Concrete Takeaways Post-CFAR

Con­crete Take­aways:

[So I re­cently vol­un­teered at a CFAR work­shop. This is part five of a five-part se­ries on how I changed my mind. It’s split into 3 sec­tions: TAPs, Heuris­tics, and Con­cepts. They get pro­gres­sively more ab­stract. It’s also quite long at around 3,000 words, so feel free to just skip around and see what looks in­ter­est­ing.]

(I didn’t post Part 3 and Part 4 on LW, as they’re more spec­u­la­tive and ar­guably less in­ter­est­ing, but I’ve linked to them on my blog if any­one’s in­ter­ested.]

This is a col­lec­tion of TAPs, heuris­tics, and con­cepts that I’ve been think­ing about re­cently. Many of them were in­spired by my time at the CFAR work­shop, but there’s not re­ally un­der­ly­ing theme be­hind it all. It’s just a col­lec­tion of ideas that are ei­ther prac­ti­cal or in­ter­est­ing.


TAPs, or Trig­ger Ac­tion Plan­ning, is a CFAR tech­nique that is used to build habits. The ba­sic idea is you pair a strong, con­crete sen­sory “trig­ger” (e.g. “when I hear my alarm go off”) with a “plan”—the thing you want to do (e.g. “I will put on my run­ning shoes”).

If you’re good at notic­ing in­ter­nal states, TAPs can also use your feel­ings or other in­ter­nal things as a trig­ger, but it’s best to try this with some­thing con­crete first to get the sense of it.

Some of the more helpful TAPs I’ve re­cently been think­ing about are be­low:

Ask for Ex­am­ples TAP:

[No­tice you have no men­tal pic­ture of what the other per­son is say­ing. → Ask for ex­am­ples.]

Ex­am­ples are good. Ex­am­ples are god. I re­ally, re­ally like them.

In con­ver­sa­tions about ab­stract top­ics, it can be easy to un­der­stand the mean­ing of the words that some­one said, yet still miss the men­tal in­tu­ition of what they’re point­ing at. Ask­ing for an ex­am­ple clar­ifies what they mean and helps you un­der­stand things bet­ter.

The trig­ger for this TAP is notic­ing that what some­one said gave you no men­tal pic­ture.

I may be ex­trap­o­lat­ing too far from too lit­tle data here, but it seems like peo­ple do try to “fol­low along” with things in their head when listen­ing. And if this men­tal nar­ra­tive, simu­la­tion, or what­ever in­ter­nal thing you’re do­ing comes up blank when some­one’s speak­ing, then this may be a sign that what they said was un­clear.

Once you no­tice this, you ask for an ex­am­ple of what gave you no men­tal pic­ture. Ideally, the other per­son can then re­spond with a more con­crete state­ment or clar­ifi­ca­tion.

Quick Fo­cus­ing TAP:

[No­tice you feel aver­sive to­wards some­thing → Be cu­ri­ous and try to source the aver­sion.]

Aver­sion Fac­tor­ing, In­ter­nal Dou­ble Crux, and Fo­cus­ing are all tech­niques CFAR teaches to help deal with in­ter­nal feel­ings of bad­ness.

While there are definite nu­ances be­tween all three tech­niques, I’ve sort of ab­stracted from the gen­eral core of “figur­ing out why you feel bad” to cre­ate an in-the-mo­ment TAP I can use to help de­bug my­self.

The trig­ger is notic­ing a men­tal flinch or an ugh field, where I in­stinc­tively shy away from look­ing too hard.

After I no­tice the feel­ing, my first step is to cul­ti­vate a sense of cu­ri­os­ity. There’s no sense of need­ing to solve it; I’m just in­ter­ested in why I’m feel­ing this way.

Once I’ve di­rected my at­ten­tion to the men­tal pain, I try to source the dis­com­fort. Us­ing some back­track­ing and check­ing mul­ti­ple threads (e.g. “is it be­cause I feel scared?”) al­lows me to figure out why. This whole pro­cess takes maybe half a minute.

When I’ve figured out the rea­son why, a sort of shift hap­pens, similar to the felt shift in fo­cus­ing. In a similar way, I’m try­ing to “ground” the neb­u­lous, un­cer­tain dis­com­fort, forc­ing it to take shape.

I’d recom­mend try­ing some Fo­cus­ing be­fore try­ing this TAP, as it’s ba­si­cally an ex­pe­d­ited ver­sion of it, hence the name.

