Social status part 1/2: negotiations over object-level preferences

Steven Byrnes5 Mar 2024 16:29 UTC

115 points

Communication Cultures Social & Cultural Dynamics Negotiation Social Status World Modeling Object level and Meta level

1.1 Introduction

Human interactions are full of little “negotiations”. My friend and I have different preferences about where to go for dinner. My boss and I have different preferences about how soon I should deliver the report. My spouse and I are both enjoying this chat, but we inevitably have slightly different (unstated) preferences about whose turn it is to speak, whether to change the subject, etc.

None of these are arguments. Everyone is having a lovely time. But they involve conflicting preferences, however mild, and these conflicts need to somehow get resolved.

These ubiquitous everyday “negotiations” have some funny properties. At the surface level, both people may put on an elaborate pretense that there is no conflict at all. (“Oh, it’s no problem, it would be my pleasure!”) Meanwhile, below the surface, the negotiation is playing out through the push-and-pull of subtle conversational implicatures, tone-of-voice, and so on. Additionally, sometimes one person will be tacitly treated as more of a “leader” within the interaction, and then that person gets a stronger vote.

Why do we do this? How does it work? In this post, I hope to shed some light on this phenomenon. That will take us on a journey through a bunch of topics including “leading and following”, utility function comparisons, “pushiness”, “ask versus guess culture”, plausible deniability, arms races, and more.

By the way, this is the first in a series of two posts:

In this post, I am focusing just on people’s differing “object-level preferences”—I prefer pizza but you prefer sushi; I want to crash at your place tonight but you’d rather I didn’t; etc.
Then the next post, “Social status part 2/2: everything else”, will layer on another heap of complexity, related to the fact that people also have preferences related to the interaction itself, like “a preference to be polite”, or “a preference to come across as a confident leader”. Thus, that next post will get into topics like dominance, prestige, status, offense, passive-aggressiveness, self-deprecation, and more.

[See footnote for backstory of why I’m writing about this topic→^[1]]

1.2 If two people in an interaction have conflicting object-level preferences, one will “mostly lead” and the other will “mostly follow” (or they could both “half lead”, etc.)

I am riffing on the excellent 2018 blog post “Making yourself small” by “Helen”, but with a terminological change—her “make yourself big / make yourself small” is my “mostly lead / mostly follow”.^[2] It might be helpful to read that, but hopefully this discussion will be self-contained.

If Alice and Beth^[3] are interacting, they will generally have non-identical “object-level preferences” about what to talk about, where to go, what to do, etc.

Definition: If Alice is “mostly leading” and Beth is “mostly following”, that means that, right now, Alice’s object-level preferences are a better predictor of what will actually happen than Beth’s. This would look like:

When Alice feels like saying something, she’s more likely to say it. When Beth feels like saying something, she’s more likely to keep her mouth shut, although she might say it if it comes up organically in conversation, or if the conversation has a clear gap, or more generally if Beth is confident that Alice would have a direct object-level desire for Beth to say that thing.
- (Beth might occasionally make an exception to her usual behavior and butt into the conversation if she feels an unusually strong desire to say something.)
Beth is more likely to be paying constant close attention to Alice,^[4] trying to suss out Alice’s opinions and preferences, spoken or unspoken, and to accommodate them rather than pushing back.
Alice may physically take up more space, whereas Beth may stay out of Alice’s way.
- This one is especially salient in animal interactions, as discussed in “Making yourself small” (with cute videos of goats and horses!). As another animal example, let’s say I have two pet hens, Nugget and Speckles. If Nugget is “mostly leading” and Speckles is “mostly following”, and they both have an object-level desire to eat from the same pile of grain, then Nugget will probably act in a way that maximally satisfies her object-level desires—she will walk straight to the center of the pile of grain, and start eating. Meanwhile, Speckles has the same object-level desire as Nugget, but she mostly won’t act on it; instead she will wait until Nugget is done eating, or perhaps she will eat from the side of the pile.
See “Social status hacks from The Improv Wiki” (2012) for more along these lines.

1.2.1 “Leading” versus “following” applies to both dominance and prestige

Note that I’m focusing here on what Alice and Beth are actually doing, not what their social motivations are. For example, consider the two cases {Alice=short-tempered boss, Beth=terrified subordinate} versus {Alice=pop icon, Beth=fan-girl}. Try rereading the above bullet points with either of those two situations in mind. I claim that the bullet points apply equally well to both those situations. But they’re still very different situations! Astute readers will recognize that example as the dominance-versus-prestige dichotomy, much more on which in the next post.

