A Developmental Framework for Rationality

[A way of look­ing at ra­tio­nal­ity as a set of tran­si­tion­ing wor­ld­views that op­er­ate un­der differ­ent as­sump­tions. The short sum­mary of pro­gres­sion looks like tech­niques → habits → in­tro­spec­tion → feel­ings → mul­ti­ple on­tolo­gies. Skills which ap­pear to be at odds with one an­other can be seen as op­er­at­ing un­der differ­ent mod­els of what the ques­tion of ra­tio­nal­ity is even about.]

I started mindlevelup with the goal of figur­ing out self-im­prove­ment.

Over 2 years later and over 100 (!) es­says later, I’m still at it.

Look­ing back, one thing that stands out is how I can iden­tify dis­tinct shifts in the ways I’ve ap­proached this prob­lem. The con­cepts I’ve writ­ten about, they haven’t all been op­er­at­ing on the same level of ab­strac­tion or as­sump­tions. Some are con­crete tech­niques, while oth­ers are overviews of mechanisms, still oth­ers are gen­er­al­iza­tions about the pro­cess of learn­ing the tech­niques, and there’s even a few mus­ing about the na­ture of self-im­prove­ment as a whole.

I’d like to out­line what I think is a plau­si­ble way that some­one’s thoughts might evolve as they ap­proach this prob­lem, based off my own ex­pe­riences. I think this will be use­ful in iden­ti­fy­ing what the un­der­ly­ing mod­els be­hind differ­ent ra­tio­nal­ity skills looks like. In par­tic­u­lar, I’ll be us­ing this to give ad­di­tional con­text for where I think a lot of my own es­says slot into the wor­ld­views I’ll be in­tro­duc­ing.

A] Tech­niques Rule:

You’re on a quest to find the One True Ra­tion­al­ity.

When you first en­counter ra­tio­nal­ity, it’s in the con­text of cog­ni­tive bi­ases, pesky glitches which lead us down sub­op­ti­mal paths when we can clearly see a bet­ter way for­ward.

For ex­am­ple:

  1. When we say we’re go­ing to finish some­thing in an hour, it’ll re­ally take us more than that. Some­how, pro­jects ba­si­cally never get finished on sched­ule, even when ev­ery­one knows this sort of thing hap­pens all the time!

  2. Lots of say that we want to ex­er­cise, and some­how only a few of us ac­tu­ally get it done. This dis­par­ity ex­ists for just about any­thing else, from read­ing more books, writ­ing more code, or meet­ing new peo­ple!

  3. Peo­ple on differ­ent sides of a topic will get into a room to share their opinions. And, yet, when both sides leave, they end up more sure of their origi­nal opinion then be­fore they be­gan to ar­gue!

Given these ob­ser­va­tions, you look at self-im­prove­ment as one of over­rid­ing your body’s silly built-in de­faults. The right thing to do, then, is find­ing al­gorithms that can out­perform your cur­rent ones, and then do­ing those in­stead. Un­der this frame­work, tech­niques are very ap­peal­ing.

They’re magic spells you can plug-and-chug to im­me­di­ately swap up your de­faults—just chant, cast, and op­ti­mize!

Things I’ve writ­ten which fall un­der this view­point:

  1. Fight­ing Pro­cras­ti­na­tion is all about differ­ent ways to try and deal with pro­cras­ti­na­tion. Precom­mit­ment was a big thing that I recom­mended in that es­say which I think is quite ap­peal­ing un­der a tech­nique-based wor­ld­view. It’s sim­ple to de­scribe and has all the hal­l­marks of en­forc­ing your “ra­tio­nal” de­sires onto your “ir­ra­tional” self.

  2. IAT and CEV is about a way of query­ing other parts of you to see what ac­tions are “ac­tu­ally” best for you. I think it falls un­der this mind­set when you use it as a tool to ac­com­mo­date for things hy­per­bolic dis­count­ing, e.g. ask­ing your­self “Would I re­ally re­flec­tively en­dorse this ac­tion?”

  3. Refer­ence Class Fore­cast­ing, Mur­phyjitsu, and Back-plan­ning, the three tech­niques fea­tured in the Plan­ning 101 primer very much es­pouse this view of “All you have to do is <this other thing> rather than <naive thing you already do> and then things will mag­i­cally get bet­ter!”

There’s a definite sense of wran­gling with your­self here, of say­ing things like “Stupid body! I clearly know what the cor­rect thing to do is, and you’re get­ting in my way!”