Rule of Reflex­ivity TAP:

[No­tice you’re judg­ing some­one → Re­call an in­stance where you did some­thing similar /​ con­struct a plau­si­ble in­ter­nal nar­ra­tive]

[No­tice you’re mak­ing an ex­cuse → Re­call times where oth­ers used this ex­cuse and up­date on how you re­act in the fu­ture.]

This is a TAP that was born out of my ob­ser­va­tion that our ex­cuses seem way more self-con­sis­tent when we’re the ones say­ing then. (Oh, why hello there, Fun­da­men­tal At­tri­bu­tion Er­ror!) The point of prac­tic­ing the Rule of Reflex­ivity is to build em­pa­thy.

The Rule of Reflex­ivity goes both ways. In the first case, you want to no­tice if you’re judg­ing some­one. This might feel like as­cribing a value judg­ment to some­thing they did, e.g. “This per­son is stupid and made a bad move.”

The re­sponse is to re­call times where ei­ther you did some­thing similar or (if you think you’re perfect) think of a plau­si­ble set of events that might have caused them to act in this way. Re­mem­ber that most peo­ple don’t think they’re act­ing stupidly; they’re just do­ing what seems like a good idea from their per­spec­tive.

In the sec­ond case, you want to no­tice when you’re try­ing to jus­tify your own ac­tions. If the ex­cuses you your­self make sus­pi­ciously sound like things you’ve heard oth­ers say be­fore, then you may want to jump less likely to im­me­di­ately dis­miss­ing them in the fu­ture.

Keep Calm TAP:

[No­tice you’re start­ing to get an­gry → Take a deep breath → Speak softer and slower]

Okay, so this TAP is prob­a­bly not easy to do be­cause you’re work­ing against a biolog­i­cal re­sponse. But I’ve found it use­ful in sev­eral in­stances where oth­er­wise I would have got­ten into a deeper ar­gu­ment.

The trig­ger, of course, is notic­ing that you’re an­gry. For me, this feels like an in­creased tight­ness in my chest and a de­sire to raise my voice. I may feel like a cher­ished be­lief of mine is be­ing at­tacked.

Once I no­tice these signs, I re­mem­ber that I have this TAP which is about stay­ing calm. I think some­thing like, “Ah yes, I’m get­ting an­gry now. But I pre­vi­ously already made the de­ci­sion that it’d be a bet­ter idea to not yell.”

After that, I take a deep breath, and I try to open up my stance. Then I re­mem­ber to speak in a slower and quieter tone than pre­vi­ously. I find this TAP es­pe­cially helpful in ar­gu­ments—ahem, col­lab­o­ra­tive searches for the truth—where things get a lit­tle too ex­cited on both sides.


Heuris­tics are al­gorithm-like things you can do to help get bet­ter re­sults. I think that it’d be pos­si­ble to turn many of the heuris­tics be­low into TAPs, but there’s a sense of de­liber­ately think­ing things out that sep­a­rates these from just the “mind­less” ac­tions above.

As more for­mal pro­ce­dures, these heuris­tics do re­quire you to re­mem­ber to Take Time to do them well. How­ever, I think that the sorts of benefits you get from make it worth the slight in­vest­ment in time.

Mod­ified Mur­phyjitsu: The Time Travel Reframe:

(If you haven’t read up on Mur­phyjitsu yet, it’d prob­a­bly be good to do that first.)

Mur­phyjitsu is based off the idea of a pre­mortem, where you imag­ine that your pro­ject failed and you’re look­ing back. I’ve always found this to be a weird tem­po­ral fram­ing, and I re­al­ized there’s a po­ten­tially eas­ier way to de­scribe things:

Say you’re sit­ting at your desk, get­ting ready to write a re­port on in­tertem­po­ral travel. You’re con­fi­dent you can finish be­fore the hour is over. What could go wrong? Clos­ing Face­book, you be­gin to start typ­ing.

Sud­denly, you hear a loud CRACK! A burst of light floods your room as a figure pops into ex­is­tence, dark and silhou­et­ted by the bright­ness be­hind it. The light re­cedes, and the figure crum­ples to the ground. Float­ing in the air is a whirring gizmo, filled with turn­ing gears. Strangely enough, your at­ten­tion is drawn from the gizmo to the per­son on the ground:

The figure has a fa­mil­iar sort of shape. You ap­proach, ten­ta­tively, and find the split­ting image of your­self! The per­son stirs and speaks.

