The 5-Second Level

To de­velop meth­ods of teach­ing ra­tio­nal­ity skills, you need to learn to fo­cus on men­tal events that oc­cur in 5 sec­onds or less. Most of what you want to teach is di­rectly on this level; the rest con­sists of chain­ing to­gether skills on this level.

As our first ex­am­ple, let’s take the vi­tal ra­tio­nal­ist skill, “Be spe­cific.”

Even with peo­ple who’ve had mod­er­ate amounts of ex­po­sure to Less Wrong, a fair amount of my helping them think effec­tively of­ten con­sists of my say­ing, “Can you give me a spe­cific ex­am­ple of that?” or “Can you be more con­crete?”

A cou­ple of for­ma­tive child­hood read­ings that taught me to be spe­cific:

“What is meant by the word red?
”It’s a color.“
”What’s a color?“
”Why, it’s a qual­ity things have.“
”What’s a qual­ity?
”Say, what are you try­ing to do, any­way?”

You have pushed him into the clouds. If, on the other hand, we ha­bit­u­ally go down the ab­strac­tion lad­der to lower lev­els of ab­strac­tion when we are asked the mean­ing of a word, we are less likely to get lost in ver­bal mazes; we will tend to “have our feet on the ground” and know what we are talk­ing about. This habit dis­plays it­self in an an­swer such as this:

“What is meant by the word red?
”Well, the next time you see some cars stopped at an in­ter­sec­tion, look at the traf­fic light fac­ing them. Also, you might go to the fire de­part­ment and see how their trucks are painted.”

-- S. I. Hayakawa, Lan­guage in Thought and Action


“Be­ware, de­mon!” he in­toned hol­lowly. “I am not with­out defenses.“
”Oh yeah? Name three.”

-- Robert Asprin, Another Fine Myth

And now, no sooner does some­one tell me that they want to “fa­cil­i­tate com­mu­ni­ca­tions be­tween man­agers and em­ploy­ees” than I say, “Can you give me a con­crete ex­am­ple of how you would do that?” Hayakawa taught me to dis­t­in­guish the con­crete and the ab­stract; and from that small pas­sage in Asprin, I picked up the dread­ful per­sonal habit of call­ing peo­ple’s bluffs, of­ten us­ing the spe­cific phrase, “Name three.”

But the real sub­ject of to­day’s les­son is how to see skills like this on the 5-sec­ond level. And now that we have a spe­cific ex­am­ple in hand, we can pro­ceed to try to zoom in on the level of cog­ni­tive events that hap­pen in 5 sec­onds or less.

Over-ab­strac­tion hap­pens be­cause it’s easy to be ab­stract. It’s eas­ier to say “red is a color” than to pause your thoughts for long enough to come up with the ex­am­ple of a stop sign. Ab­strac­tion is a path of least re­sis­tance, a form of men­tal laz­i­ness.

So the first thing that needs to hap­pen on a timescale of 5 sec­onds is per­cep­tual recog­ni­tion of highly ab­stract state­ments un­ac­com­panied by con­crete ex­am­ples, ac­com­panied by an au­to­matic aver­sion, an ick re­ac­tion—this is the trig­ger which in­vokes the skill.

Then, you have ac­tion­able stored pro­ce­dures that as­so­ci­ate to the trig­ger. And “come up with a con­crete ex­am­ple” is not a 5-sec­ond-level skill, not an ac­tion­able pro­ce­dure, it doesn’t trans­form the prob­lem into a task. An ac­tion­able men­tal pro­ce­dure that could be learned, stored, and as­so­ci­ated with the trig­ger would be “Search for a mem­ory that in­stan­ti­ates the ab­stract state­ment”, or “Try to come up with hy­po­thet­i­cal ex­am­ples, and then dis­card the lousy ex­am­ples your imag­i­na­tion keeps sug­gest­ing, un­til you fi­nally have a good ex­am­ple that re­ally shows what you were origi­nally try­ing to say”, or “Ask why you were mak­ing the ab­stract state­ment in the first place, and re­call the origi­nal men­tal causes of your mak­ing that state­ment to see if they sug­gest some­thing more con­crete.”

Or to be more spe­cific on the last men­tal pro­ce­dure: Why were you try­ing to de­scribe red­ness to some­one? Did they just run a red traf­fic light?