This is also char­ac­ter­ized by a sense of want­ing to col­lect tech­niques. Given the ini­tial prob­lem specs of over­rid­ing de­faults, it feels like you need a skill to counter each bias you have. Each coun­ter­mea­sure you learn about feels like a lit­tle boost up. You hunger for more spells.

And yet, de­spite hav­ing read up about all of these great spells, you can’t seem to re­call the right in­can­ta­tions and wand move­ments to call upon the right one in the right mo­ment…

B] Build­ing Au­to­mat­ic­ity:

The prob­lem, you rea­son, isn’t just about beat­ing bi­ases,

“Just do X in­stead of Y” isn’t a vi­able strat­egy if there isn’t a way to ac­tu­ally get your­self to do X at the right mo­ments. But it seems quite difficult to be in­ten­tional all the time, to be able to re­flec­tively say, “Ah, and now is the right time to cast my spell Pre­morta, which will show me an in­stance of the fu­ture where things have gone ter­ribly wrong!”

That just won’t do.

Thus you de­cide to be­come a cy­borg.

What mat­ters, af­ter all, you rea­son, is just that the de­sired al­gorithm gets ex­e­cuted at the right time. You turn to habits, to­wards in­stal­ling the afore­men­tioned al­gorithms into your­self. There’s a sub­tle shift in fo­cus that comes in. Whereas you pre­vi­ously thought about tech­niques in terms of nul­lify­ing your bi­ases, the tech­niques them­selves come into full fo­cus here: It’s about div­ing deep into the ques­tion of what it “re­ally” means to prac­tice a ra­tio­nal­ity skill.

Key as­sump­tions here are that hu­mans are stim­u­lus-re­sponse ma­chines, and that any ra­tio­nal­ity tech­nique can be turned into a habit. One big in­sight at this stage is the dis­tinc­tion be­tween declar­a­tive and pro­ce­du­ral knowl­edge—it’s very pos­si­ble for there to be an ex­pla­na­tion of a ra­tio­nal­ity tech­nique with­out giv­ing any ac­tual good in­for­ma­tion on how to im­ple­ment said tech­nique in real life.

Things I’ve writ­ten which fall un­der this view­point:

  1. TAPs are the core tech­nique here which bring ev­ery­thing to­gether. TAPs are the ul­ti­mate meta-skill; here they ex­ist as a pro­ce­du­ral­ized pro­cess that al­lows you to learn other pro­ce­du­ral­ized skills.

  2. Hunt­ing for Prac­ti­cal­ity has, as its core mes­sage the idea that you should be op­er­a­tional­iz­ing, trans­lat­ing any ad­vice you have into its most ac­tion­able form by ask­ing your­self “How do I see my­self ac­tu­ally act­ing differ­ently as a re­sult of <what­ever in­for­ma­tion I just got> in the fu­ture?” It’s all about speci­fic­ity and ac­tion­abil­ity.

  3. There Is No Akra­sia re­jects the idea of akra­sia as an all-en­com­pass­ing bo­gey­man which causes all the pro­duc­tivity prob­lems. In­stead, it pushes you to look di­rectly at what’s go­ing wrong and come up with a spe­cific fix for that thing, what­ever it is.

Armed with your Ham­mer of Re­duc­tion­ism and the pow­er­ful weld­ing power of TAPs, you get to work break­ing things down and in­stal­ling them. Things go well. You’re not just all talk any­more—you can point to spe­cific in­stances where you have in­deed performed bet­ter.

And yet, it’s not smooth sailing. Some of the soft­ware you in­stall quickly be­comes ob­so­lete and stops work­ing. Other times, your legacy code—your old de­faults—seem to in­ter­act in quite chaotic ways with all of these new things you’re in­stal­ling. Maybe, you think, it’s time to take an­other peek at that mas­sive code­base…

C] Go­ing Mental

You’re back here again.

You had one look at the code­base, and it was a to­tal mess. No doc­u­men­ta­tion and a bunch of hard-coded vari­ables.

So, against all hope, you won­der if maybe you can gain some more in­sights by look­ing at ex­actly what’s go­ing wrong with the in­stal­la­tion pro­cess it­self. No in-depth in­ter­nals anal­y­sis needed, thank you very much.

You turn your at­ten­tion in­wards, on the prac­tice, rather than what’s be­ing done be­hind the scenes. By pay­ing at­ten­tion to the sen­sa­tions of en­gag­ing in ra­tio­nal­ity, though, you re­al­ize that fo­cus­ing only on ex­e­cu­tion is fu­tile: The phe­nomenon of learn­ing skills is still a men­tal one, no mat­ter how much you’d like to ab­stract away the messi­ness of the mind and fo­cus on im­ple­men­ta­tion.

