo­ples will­ing­ness to know­ingly use tricks to pas a class, or get bet­ter grades. Here are two ex­am­ples that I re­mem­berer class­mates openly talk­ing about:

• Use ex­tra sloppy hand­writ­ing to coun­cil spel­ling mis­takes.

• On an open ques­tion, just write ev­ery­thing you can think of, that seems at least semi rele­vant and hope that you in­cluded what ever the teacher was get­ting at, some where in there.

• my sum­mary:

• For ab­stract con­cepts we need to drill down to re­late them to em­piri­cal testing

• We shouldn’t be satis­fied with just say­ing the key­words, but seek to gen­uinly un­der­stand what’s go­ing on.

• The hu­man brains tends to value what it gives so­cial ap­proval than which it matches phys­i­cal re­al­ity.

• Test as much as pos­si­ble.

• One thing I have no­ticed re­lat­ing to this in school is that on tests some­times I put down an an­swer on a quiz that I know is wrong be­cause the teacher will give me points if I put that. For ex­am­ple, on a Phys­i­cal Ed­u­ca­tion quiz the teacher briefly talked about how sugar af­fects the hu­man body. One of the ques­tions was mul­ti­ple choice and it said “Su­gar is a...” and I se­lected car­bo­hy­drate even though I knew it was wrong be­cause that is what we were taught.

• Why is sugar not a car­bo­hy­drate?

• The words sugar and car­bo­hy­drate are ba­si­cally syn­onyms. Carbs can be bro­ken down into glu­cose, galac­tose, fruc­tose and man­nose, which are called monosachar­rides aka sim­ple sug­ars. Sacharr is the Greek word for sugar. So I am not sure why you think you put down the wrong an­swer,

• “Su­gar” and “car­bo­hy­drate” are not syn­onyms (starch is a car­bo­hy­drate but not a sugar), but sugar is in­deed a car­bo­hy­drate. I do not know why Eng­ineerofS­cience thinks “car­bo­hy­drate” was a wrong an­swer on that quiz.

• I said “ba­si­cally”. :)

Starch is a polysachar­ride, the Greek word for sugar is sacharr. Starch is made up of long-chains of glu­cose, and glu­cose is sugar. The body breaks down the polysachar­rides it in­gests down into sim­ple sug­ars like glu­cose. So what­ever the origi­nal form, it always ends up be­ing sugar.

But yes, what you say is also cor­rect. All sug­ars are car­bo­hy­drates, but not all car­bo­hy­drates are sug­ars (like starch for ex­am­ple, but even starch can be fur­ther bro­ken down into sug­ars)…

• One of the ques­tions was mul­ti­ple choice and it said “Su­gar is a...” and I se­lected car­bo­hy­drate even though I knew it was wrong be­cause that is what we were taught.

It’s not a ques­tion of right or wrong but of how the term sugar is defined. Pop­u­lar defi­ni­tions of sugar like Wikipe­dia’s say: “Su­gar is the gen­er­al­ized name for sweet, short-chain, sol­u­ble car­bo­hy­drates, many of which are used in food. ”

• Here’s what you ac­tu­ally wanted to link to for “look­ing back