(And then what kind of ex­er­cise can you run some­one through, which will get them to dis­t­in­guish red traf­fic lights from green traf­fic lights? What could teach some­one to dis­t­in­guish red from green?)

When you ask how to teach a ra­tio­nal­ity skill, don’t ask “How can I teach peo­ple to be more spe­cific?” Ask, “What sort of ex­er­cise will lead peo­ple through the part of the skill where they per­cep­tu­ally rec­og­nize a state­ment as overly ab­stract?” Ask, “What ex­er­cise teaches peo­ple to think about why they made the ab­stract state­ment in the first place?” Ask, “What ex­er­cise could cause peo­ple to form, store, and as­so­ci­ate with a trig­ger, a pro­ce­dure for go­ing through hy­po­thet­i­cal ex­am­ples un­til a good one or at least ad­e­quate one is in­vented?”

Com­ing up with good ways to teach men­tal skills re­quires think­ing on the 5-sec­ond level, be­cause un­til you’ve reached that level of in­tro­spec­tive con­crete­ness, that fine­ness of gran­u­lar­ity, you can’t rec­og­nize the el­e­ments you’re try­ing to teach; you can’t rec­og­nize the pat­terns of thought you’re try­ing to build in­side a mind.

To come up with a 5-sec­ond de­scrip­tion of a ra­tio­nal­ity skill, I would sug­gest zoom­ing in on a con­crete case of a real or hy­po­thet­i­cal per­son who (a) fails in a typ­i­cal fash­ion and (b) suc­cess­fully ap­plies the skill. Break down their in­ter­nal ex­pe­rience into the small­est gran­ules you can man­age: per­cep­tual clas­sifi­ca­tions, con­texts that evoke emo­tions, fleet­ing choices made too quick for ver­bal con­sid­er­a­tion. And then gen­er­al­ize what they’re do­ing while stay­ing on the 5-sec­ond level.

Start with the con­crete ex­am­ple of the per­son who starts to say “Red is a color” and cuts them­selves off and says “Red is what that stop sign and that fire en­g­ine have in com­mon.” What did they do on the 5-sec­ond level?

  1. Per­cep­tu­ally rec­og­nize a state­ment they made as overly ab­stract.

  2. Feel the need for an ac­com­pa­ny­ing con­crete ex­am­ple.

  3. Be suffi­ciently averse to the lack of such an ex­am­ple to avoid the path of least re­sis­tance where they just let them­selves be lazy and ab­stract.

  4. As­so­ci­ate to and ac­ti­vate a stored, ac­tion­able, pro­ce­du­ral skill, e.g:
    4a. Try to re­mem­ber a mem­ory which matches that ab­stract thing you just said.
    4b. Try to in­vent a spe­cific hy­po­thet­i­cal sce­nario which matches that ab­stract thing you just said.
    4c. Ask why you said the ab­stract thing in the first place and see if that sug­gests any­thing.


    • Be­fore even 1: They rec­og­nize that the no­tion of “con­crete” means things like fold­ing chairs, events like a young woman buy­ing a vanilla ice cream, and the num­ber 17, i.e. spe­cific enough to be vi­su­al­ized; and they know “red is a color” is not spe­cific enough to be satis­fy­ing. They per­cep­tu­ally rec­og­nize (this is what Hayakawa was try­ing to teach) the car­di­nal di­rec­tions “more ab­stract” and “less ab­stract” as they ap­ply within the land­scape of the mind.

      If you are think­ing on this level of gran­u­lar­ity, then you’re much more likely to come up with a good method for teach­ing the skill “be spe­cific”, be­cause you’ll know that what­ever ex­er­cise you come up with, it ought to cause peo­ple’s minds to go through events 1-4, and provide ex­am­ples or feed­back to train per­cep­tion 0.

