Noticing Frame Differences

Pre­vi­ously: Keep­ing Beliefs Cruxy


When dis­agree­ments per­sist de­spite lengthy good-faith com­mu­ni­ca­tion, it may not just be about fac­tual dis­agree­ments – it could be due to peo­ple op­er­at­ing in en­tirely differ­ent frames — differ­ent ways of see­ing, think­ing and/​or com­mu­ni­cat­ing.

If you can’t no­tice when this is hap­pen­ing, or you don’t have the skills to nav­i­gate it, you may waste a lot of time.

Ex­am­ples of Broad Frames

Gears-ori­ented Frames

Bob and Alice’s con­ver­sa­tion is about cause and effect. Nei­ther of them are plan­ning to take di­rect ac­tions based on their con­ver­sa­tion, they’re each just in­ter­ested in un­der­stand­ing a par­tic­u­lar do­main bet­ter.

Bob has a model of the do­main that in­cludes gears A, B, C and D. Alice has a model that in­cludes gears C, D and F. They’re able to ex­change in­for­ma­tion, and their in­for­ma­tion is com­pat­i­ble,and they each end up with a shared model of how some­thing works.

There are other ways this could have gone. Ben Pace cov­ered some of them in a sketch of good com­mu­ni­ca­tion:

  • Maybe they dis­cover their mod­els don’t fit, and one of them is wrong

  • Maybe com­bin­ing their mod­els re­sults in a sur­pris­ing, coun­ter­in­tu­itive out­come that takes them awhile to ac­cept.

  • Maybe they fail to in­te­grate their mod­els, be­cause they were work­ing at differ­ent lev­els of ab­strac­tion and didn’t re­al­ize it.

Some­times they might fall into sub­tler traps.

Maybe the thing Alice is call­ing “Gear C” is ac­tu­ally differ­ent from Bob’s “Gear C”. It turns out that they were us­ing the same words to mean differ­ent things, and even though they’d both read blog­posts warn­ing them about that they didn’t no­tice.

So Bob tries to slot Alice’s gear F into his gear C and it doesn’t fit. If he doesn’t already have rea­son to trust Alice’s epistemics, he may con­clude Alice is crazy (in­stead of them refer­ring to sub­tly differ­ent con­cepts).

This may cause con­fu­sion and dis­trust.

But, the point of this blog­post is that Alice and Bob have it easy.

They’re ac­tu­ally try­ing to have the same con­ver­sa­tion. They’re both try­ing to ex­change ex­plicit mod­els of cause-and-effect, and come away with a clearer un­der­stand­ing of the world through a re­duc­tion­ist lens.

There are many other frames for a con­ver­sa­tion though.

Feel­ings-Ori­ented Frames

Clark and Dwight are ex­plor­ing how they feel and re­late to each other.

The fo­cus of the con­ver­sa­tion might be nav­i­gat­ing their par­tic­u­lar re­la­tion­ship, or helping Clark un­der­stand why he’s been feel­ing frus­trated lately

When the Lan­guage of Feel­ings jus­tifies it­self to the Lan­guage of Gears, it might say things like: “Feel­ings are im­por­tant in­for­ma­tion, even if it’s fuzzy and hard to pin down or build ex­plicit mod­els out of. If you don’t have a way to listen and make sense of that in­for­ma­tion, your model of the world is go­ing to be im­pov­er­ished. This in­volves some­times look­ing at things through lenses other than what you can ex­plic­itly ver­bal­ize.”

I think this is true, and im­por­tant. The peo­ple who do their think­ing through a gear-cen­tric frame should be pay­ing at­ten­tion to feel­ings-cen­tric frames for this rea­son. (And mean­while, feel­ings them­selves to­tally have gears that can be un­der­stood through a mechanis­tic frame­work)

But for many peo­ple that’s not ac­tu­ally the point when look­ing through a feel­ings-cen­tric frame. And not un­der­stand­ing this may lead to fur­ther dis­con­nect if a Gearsy per­son and a Feel­ingsy per­son are try­ing to talk.

“Yeah feel­ings are in­for­ma­tion, but, also, like, man, you’re a hu­man be­ing with all kinds of fas­ci­nat­ing emo­tions that are an im­por­tant part of who you are. This is su­per in­ter­est­ing! And there’s a way of mak­ing sense of it that’s nec­es­sar­ily ex­pe­ri­en­tial rather than about ex­plicit, com­mu­ni­ca­ble knowl­edge.”