You can’t re­move your mind from the equa­tion.

Things I’ve writ­ten which ex­plore this foray into the ne­ces­sity of the in­ter­nal ex­pe­rience of learn­ing ra­tio­nal­ity:

  1. Fad­ing Novelty is about how any skill we try to learn is go­ing to be sub­ject to our brain’s ten­dency to even­tu­ally be­come ac­cus­tomed to the new­ness of said skill. Thus, we should ex­pect ini­tial high in­ter­est in a skill (per­haps rep­re­sen­ta­tive of some­one see­ing a new tech­nique) which even­tu­ally fades as time goes on (per­haps to find yet an­other skill to learn).

  2. Con­cep­tual Similar­ity Does Not Im­ply Ac­tion­able Similar­ity looks at the challenges that can come up when our brains sub­sti­tute an eas­ier ques­tion for a harder one be­cause it thinks the an­swer to both is the same. Con­cretely, some­one might im­me­di­ately dis­miss ad­vice that sounds “ob­vi­ous” as it’s eas­ier to an­swer the ques­tion of “Does this ad­vice sound novel?” rather than “Would I benefit from tak­ing this ad­vice?”

  3. Re­place Stereo­types With Ex­pe­riences is about how can of­ten let af­fect and aes­thet­ics be ma­jor de­ci­sion-mak­ing fac­tors, when in fact ex­pe­rience is a far bet­ter guide. There’s a mix-up hap­pen­ing here where your feel­ings about X get con­flated with your feel­ings while do­ing X.

  4. Ex­pli­ca­tion ex­plores how vague­ness can be com­fort­ing and be a place of re­treat for our poorly-made plans. It’s only by div­ing straight into the scary-feel­ing un­cer­tainty and mak­ing spe­cific claims/​plans that give us an op­por­tu­nity to re­ceive feed­back and ver­ifi­ca­tion.

Hu­mans have an in­ter­nal uni­verse. Ra­tion­al­ity, as a thing hu­mans do, is go­ing to in­ter­face with said uni­verse at some level. When mak­ing de­ci­sions or prac­tic­ing ra­tio­nal­ity, then, the ways our minds in­ter­act with our men­tal con­cep­tions of ra­tio­nal­ity do in fact play a role. That much is clear now.

And yet, you can’t find good solu­tions by just de­tachedly ob­serv­ing things from a dis­tance.

Sigh­ing, you roll up your sleeves.

It’s time for a chat with your in­ner demons.

D] The Hu­man Align­ment Prob­lem:

“Hello Sys­tem 1, my old friend.”

You ap­proach your­self ten­ta­tively.

“I tried go­ing against you, mold­ing you to my will. Then I tried ig­nor­ing you en­tirely, hop­ing I could do well with­out you. But it looks like I’ll need your help af­ter all. Turns out the real Ra­tion­al­ity was in­side me all along.”

You say noth­ing in re­sponse.

(It’s your­self, af­ter all.)

“How are we feel­ing right now?” you ask.

Which, of course, you already know the an­swer to.

Anger. Aver­sion. Calm. Sor­row. Worry. Want. Plea­sure. Joy.

You have lots of feel­ings.

And they’re all im­por­tant.


Given the in­trin­si­cally men­tal por­tion of learn­ing ra­tio­nal­ity, the di­rect ap­proaches given by tech­niques and habits are miss­ing a cru­cial com­po­nent. Namely, there are situ­a­tions where, de­spite your best efforts to set up a sys­tem to do X, some­thing in­side of you is re­sist­ing. The sen­sa­tion of forc­ing your­self to do some­thing is not a pleas­ant one.

So in­stead of try­ing to bind the de­mon to your will, you fuse with the de­mon.

Prac­ti­cally speak­ing, this means com­ing to terms with all of the in­tu­itive, word­less, gut pieces of your­self.

One com­po­nent of this frame is let­ting go of some sense of di­rect con­trol over your­self. You ac­cept that mo­ti­va­tion isn’t always some­thing you can hack to­gether with a for­mula. It’s not that you en­dorse a wor­ld­view where mo­ti­va­tion is mag­i­cal and ir­re­ducible, but just that, in­stru­men­tally speak­ing, there are ways of be­com­ing driven which don’t in­volve think­ing, “And now I must mo­ti­vate my­self to get X done!”