(For a while, I was thinking of dominance and prestige as two mostly-unrelated topics. Now I recognize that they have a major area of overlap, namely everything in this post. Again, the next post will get into how they differ.)

1.3 The “Weighted Average” Toy model: A range from “0% leader” to “100% leader” in an interaction, summing to 100%

To grossly oversimplify, let’s say that Alice and Beth are in an interaction, and Alice’s object-level preferences are captured by the utility function $U_{A}$ , while Beth’s are captured by $U_{B}$ .

My model will involve calculating a weighted average of $U_{A}$ and $U_{B}$ , with the amount of “leading” corresponding to the weights of the weighted average. (If you’re about to start complaining about utility functions being affine-transformation-invariant, then hang on, I’ll get to that in §1.3.2 below.)

For example:

In the extreme leader-follower case, Alice is 100% leader, and Beth 0%, and then the group will act entirely according to $U_{A}$ , while $U_{B}$ doesn’t matter at all. Alice just says and does whatever she thinks is best to say or do, and Beth goes along with it.
In the ideal peer case, Alice and Beth are both 50% leader, and then the group will act according to the average of their preferences, $0.5 \times U_{A} + 0.5 \times U_{B}$ . In other words, both parties will have an equal say in what ends up happening.
In between, let’s say Alice is a boss and Beth is a junior employee, at a meeting to decide the scope of responsibilities and deadlines for Beth’s project, within a relatively hierarchical, top-down organization. In this situation, Alice might be 90% leading the interaction, and Beth 10%: Alice is mostly making the decisions, and Alice has the final say, but if Beth has a strong preference about a specific aspect of the discussion, then Beth might gingerly push back on Alice, and then Alice may well be accommodating. The final group decisions would follow something like $0.9 \times U_{A} + 0.1 \times U_{B}$ —a weighted average, where Alice’s preferences carry most of the weight.

The real interaction is of course a far more complicated and subtle negotiation than this weighted-average toy model, but I still think it’s a helpful mental picture.

This weighted-average model is at least part of why people say status is zero-sum^[5]: In an interaction between Alice and Beth, it is mathematically required that the amount of leading for Alice and Beth sum to 100%. So Alice can only lead more to the extent that Beth leads less, and vice-versa.^[6] Leading isn’t exactly the same as “status”—more on which in the next post—but here we have our first taste of zero-sum-ness.

Here’s an intuitive way to think about leading-ness summing to 100% and no more: if my object-level preference is to chat with you about trains but not dinosaurs, and yours is to chat with me about dinosaurs but not trains, then at the end of the day, we can’t both have our object-level preferences fully satisfied. This is a zero-sum aspect of the interaction.

Of course, in polite society, we would typically both be “50%-leading”, or at least “close-to-50% leading”, and thus we would deftly and implicitly negotiate a compromise. Maybe the conversation topic would bounce back and forth between trains and dinosaurs, or we would find something else entirely to talk about, or we would stop talking and go watch TV, or we would find an excuse to cordially end our interaction, etc.

1.3.1 Worked example

Two people, Alice and Beth, are employees in charge of ordering food for a big upcoming meeting, and they need to quickly decide between ordering pizza versus sushi. Alice has “3 utils” of object-level utility for ordering pizza and “4 utils” for sushi, whereas Beth has “5 utils” for pizza and “2 utils” for sushi. Here’s a table that we’ll refer back to:

	Pizza	Sushi
Alice	3 utils	4 utils
Beth	5 utils	2 utils
⁵⁰⁄₅₀ peer/peer scenario	4 utils ✓	3 utils
⁹⁰⁄₁₀ boss/intern scenario	3.2 utils	3.8 utils ✓

Thus:

If Alice and Beth are peers in the interaction, each 50% leading the interaction, then we average $U_{A}$ and $U_{B}$ and they order pizza. What this looks like is: Beth has a stronger preference for ordering pizza, and that outweighs Alice’s milder preference for sushi. Alice communicates her mild preference by saying things like “Well, some of the attendees had pizza yesterday, so might prefer sushi, but I dunno, I’m sure pizza would be fine…” while Beth communicates her somewhat stronger preference by saying somewhat stronger things like “I was gonna say, umm, I really think pizza would be more appropriate for this kind of meeting … you don’t think so?”. (More on this kind of speech in §1.4 below.)
If Alice is a boss and Beth is a junior intern, in a somewhat hierarchical corporate culture, then maybe Alice will 90% lead and Beth 10%. We take a weighted average of $U_{A}$ and $U_{B}$ and they order sushi. What this looks like is: Beth feels generally cowed by Alice, and thus will “mostly follow”, communicating her preferences in an indirect and tentative way, always trying to preemptively suss out what Alice’s opinion is, and being loath to push back on it. So Alice’s mild preference for ordering sushi will win the day.