Frames of Power and Negotiation

Dom­i­nance and Threat

Erica is Frank’s boss. They’re dis­cussing whether the pro­ject Frank has been lead­ing should con­tinue, or whether it should stop and all the peo­ple on Frank’s team re­as­signed.

Frank ar­gues there’s a bunch of rea­sons his pro­ject is im­por­tant to the com­pany (i.e. it pro­vides fi­nan­cial value). He also ar­gues that it’s good for morale, and that can­cel­ling the pro­ject would make his team feel alienated and dis­re­spected.

Erica ar­gues back that there are other pro­jects that are more fi­nan­cially valuable, and that his team’s feel­ings aren’t im­por­tant to the com­pany.

It so hap­pens that Frank had been up for a pro­mo­tion soon, and that would put him (go­ing for­ward) on more even foot­ing with Erica, rather than her be­ing his su­pe­rior.

It’s not (nec­es­sar­ily) about the facts, or feel­ings.

If Alice and Bob wan­dered by, they might no­tice Erica or Frank seem­ing to make some­what ba­sic rea­son­ing mis­takes about how much money the pro­ject would make or why it was valuable. Naively, Alice might point out that they seem to be en­gag­ing in mo­ti­vated rea­son­ing.

If Clark or Dwight wan­dered by, they might no­tice that Erica doesn’t seem to re­ally be en­gag­ing with Frank’s wor­ries about team morale. Naively, Clark might say some­thing like “Hey, you don’t seem to re­ally be pay­ing at­ten­tion to what Frank’s team is ex­pe­rienc­ing, and this is prob­a­bly rele­vant to ac­tu­ally hav­ing the com­pany be suc­cess­ful.”

But the con­ver­sa­tion is not about shar­ing mod­els, and it’s not about un­der­stand­ing feel­ings. It’s not even nec­es­sar­ily about “what’s best for the com­pany.”

Their con­ver­sa­tion is a ne­go­ti­a­tion. For Erica and Frank, most of what’s at stake are their own fi­nan­cial in­ter­ests, and their so­cial sta­tus within the com­pany.

The dis­cus­sion is a chess board. Fi­nan­cial mod­els, worker morale, and ex­plicit ver­bal ar­gu­ments are more like game pieces than any­thing to be taken at face value.

This might be fully trans­par­ent to both Erica and Frank (such that nei­ther even con­sid­ers the other de­cep­tive). Or, they might both earnestly be­lieve what they’re say­ing – but nonethe­less, if you try to in­ter­pret the con­ver­sa­tion as a prac­ti­cal de­ci­sion about what’s best for the com­pany, you’ll come away con­fused.

The Lan­guage of Trade

Ge­orge and Han­nah are ne­go­ti­at­ing a trade.

Like Erica and Frank, this is ul­ti­mately a con­ver­sa­tion about what Ge­orge and Han­nah want.

A po­ten­tial differ­ence is that Erica and Frank might think of their situ­a­tion as zero-sum, and there­fore most of the re­s­olu­tion has more to do with figur­ing out “who would win in a poli­ti­cal fight?”, and then hav­ing the coun­ter­fac­tual loser back down.

Whereas Ge­orge/​Han­nah might be ac­tively look­ing for pos­i­tive sum trades, and in the event that they can’t find one, they just go about their lives with­out get­ting in each other’s way.

(Erica and Frank might also look for op­por­tu­ni­ties to trade, but do­ing so hon­estly might first re­quire them to es­tab­lish the de­gree to which their de­sires are mu­tu­ally in­com­pat­i­ble and who would win a dom­i­nance con­test. Then, hav­ing es­tab­lished their re­spec­tive po­si­tions, they might speak plainly about what they have to offer each other)


Notic­ing ob­vi­ous frame differences

So the first skill here, is notic­ing when you’re hav­ing wildly differ­ent ex­pec­ta­tions about what sort of con­ver­sa­tion you’re hav­ing.

If Ge­orge is look­ing for a trade and Frank is look­ing for a fight, Ge­orge might find him­self sud­denly bruised in ways he wasn’t pre­pared for. And/​or, Frank might have ran­domly de­stroyed re­sources when there’d been an op­por­tu­nity for pos­i­tive sum in­ter­ac­tion.

Or: If Dwight says “I’m feel­ing so frus­trated at work. My boss is con­stantly be­lit­tling me”, and then Bob leaps in with an ex­pla­na­tion of why his boss is do­ing that and maybe try­ing to fix it…

Well, this one is at least a stereo­typ­i­cal re­la­tion­ship failure mode you’ve prob­a­bly heard of be­fore (where Dwight might just want val­i­da­tion).