      Next ex­am­ple of think­ing on the 5-sec­ond scale: I pre­vi­ously asked some peo­ple (es­pe­cially from the New York LW com­mu­nity) the ques­tion “What makes ra­tio­nal­ists fun to be around?“, i.e., why is it that once you try out be­ing in a ra­tio­nal­ist com­mu­nity you can’t bear the thought of go­ing back? One of the pri­mary qual­ities cited was “Be­ing non-judg­men­tal.” Two differ­ent peo­ple came up with that ex­act phrase, but it struck me as be­ing not pre­cisely the right de­scrip­tion—ra­tio­nal­ists go around judg­ing and es­ti­mat­ing and weigh­ing things all the time. (Notic­ing small dis­cor­dances in an im­por­tant de­scrip­tion, and re­act­ing by try­ing to find an ex­act de­scrip­tion, is an­other one of those 5-sec­ond skills.) So I pon­dered, try­ing to come up with a more spe­cific image of ex­actly what it was we weren’t do­ing, i.e. Be­ing Spe­cific, and af­ter fur­ther vi­su­al­iza­tion it oc­curred to me that a bet­ter de­scrip­tion might be some­thing like this: If you are a fel­low mem­ber of my ra­tio­nal­ist com­mu­nity and you come up with a pro­posal that I dis­agree with—like “We should all prac­tice ly­ing, so that we feel less pres­sure to be­lieve things that sound good to en­dorse out loud”—then I may ar­gue with the pro­posal on con­se­quen­tial­ist grounds. I may judge. But I won’t start say­ing in im­mense in­dig­na­tion what a ter­rible per­son you must be for sug­gest­ing it.

      Now I could try to ver­bally define ex­actly what it is we don’t do, but this would fail to ap­proach the 5-sec­ond level, and prob­a­bly also fail to get at the real qual­ity that’s im­por­tant to ra­tio­nal­ist com­mu­ni­ties. That would merely be an­other at­tempt to leg­is­late what peo­ple are or aren’t al­lowed to say, and that would make things less fun. There’d be a new ac­cu­sa­tion to worry about if you said the wrong thing—“Hey! Good ra­tio­nal­ists don’t do that!” fol­lowed by a de­bate that wouldn’t be ex­pe­rienced as pleas­ant for any­one in­volved.

      In this case I think it’s ac­tu­ally eas­ier to define the thing-we-avoid on the 5-sec­ond level. Per­son A says some­thing that Per­son B dis­agrees with, and now in Per­son B’s mind there’s an op­tion to go in the di­rec­tion of a cer­tain poi­sonous plea­sure, an op­por­tu­nity to ex­pe­rience an emo­tional burst of righ­teous in­dig­na­tion and a feel­ing of su­pe­ri­or­ity, a chance to cas­ti­gate the other per­son. On the 5-sec­ond level, Per­son B re­jects this temp­ta­tion, and in­stead in­vokes the pro­ce­dure of (a) paus­ing to re­flect and then (b) talk­ing about the con­se­quences of A’s pro­posed policy in a tone that might per­haps be wor­ried (for the way of ra­tio­nal­ity is not to re­fuse all emo­tion) but nonethe­less is not filled with righ­teous out­rage and in­dig­na­tion which de­mands that all oth­ers share that in­dig­na­tion or be like­wise cas­ti­gated.

      (Which in prac­tice, makes a re­ally huge differ­ence in how much ra­tio­nal­ists can re­lax when they are around fel­low ra­tio­nal­ists. It’s the differ­ence be­tween hav­ing to care­fully tip­toe through a minefield and be­ing free to run and dance, know­ing that even if you make a mis­take, it won’t so­cially kill you. You’re even al­lowed to say “Oops” and change your mind, if you want to back­track (but that’s a whole ’nother topic of 5-sec­ond skills)...)

      The point of 5-sec­ond-level anal­y­sis is that to teach the pro­ce­du­ral habit, you don’t go into the evolu­tion­ary psy­chol­ogy of poli­tics or the game the­ory of pun­ish­ing non-pun­ish­ers (by which the in­dig­nant de­mand that oth­ers agree with their in­dig­na­tion), which is un­for­tu­nately how I tended to write back when I was writ­ing the origi­nal Less Wrong se­quences. Rather you try to come up with ex­er­cises which, if peo­ple go through them, causes them to ex­pe­rience the 5-sec­ond events—to feel the temp­ta­tion to in­dig­na­tion, and to make the choice oth­er­wise, and to as­so­ci­ate al­ter­na­tive pro­ce­du­ral pat­terns such as paus­ing, re­flect­ing, and ask­ing “What is the ev­i­dence?” or “What are the con­se­quences?”