“I’m you from one week into the fu­ture,” your fu­ture self croaks. Your fu­ture self tries to tries to get up, but sinks down again.

“Oh,” you say.

“I came from the fu­ture to tell you…” your tem­po­ral clone says in a scratchy voice.

“To tell me what?” you ask. Already, you can see the whispers of a sce­nario form­ing in your head…

Fu­ture Your slowly says, “To tell you… that the re­port on in­tertem­po­ral travel that you were go­ing to write… won’t go as planned at all. Your best-case es­ti­mate failed.”

“Oh no!” you say.

Some­how, though, you aren’t sur­prised…

At this point, what plau­si­ble rea­sons for your failure come to mind?

I hy­poth­e­size that the time-travel re­frame I provide here for Mur­phyjitsu en­gages similar parts of your brain as a pre­mortem, but is 100% more ex­cit­ing to use. In all se­ri­ous­ness, I think this is a re­frame that is eas­ier to grasp com­pared to the twisted “imag­ine you’re in the fu­ture look­ing back into the past, which by the way hap­pens to be you in the pre­sent” fram­ing nor­mal Mur­phyjitsu uses.

The ac­tual (non-dra­ma­tized) word­ing of the heuris­tic, by the way, is, “Imag­ine that Fu­ture You from one week into the fu­ture comes back tel­ling you that the plan you are about to em­bark on will fail: Why?”

Low on Time? Power On!

Often, when I find my­self low on time, I feel less com­pel­led to try. This seems sort of like an in­stance of failing with aban­don, where I think some­thing like, “Oh well, I can’t pos­si­bly get any­thing done in the re­main­ing time be­tween event X and event Y”.

And then I find my­self do­ing quite lit­tle as a re­sponse.

As a re­sult, I’ve de­cided to in­ter­nal­ize the idea that be­ing low on time doesn’t mean I can’t make mean­ingful progress on my prob­lems.

This a very Re­solve-es­que tech­nique. The idea is that even if I have only 5 min­utes, that’s enough to get things down. There’s lots of use­ful things I can pack into small time chunks, like think­ing, brain­storm­ing, or do­ing some Quick Fo­cus­ing.

I’m hop­ing to com­bat the sense of ap­a­thy /​ listless­ness that creeps in when time draws to a close.

Su­per­charge Mo­ti­va­tion by Prop­a­gat­ing Emo­tional Bonds:

[Dis­claimer: I sus­pect that this isn’t an op­ti­mal mo­ti­va­tion strat­egy, and I’m sure there are peo­ple who will ob­ject to hav­ing bonds based on oth­ers rather than them­selves. That’s okay. I think this tech­nique is effec­tive, I use it, and I’d like to share it. But if you don’t think it’s right for you, feel free to just move along to the next thing.]

CFAR used to teach a skill called Prop­a­gat­ing Urges. It’s now been largely sub­sumed by In­ter­nal Dou­ble Crux, but I still find Prop­a­gat­ing Urges to be a pow­er­ful con­cept.

In short, Prop­a­gat­ing Urges hy­poth­e­sizes that mo­ti­va­tion prob­lems are caused be­cause the im­plicit parts of our­selves don’t see how the bor­ing things we do (e.g. filing taxes) causally re­late to things we care about (e.g. not go­ing to jail). The ac­tual tech­nique in­volves walk­ing through the causal chain in your mind and some visceral imagery ev­ery step of the way to get the im­plicit part of your­self on board.

I’ve taken the same gen­eral prin­ci­ple, but I’ve fo­cused it en­tirely on the re­la­tion­ships I have with other peo­ple. If all the parts of me re­al­ize that do­ing some­thing would greatly hurt those I care about, this be­comes a stronger mo­ti­va­tion than most ex­ter­nal in­cen­tives.

For ex­am­ple, I walked through an elab­o­rate in­ter­nal simu­la­tion where I wanted to stop do­ing a Thing. I imag­ined some­one I cared deeply for find­ing out about my Thing-habit and be­ing ab­solutely deeply dis­ap­pointed. I fo­cused on the sheer emo­tional weight that such dis­ap­point­ment would cause (fa­cial ex­pres­sions, what they’d feel in­side, the whole deal).

I now have a deep in­junc­tion against do­ing the Thing, and all the parts of me are in agree­ment be­cause we agree that such a Thing would hurt other peo­ple and that’s ob­vi­ously bad.

The ba­sic steps for Prop­a­gat­ing Emo­tional Bonds looks like:

  • Figure out what thing you want to do more of or stop do­ing.