Un­tan­gling Emo­tions, Beliefs and Goals

A more in­ter­est­ing ex­am­ple of Gears-and-Feel­ings might be some­thing like:

Alice and Dwight are talk­ing about what ca­reer op­tions Dwight should con­sider. (Dwight is cur­rently an artist, not mak­ing much money, and has de­cided they want to try some­thing else)

Alice says “Have you con­sid­ered be­com­ing a pro­gram­mer? I hear they make a lot of money and you can get started with a 3 month boot­camp.”

Dwight says “Gah, don’t talk to me about pro­gram­ming.”

It turns out that Dwight’s dad always pushed him to learn pro­gram­ming, in a fairly au­thor­i­tar­ian way. Now Dwight feels a bunch of ugh­i­ness around pro­gram­ming, with a mix­ture of “You’re not the boss of me! I’mma be an artist in­stead!”

In this situ­a­tion, per­haps the best op­tion might be to say: “okay, seems like pro­gram­ming isn’t a good fit for Dwight,” and move on.

But it might also be that pro­gram­ming is ac­tu­ally a good op­tion for Dwight to con­sider… it’s just that the con­ver­sa­tion can’t pro­ceed in the straight­for­ward cost/​benefit anal­y­sis frame that Alice was ex­plor­ing.

Dwight mak­ing mean­ingful up­dates on whether pro­gram­ming is good for him de­pends on un­tan­gling his emo­tions, and/​or ex­plor­ing the re­la­tion­ship be­tween his ex­plicit mod­els and his messier in­ter­nals. It might re­quire mak­ing piece with some long­stand­ing is­sues with his father, or learn­ing to de­tach them from the “should I be a pro­gram­mer” ques­tion.

It might be that the most use­ful thing Alice can do is give him the space to work through that on his own.

If Dwight trusts Alice to shift into a feel­ings-ori­ented frame­work (or a frame­work that at least in­cludes feel­ing), Alice might be able to di­rectly help him with the pro­cess.

It may also be that this pre­req­ui­site trust doesn’t ex­ist, or that Dwight just doesn’t want to have this con­ver­sa­tion, in which case it’s prob­a­bly just best to move on to an­other topic.


Sub­tle differ­ences be­tween frames

This gets much more com­pli­cated when you ob­serve that a) there’s lots of slight vari­a­tions on frames, and b) many peo­ple and con­ver­sa­tions in­volve a mix­ture of frames.

It’s not that hard to no­tice that one per­son is in a feel­ings-cen­tric frame while an­other per­son is in a gears-cen­tric frame. But things can ac­tu­ally get even more con­fus­ing if two peo­ple share a broad frame (and so think they should be speak­ing the same lan­guage), but ac­tu­ally they’re com­mu­ni­cat­ing in two differ­ent sub­frames.

Ex­am­ple differ­ences be­tween gears-frames

Con­sider vari­a­tions of Alice and Bob – both fo­cused on causal mod­els – who are com­ing from these differ­ent van­tage points:

Goal-ori­ented vs Cu­ri­os­ity-driven conversation

Alice is try­ing to solve a spe­cific prob­lem (say, get a par­tic­u­lar car en­g­ine fixed), and Bob thinks they’re just, like, hav­ing a free­wheel­ing con­ver­sa­tion about car en­g­ines and how neat they are (and if their cu­ri­os­ity took them in a differ­ent di­rec­tion they might shift the con­ver­sa­tion to­wards some­thing that had noth­ing to do with car en­g­ines).

De­bate vs Doublecrux

Alice is try­ing to pre­sent ar­gu­ments for her side, and ex­pects Bob to re­fute those ar­gu­ments or pre­sent differ­ent ar­gu­ments. The bur­den of pre­sent­ing a good case is on Bob.

Whereas Bob thinks they’re try­ing to mu­tu­ally con­verge on true be­liefs (which might mean adopt­ing to­tally new po­si­tions, and might in­volve each per­son fo­cus­ing on how to change their own mind rather than their part­ner’s)

Spe­cific ontologies

If one per­son is, say, re­ally into eco­nomics, then they might nat­u­rally frame ev­ery­thing in terms of trans­ac­tions. Some­one else might be re­ally into pro­gram­ming and see ev­ery­thing as ab­stracted func­tions that call each other.

They might keep phras­ing things in terms that fit their preferred on­tol­ogy, and have a hard time pars­ing state­ments from a differ­ent on­tol­ogy.