Hopefully this makes sense so far.

1.3.2 What about intersubjective utility comparisons—a.k.a., that thing about affine transformations?

Pedantic readers may be complaining that utility functions are affine-transformation-invariant, and therefore you’re not allowed to average them like that. I’ll refer to the worked example above to spell out what these pedantic readers are worried about, and then we’ll get to how I respond.

The problem is that “utils” are a meaningless unit. For example, let’s rescale Alice’s utility function $U_{A}$ , such that Alice has 3000 utils for pizza and 4000 utils for sushi. But we’ll keep Beth’s the same as above: 5 utils for pizza and 2 utils for sushi. The weighted averages are now totally different: as you can check for yourself, Alice & Beth would now order sushi in both situations. This seems like cheating—stacking the deck in Alice’s favor. But it’s not obvious what went wrong. The rescaled utility function for Alice is just as valid a description of Alice’s preferences as before. What’s going on here? Is there some grounds to say that this utility function definition is wrong, and the previous one is right?

I’ll give two answers to this conundrum.

I’ll start with my main answer, which is more intuitive and which I’ll mostly be relying on in this post. In everyday life, if I say “Alice wants Thing A much more than Beth wants Thing B”, then this is a perfectly normal thing to say, and it has a substantive and widely-understood meaning.

In the table shown above, Alice gets 1 extra util from ordering sushi, and Beth gets 3 extra utils from ordering pizza, so we can say “Beth wants pizza more than Alice wants sushi”. After rescaling, Alice gets 1000 extra utils from ordering sushi, and Beth still gets 3 from pizza, so we can say “Alice wants sushi much more than Beth wants pizza”. So these are not the same situation.

In other words, insofar as humans are generally psychologically similar, we can and should describe them with similar $U$ ’s. So the comparison is not arbitrary.

My second answer is for the benefit of readers who absolutely reject the notion that we can ever compare different people’s desire-strengths. For those readers, let’s start by considering the following. Suppose Alice and Beth are interacting. Then there is no observable difference between:

Alice has utility function $U$ and is 50% leading; and Beth has utility function $V$ and is 50% leading
Alice has utility function $3 U$ and is 25% leading; and Beth has utility function $V$ and is 75% leading

So, if we reject intersubjective utility comparisons, then it follows that “% leading” is an unobservable parameter, and hence meaningless. But wait—no it isn’t! Its absolute level is meaningless, but its changes are meaningful and important. Consider the following:

Scenario 1: Every day for the past month, Bud The Bully demands that Ned The Nerd hand over his lunch money. But today, Ned is extremely hungry. So Ned, shaking with fear, refuses: “I’m sorry Bud, I need that, you can’t have it, I’m really sorry.”
Scenario 2: Every day for the past month, Bud The Bully demands that Ned The Nerd hand over his lunch money. But today, Ned has decided that he should start sticking up for himself, even at the risk of getting beaten up. So Ned, shaking with fear, refuses: “I’m sorry Bud, I need that, you can’t have it, I’m really sorry.”

In Scenario 1, the thing that changed today is one particular entry in Ned’s object-level utility function. In Scenario 2, the thing that changed today is Ned’s general willingness to assert his preferences in relation to Bud. For example, if Bud had instead bullied Ned in a manner unrelated to eating—say, by mocking Ned’s freckles—then Bud would get pushback in Scenario 2 but not Scenario 1.

So in Scenario 2 (but not directly in 1), Ned is trying to change the general nature of his relationship with Bud. Correspondingly, in my model, I would say that in Scenario 2 (but not 1), Ned is trying to follow less and lead more in relation to Bud.