      What would be an ex­er­cise which de­vel­ops that habit? I don’t know, al­though it’s worth not­ing that a lot of tra­di­tional ra­tio­nal­ists not as­so­ci­ated with LW also have this skill, and that it seems fairly learn­able by os­mo­sis from watch­ing other peo­ple in the com­mu­nity not be in­dig­nant. One method that seems worth test­ing would be to ex­pose peo­ple to as­ser­tions that seem like ob­vi­ous temp­ta­tions to in­dig­na­tion, and get them to talk about ev­i­dence or con­se­quences in­stead. Say, you pro­pose that eat­ing one-month-old hu­man ba­bies ought to be le­gal, be­cause one-month-old hu­man ba­bies aren’t as in­tel­li­gent as pigs, and we eat pigs. Or you could start talk­ing about fem­i­nism, in which case you can say pretty much any­thing and it’s bound to offend some­one. (Did that last sen­tence offend you? Pause and re­flect!) The point be­ing, not to per­suade any­one of any­thing, but to get them to in­tro­spec­tively rec­og­nize the mo­ment of that choice be­tween in­dig­na­tion and not-in­dig­na­tion, and walk them through an al­ter­na­tive re­sponse, so they store and as­so­ci­ate that pro­ce­du­ral skill. The ex­er­cise might fail if the con­text of a school-ex­er­cise meant that the in­dig­na­tion never got started—if the temp­ta­tion/​choice were never ex­pe­rienced. But we could try that teach­ing method, at any rate.

      (There’s this 5-sec­ond skill where you re­spond to men­tal un­cer­tainty about whether or not some­thing will work, by imag­in­ing test­ing it; and if it looks like you can just go test some­thing, then the thought oc­curs to you to just go test it. To teach this skill, we might try show­ing peo­ple a list of hy­pothe­ses and ask­ing them to quickly say on a scale of 1-10 how easy they look to test, be­cause we’re try­ing to teach peo­ple a pro­ce­du­ral habit of per­cep­tu­ally con­sid­er­ing the testable­ness of ideas. You wouldn’t give peo­ple lots of time to think, be­cause then that teaches a pro­ce­dure of go­ing through com­plex ar­gu­ments about testa­bil­ity, which you wouldn’t use rou­tinely in real life and would end up as­so­ci­at­ing pri­mar­ily to a school-con­text where a defen­si­ble ver­bal ar­gu­ment is ex­pected.)

      I should men­tion, at this point, that learn­ing to see the 5-sec­ond level draws heav­ily on the in­tro­spec­tive skill of vi­su­al­iz­ing men­tal events in spe­cific de­tail, and main­tain­ing that in­tro­spec­tive image in your mind’s eye for long enough to re­flect on it and an­a­lyze it. This may take prac­tice, so if you find that you can’t do it right away, in­stinc­tively re­act by feel­ing that you need more prac­tice to get to the lovely re­ward, in­stead of in­stinc­tively giv­ing up.

      Has ev­ery­one learned from these ex­am­ples a per­cep­tual recog­ni­tion of what the “5-sec­ond level” looks like? Of course you have! You’ve even in­stalled a men­tal habit that when you or some­body else comes up with a sup­pos­edly 5-sec­ond-level de­scrip­tion, you au­to­mat­i­cally in­spect each part of the de­scrip­tion to see if it con­tains any block units like “Be spe­cific” which are ac­tu­ally high-level chunks.

      Now, as your ex­er­cise for learn­ing the skill of “Re­solv­ing cog­ni­tive events to the 5-sec­ond level”, take a ra­tio­nal­ist skill you think is im­por­tant (or pick a ran­dom LW post from How To Ac­tu­ally Change Your Mind); come up with a con­crete ex­am­ple of that skill be­ing used suc­cess­fully; de­com­pose that us­age to a 5-sec­ond-level de­scrip­tion of per­cep­tual clas­sifi­ca­tions and emo­tion-evok­ing con­texts and as­so­ci­a­tive trig­gers to ac­tion­able pro­ce­dures etcetera; check your de­scrip­tion to make sure that each part of it can be vi­su­al­ized as a con­crete men­tal pro­cess and that there are no non-ac­tion­able ab­stract chunks; come up with a teach­ing ex­er­cise which seems like it ought to cause those sub-5-sec­ond events to oc­cur in peo­ple’s minds; and then post your anal­y­sis and pro­posed ex­er­cise in the com­ments. Hope to hear from you soon!