Ex­am­ple differ­ences be­tween feel­ings-frames

“Mu­tual Con­nec­tion” vs “Turn Based Shar­ing”

Clark might be try­ing to share feel­ings for the sake of build­ing con­nec­tion (shar­ing back and forth, get­ting into a flow, get­ting res­o­nance).

Whereas Dwight might think the point is more for each of them to fully share their own ex­pe­rience, while the other one listens and takes up as lit­tle space as pos­si­ble.

“I Am My Feel­ings” vs “My Feel­ings are Ob­jects”

Clark might highly self iden­tify with his feel­ings (in a sort of Ro­man­tic frame­work). Dwight might care a lot about un­der­stand­ing his feel­ings but see them as tem­po­rary ob­jects in his ex­pe­rience (sort of Bud­dhist)


Con­crete ex­am­ple: The FOOM Debate

One of my origi­nal mo­ti­va­tions for this post was the Yud­kowsky/​Han­son Foom De­bate, where much ink was spilled but AFAICT nei­ther Yud­kowsky nor Han­son changed their mind much.

I re­cently re-read through some por­tions of it. The de­bate seemed to fea­ture sev­eral of the “differ­ences within gears-ori­en­ta­tion” listed above:

Spe­cific on­tolo­gies: Han­son is steeped in eco­nomics and sees it as the ob­vi­ous lens to look at AI, evolu­tion and other ma­jor his­tor­i­cal forces. Yud­kowsky in­stead sees things through the lens of op­ti­miza­tion, and how to de­velop a causal un­der­stand­ing of what re­cur­sive op­ti­miza­tion means and where/​whether we’ve seen it his­tor­i­cally.

Goal vs Cu­ri­os­ity: I have an over­all sense that Yud­kowsky is more ac­tion ori­ented – he’s speci­fi­cally set­ting out to figure out the most im­por­tant things to do to in­fluence the far fu­ture. Whereas Han­son mostly seems to see his job as “be a pro­fes­sional economist, who looks at var­i­ous situ­a­tions through an eco­nomic lens and see if that leads to in­ter­est­ing in­sights.”

Dis­cus­sion for­mat: Through­out the dis­cus­sion, Han­son and Yud­kowsky are ar­tic­u­lat­ing their points us­ing very differ­ent styles. On my re­cent read-through, I was im­pressed with the de­gree and man­ner to which they dis­cussed this ex­plic­itly:

Eliezer notes:

I think we ran into this same clash of styles last time (i.e., back at Oxford). I try to go through things sys­tem­at­i­cally, lo­cate any pos­si­ble points of dis­agree­ment, re­solve them, and con­tinue. You seem to want to jump di­rectly to the dis­agree­ment and then work back­ward to find the differ­ing premises. I worry that this puts things in a more dis­agree­able state of mind, as it were—con­ducive to feed-back­ward rea­son­ing (ra­tio­nal­iza­tion) in­stead of feed-for­ward rea­son­ing.
It’s prob­a­bly also worth bear­ing in mind that these kinds of metadis­cus­sions are im­por­tant, since this is some­thing of a trailblaz­ing case here. And that if we re­ally want to set up con­di­tions where we can’t agree to dis­agree, that might im­ply set­ting up things in a differ­ent fash­ion than the usual In­ter­net de­bates.

Han­son re­sponds:

When I at­tend a talk, I don’t im­me­di­ately jump on any­thing a speaker says that sounds ques­tion­able. I wait un­til they ac­tu­ally make a main point of their talk, and then I only jump on points that seem to mat­ter for that main point. Since most things peo­ple say ac­tu­ally don’t mat­ter for their main point, I find this to be a very use­ful strat­egy. I will be very sur­prised in­deed if ev­ery­thing you’ve said mat­tered re­gard­ing our main point of dis­agree­ment.

I found it in­ter­est­ing that I find both these points quite im­por­tant – I’ve run into each failure mode be­fore. I’m un­sure how to nav­i­gate be­tween this rock and hard place.


My main goal with this es­say was to es­tab­lish frame-differ­ences as an im­por­tant thing to look out for, and to de­scribe the con­cept from enough differ­ent an­gles to (hope­fully) give you a gen­eral sense of what to look for, rather than a sin­gle failure mode.

What to do once you no­tice a frame-differ­ence de­pends a lot on con­text, and un­for­tu­nately I’m of­ten un­sure what the best ap­proach is. The next few posts will ap­proach “what has some­times worked for me”, and (per­haps more sadly) “what hasn’t.”