What you gain in re­turn is a great deal more self-trust. You no longer need to be the per­son that watches over your own shoulder, mak­ing sure that you get things done. It’s now much less about “forc­ing” your­self to do things be­cause the things you’re do­ing are things you want to do any­way. In­ter­nal con­flict is gen­er­ally re­moved from the equa­tion be­cause you’re giv­ing all of your differ­ent sides a voice.

It’s not just that you’ve got a new tech­nique that draws on a new way of get­ting tech­niques down. This is a view­point that isn’t about the tech­niques.

Things I’ve writ­ten which are about this shift into feel­ing out your feel­ings:

  1. Feel­ings Mat­ter. As ad­ver­tised. A slightly longer ex­pla­na­tion about how I think about feel­ings and some of the so­cial in­cen­tives for not want­ing to pub­li­cly en­dorse feel­ings as im­por­tant (e.g. be­ing as­so­ci­ated with silly self-help gu­rus).

  2. Ly­ing On The Ground is about cul­ti­vat­ing feel­ings and sen­sa­tions by, as the name sug­gests, ly­ing on the ground. It pre­sents the idea of pay­ing at­ten­tion to feel­ings as, not ex­actly a pro­duc­tivity tech­nique, but as a use­ful thing in it­self.

  3. Re­cov­er­ing From Failure looks at the com­mit­ments you make with your­self. It takes an hon­est look at what’s go­ing wrong when you break them. It op­er­ates off the as­sump­tion that all the parts of your­self have needs, and com­mit­ments you break rep­re­sent un­met needs.

I think the best ex­po­si­tion to this view­point is the Re­plac­ing Guilt se­ries by Nate Soares.

Be­ing whole with all of your­self is a pow­er­ful place to be in. You’re able to re­solve your in­ner aver­sions, which are of­ten up­stream block­ades for lots of en­deav­ors. More­over, words like “pro­cras­ti­na­tion” and “mo­ti­va­tion” are a lot less mean­ingful now that you’ve blurred the dis­tinc­tion be­tween your “wants” and your “shoulds”.

You just do them. Be­cause.

(What more rea­son do you need?)

And yet, even with this state of mind, you find your­self tempted...

E] Paradig­mas­ter:

Hav­ing al­igned your­self, you have a bet­ter mea­sure of your­self. You know there are some things you can’t do.

Or can you?

Through­out your trav­els, you hear whispers. You hear whispers of other pow­ers. Other pow­ers which promise more than al­ign­ment:

  1. Re­solve, an abil­ity to sur­pass your limits, to en­dure any trial, to always power on.

  2. At­trac­tor The­ory, a grace­ful way to go with the flow and yet pre­serve your au­ton­omy and will.

  3. Fold­ing, a self-aware­ness even deeper than al­ign­ment, which al­lows you to change your­self at a fun­da­men­tal level.

Which should you choose?

You’ve always been a good stu­dent.

You study them all.

Time passes.

You now wear more than one hat.

You’ve got lots of mod­els. Some of them con­flict, but they all have their use-cases.

You’re a mas­ter at know­ing which model to use at the right situ­a­tions. Sure, there might be some un­der­pin­nings/​as­sump­tions that don’t change much from one frame to the next, but the point is that you’re fluid enough to rec­og­nize which way of ap­proach­ing things is the most effec­tive.

The most rep­re­sen­ta­tive work I have of this frame is the In­stru­men­tal Ra­tion­al­ity se­quence as a whole, which is per­haps cheat­ing be­cause it isn’t so much as one thing which switches be­tween mod­els, but that the differ­ent es­says con­tained therein use differ­ent mod­els.

At the end, you found Many True Ra­tion­al­ities.

And yet, you have to won­der...the whole point of all these Ra­tion­al­ities is to get things done, af­ter all. And, if you look at it hard enough, isn’t still just about coun­ter­ing cog­ni­tive bi­ases? It’s not like the cen­tral prob­lem has re­ally changed.

Looks like you’re off on a quest to find the One True Ra­tion­al­ity.


Those fa­mil­iar with Ke­gan’s stages of de­vel­op­ment will likely see par­allels here, es­pe­cially as I my­self was read­ing In Over Our Heads dur­ing the writ­ing of this es­say.

I took care to try and paint this frame­work, not nec­es­sar­ily one where each stage is bet­ter, but as a set of grow­ing con­sid­er­a­tions. I think the most use­ful in­sight I can offer here is where you start to see differ­ent ra­tio­nal­ity tech­niques emerge as a con­se­quence of try­ing to solve differ­ent com­po­nents (prac­ti­cal, men­tal, on­tolog­i­cal, etc.) of the self-im­prove­ment prob­lem.