  • Imag­ine what some­one you care about would think or say.

  • Really fo­cus on how visceral that feel­ing would be.

  • Re­hearse the chain of rea­son­ing (“If I do this, then X will feel bad, and I don’t want X to feel bad, so I won’t do it”) a few times.

Take Time in So­cial Con­texts:

Often, in so­cial situ­a­tions, when peo­ple ask me ques­tions, I feel an un­der­ly­ing pres­sure to an­swer quickly. It feels like if I don’t an­swer in the next ten sec­onds, some­thing’s wrong with me. (School may have con­tributed to this). I don’t ex­actly know why, but it just feels like it’s ex­pected.

I also think that be­ing forced to hurry isn’t good for think­ing well. As a re­sult, some­thing helpful I’ve found is when some­one asks some­thing like, “Is that all? Any­thing else?” is to Take Time.

My re­sponse is some­thing like, “Okay, wait, let me ac­tu­ally take a few min­utes.” At which point, I, uh, ac­tu­ally take a few min­utes to think things through. After say­ing this, it feel like it’s now so­cially per­mis­si­ble for me to take some time think­ing.

This has proven in sev­eral con­texts where, had I not Taken Time, I would have for­got­ten to bring up im­por­tant things or missed key failure-modes.

Ground Men­tal No­tions in Real­ity not by Pla­ton­ics:

One of the pro­posed rea­sons that peo­ple suck at plan­ning is that we don’t ac­tu­ally think about the de­tails be­hind our plans. We end up think­ing about them in vague black-box-style con­cepts that hide all the scary un­known un­knowns. What we’re left with is just the con­cept of our task, rather than a deep un­der­stand­ing of what our task en­tails.

In fact, this seems fairly similar to the the “pro­to­type model” that oc­curs in scope in­sen­si­tivity.

I find this is es­pe­cially prob­le­matic for tasks which look noth­ing like their con­cepts. For ex­am­ple, my men­tal rep­re­sen­ta­tion of “do­ing math” con­jures images of great math­e­mat­i­ci­ans, in­tri­cate con­nec­tions, and fan­tas­tic con­cepts like un­countable sets.

Of course, ac­tu­ally do­ing math looks more like writ­ing stuff on pa­per, slog­ging through text­books, and bang­ing your head on the table.

My brain doesn’t differ­en­ti­ate well be­tween do­ing a task and the af­fect as­so­ci­ated with the task. Thus I think it can be use­ful to try and no­tice when our brains our do­ing this sort of black-box­ing and in­stead “un­pack” the con­cepts.

This means get­ting bet­ter cor­re­spon­dences be­tween our men­tal con­cep­tions of tasks and the tasks them­selves, so that we can hope­fully ac­tu­ally choose bet­ter.

3 Con­ver­sa­tion Tips:

I of­ten for­get what it means to be hav­ing a good con­ver­sa­tion with some­one. I think I miss op­por­tu­ni­ties to learn from oth­ers when talk­ing with them. This is my handy 3-step list of Con­ver­sa­tion Tips to get more value out of con­ver­sa­tions:

1) “Steal their Magic”: Figure out what other peo­ple are re­ally good at, and then get in­spired by their awe­some­ness and think of ways you can be­come more like that. Learn from what other peo­ple are do­ing well.

2) “Find the LCD”/​”In­tel­lec­tu­ally Es­ca­late”: Figure out where your in­tel­li­gence matches theirs, and learn some­thing new. Fo­cus on Ac­tu­ally Try­ing to bridge those in­fer­en­tial dis­tances. In con­ver­sa­tions, this means fo­cus­ing on the limits of ei­ther what you know or what the other per­son knows.

3) “Con­vince or Be Con­vinced”: (This is a John Sal­vatier idea, and it also fol­lows from the above.) Fo­cus on max­i­miz­ing your per­sua­sive abil­ity to con­vince them of some­thing. Or be con­vinced of some­thing. Either way, fo­cus on up­dat­ing be­liefs, be it your own or the other party’s.

Be The Noodly Ap­pendages of the Su­per­in­tel­li­gence You Wish To See in the World:

CFAR co-founder Anna Sala­mon has this awe­some re­frame similar to IAT which asks, “Say a su­per­in­tel­li­gence ex­ists and is try­ing to take over the world. How­ever, you are its only agent. What do you do?”

I’ll ad­mit I haven’t used this one, but it’s su­per cool and not some­thing I’d thought of, so I’m in­clud­ing it here.


Con­cepts are just things in the world I’ve iden­ti­fied and drawn some bound­aries around. They are farthest from the pipeline that goes from ideas to TAPs, as con­cepts are just ideas. Still, I do think these con­cepts “bot­tom out” at some point into prac­ti­cal­ity, and I think play­ing around with them could yield in­ter­est­ing re­sults.

Paperspace =/​= Mindspace:

I tend to write things down be­cause I want to re­mem­ber them. Re­cently, though I’ve no­ticed that rather act as an ex­ten­sion of my brain, I seem to treat things I write down as no longer in my own head. As in, if I write some­thing down, it’s not nec­es­sar­ily eas­ier for me to re­call it later.

It’s as if by “offload­ing” the thoughts onto pa­per, I’ve cleared them out of my brain. This seems sub­op­ti­mal, be­cause a big rea­son I write things down is to ce­ment them more deeply within my head.

I can still ac­cess the thoughts if I’m ask­ing my­self ques­tions like, “What did I write down yes­ter­day?” but only if I’m speci­fi­cally sort­ing for things I write down.

The point is, I want stuff I write down on pa­per to be, not where I store things, but merely a sign of what’s stored in­side my brain.

Outreach: Fo­cus on Your Tar­get’s Tar­get:

One in­ter­est­ing idea I got from the CFAR work­shop was that of think­ing about your­self as a ra­dioac­tive vam­pire. Um, I mean, think­ing about your­self as a memetic vec­tor for ra­tio­nal­ity (the vam­pire thing was an ac­tual metaphor they used, though).

The in­ter­est­ing thing they men­tioned was to think, not about who you’re di­rectly in­fluenc­ing, but who your tar­gets them­selves in­fluence.

This means that not only do you have to care about the fidelity of your trans­mis­sion, but you need to think of ways to en­sure that your tar­get also does a pass­able job of pass­ing it on to their friends.

I’ve always thought about out­reach /​ memet­ics in terms of the peo­ple I di­rectly in­fluence, so look­ing at two de­grees of sep­a­ra­tion is a pretty cool thing I hadn’t thought about in the past.

I guess that if I took this ad­vice to heart, I’d prob­a­bly have to change the way that I ex­plain things. For ex­am­ple, I might want to try giv­ing more salient ex­am­ples that can be eas­ily passed on or fo­cus­ing on get­ting the in­tu­itions be­hind the ideas across.

Build in Blank Time:

Pro­fes­sor Bar­bara Oak­ley dis­t­in­guishes be­tween fo­cused and diffused modes of think­ing. Her claim is that time spent in a thoughtless ac­tivity al­lows your brain to con­tinue work­ing on prob­lems with­out con­scious in­put. This is the ba­sis of diffuse mode.

In my ex­pe­rience, I’ve found that I get in­ter­est­ing ideas or re­mem­ber im­por­tant ideas when I’m do­ing laun­dry or some­thing else similarly mind­less.

I’ve found this to be helpful enough that I’m con­sid­er­ing build­ing in “Blank Time” in my sched­ules.

My in­tu­itions here are some­thing like, “My brain is a thought-gen­er­a­tor, and it’s par­tic­u­larly ac­tive if I can pay at­ten­tion to it. But I need to be do­ing some­thing that doesn’t re­quire much of my ex­ec­u­tive func­tion to even pay at­ten­tion to my brain. So maybe hav­ing more Blank Time would be good if I want to get more ideas.”

There’s also the ad­di­tional point that meta-level think­ing can’t be done if you’re always in the mo­ment, stuck in a task. This means that, cool ideas aside, if I just want to re­ori­ent or sur­vey my cur­rent state, Blank Time can be helpful.

The 991 Rule: Few of Your Thoughts are In­sights:

The 991 Rule says that the vast ma­jor­ity of your thoughts ev­ery day are pretty bor­ing and that only about one per­cent of them are in­sight­ful.

This was gen­er­ally true for my life…and then I went to the CFAR work­shop and this rule sort of stopped be­ing ap­pro­pri­ate. (Other ex­cep­tions to this rule were EuroSPARC [now ESPR] and EAG)


I bul­l­dozed through a bunch of ideas here, some of which could have prob­a­bly gar­nered a longer post. I’ll prob­a­bly ex­plore some of these ideas later on, but if you want to talk more about any one of them, feel free to leave a com­ment /​ PM me.