Let’s conclude. To reiterate, my actual opinion is that intersubjective utility comparisons between humans (“Beth wants pizza more than Alice wants sushi”) are good and useful, albeit approximate. But if we assume for the sake of argument that intersubjective utility comparisons are fundamentally unsound, then the main thing we lose is the mental picture wherein “50% leading” looks like two peers. On the contrary, under that assumption, “50% leading” could be anything at all—it could be a king/peasant relationship, for all we know.

However, it turns out that surprisingly little of this post series will hinge on things like that. Indeed, I’ll argue in the next post (§2.5.1) that “50% leading” is not a special threshold of any fundamental importance anyway. I’m much more interested in dynamics like “Alice wants Beth to lead less”, which is about changes to how much someone is leading. And as argued above, those changes are well-defined even without intersubjective utility comparisons.

Incidentally, I suspect that a more elegant and rigorous grounding of my “weighted average” toy model probably exists,^[7] but I don’t know what it is. Oh well, this discussion here seems fine for present purposes.

In any case, while I’ve been dwelling on the differences between Scenario 1 and Scenario 2 above, they also obviously have an important thing in common, namely Ned’s external behavior. I want to be able to talk about that thing. Since I don’t want to call it “leading”, per above, I need a new, different term. Let’s call it “pushiness”, and put it into our little model as follows:

1.4 Another toy model: An unspoken map from (subjective) “desires” to (external / behavioral) “pushiness”

Here’s a diagram illustrating what I mean by “pushiness”:

The stronger my desire, the pushier my behavior. But the map from desire-strength to pushiness is different for different people in an interaction. Specifically, the person “mostly leading” will have a higher map than the person “mostly following”, as shown here:

Depending on the details of the interaction and its cultural context, this can happen via Alice moving her map upwards (“acting bossy”, or “being a strong, confident leader”, etc.), or it can look like Beth moving her map downwards (“being tentative”, or “being deferential”, or “being eager to please”, etc.), or some combination of both.

(For example—and this will make more sense after the next post—in a “prestige” interaction, Beth chooses to “mostly follow” because she admires Alice, so this interaction would probably start with Beth moving her map downwards. By contrast, in a “dominance” interaction, Alice might first assert her authority and aggression by moving her map upwards, after which Beth might also be cowed into moving her map downwards.)

1.4.1 “Ask culture versus guess culture”

Before we start this subsection, I’d better explain what is “ask culture versus guess culture”? These terms were apparently coined in a brief 2007 MetaFilter comment that went viral. Some nice later discussions include “ask vs guess culture” by Jean Hsu (2023), or “Ask and Guess” by anonymous (2010). Here’s a short version:

In “ask culture,” it’s socially acceptable to ask for a favor—staying over at a friend’s house, requesting a raise or a letter of recommendation—and equally acceptable to refuse a favor. Asking is literally just inquiring if the request will be granted, and it’s never wrong to ask, provided you know you might be refused. In “guess culture,” however, you’re expected to guess if your request is appropriate, and you are rude if you accidentally make a request that’s judged excessive or inappropriate. You can develop a reputation as greedy or thoughtless if you make inappropriate requests.
When an asker and a guesser collide, the results are awful… —source

Incidentally, as far as I can tell, the “ask versus guess culture” distinction is either the same as, or at least closely related to, the so-called “low-context versus high-context culture” distinction in anthropology.

My proposal for how to think about ask versus guess culture: As above, “mostly leading” versus “mostly following” entails being higher and lower respectively on the pushiness scale, compared to whoever you’re interacting with. But that’s a relative metric. So there is still a degree of freedom for where you are both sitting on an absolute scale. And this is how I propose to explain the phenomenon of “ask culture versus guess culture”:

My proposal for how to think about “ask culture” versus “guess culture”. Not to scale.

1.4.2 More on “culture” clashes

If Beth is expecting the situation to be more guess-culture, and Alice is expecting the same situation to be more ask-culture, then Beth will wind up mostly following, and Alice leading, even if neither Alice nor Beth had intended for that to happen. As a result, Beth might get a false impression that Alice is aggressive, rude, and presumptuous, and conversely Alice might get a false impression that Beth is tentative and low-self-confidence. Let’s spell that out with a negotiation over pizza, per the figure above:

Here’s a real-life example from the fascinating book The Culture Map by Erin Meyer (2014). In this exchange, Mr. Díaz is a Spanish boss and Mr. Chen is his Chinese employee:

MR. DÍAZ: It looks like some of us are going to have to be here on Sunday to host the client visit.
MR. CHEN: I see.
MR. DÍAZ: Can you join us on Sunday?
MR. CHEN: Yes, I think so.
MR. DÍAZ: That would be a great help.
MR. CHEN: Yes, Sunday is an important day.
MR. DÍAZ: In what way?
MR. CHEN: It’s my daughter’s birthday.
MR. DÍAZ: How nice. I hope you all enjoy it.
MR. CHEN: Thank you. I appreciate your understanding.
Díaz laughs about the situation now. “I was quite certain he had said he was coming,” Díaz says. “And Mr. Chen was quite certain he had communicated that he absolutely could not come because he was going to be celebrating his daughter’s birthday with his family.”

The misunderstanding here comes from Mr. Chen living in guess culture, such that “it’s my daughter’s birthday” is a polite but firm way to say “I would like the day off, please”. Meanwhile, Mr. Díaz lives in ask culture, and thus assumes that, if Mr. Chen had actually wanted the day off, he would have asked for the day off.

1.4.3 I think the terms “ask” and “guess” are somewhat misleading

If Mr. Chen were to read what I just wrote above, he might object as follows: “Huh? What are you talking about? I did ask for the day off! Like, OK sure, I didn’t ask for the day off in so many words. But I certainly asked for it! And if I had been talking to someone from my own culture rather than a Spaniard, they would have understood perfectly well what I was asking for.”

So maybe “guess culture” isn’t really about guessing—it’s about asking in a more subtle way.

And conversely, even self-described ask-culture denizens also do their “asking” via conversational implicatures rather than via literal words;^[8] it’s just that the conversational implicatures they use are much less subtle than the guessers—indeed, they are so non-subtle that you probably don’t even realize that conversational implicatures are involved at all, unless you’re specifically thinking about it, or unless you are (or you’re interacting with) an extraordinarily-socially-oblivious^[9] person.

So, at least to some extent, I think calling something “guess culture” is a way to say “you are communicating your requests in such a subtle way that I cannot understand them”. And that’s not just about how people are communicating, but it’s also about you, and how tuned in you are to conversational cues. (For example, I happen to be a pretty-socially-oblivious guy, so I see “guess culture” everywhere!)

In other words, I claim that ask culture and guess culture are united in the idea that we should interpret lots of utterances as implicit communication of preference-strength within a negotiation, not statements about the world that might be true or false—consider things like “It’s no problem at all!” or “I’ll expect the report by noon tomorrow.” Ask culture and guess culture are in agreement that this is the game we’re playing; they merely offer different rulebooks for translating between utterances and preference-strengths. If you’re unhappy about the game itself, you and your close friends can try collaboratively experimenting with avant garde alternatives like “tell culture”.

1.4.4 I think plausible deniability is mostly orthogonal to ask-vs-guess

Plausible deniability is frequently brought up in the context of ask-vs-guess—see for example “Conversation Deliberately Skirts the Border of Incomprehensibility” by Scott Alexander (2017), or “The Logic of Indirect Speech” by Steven Pinker, Martin Nowak, and James Lee (2008).

Before we get into the relationship between plausible deniability and ask-versus-guess, here’s how I think about plausible deniability more broadly. I think plausible deniability is a special case of a more general dynamic where you are concerned about possible bad consequences from “leading too much” / “being too pushy” in a certain context or topic. There are some cases where expressing a preference at all, no matter how weakly, might put you at risk of being perceived as “leading too much”. An example, following Pinker, is that you might have an object-level preference to bribe an official, but you’re not sure whether they’re corrupt or not, and some officials may treat even a very weak desire to bribe them as violating a taboo. In this kind of situation, your least-bad option might be to say something that leaves the listener with 55% credence that you are expressing a nonzero preference and 45% that you aren’t. If the listener immediately reacts negatively, then oh well, at least the blowback shouldn’t be too bad. If they don’t react negatively, you can try slightly less subtlety, so that the listener shifts to 70%/30%, and then see how they react, and so on.

I think plausible deniability is sometimes relevant, but not usually. Usually we want to be polite, but we also want the listener to actually know what our preferences are, rather than having the listener feel genuinely confused and uncertain. (And if the listener is not genuinely confused and uncertain, then the “deniability” is not in fact “plausible”!) So, there are definitely some cases where plausible deniability is a central factor (as in the bribery example above), and there are many other cases where it’s a consideration on the margin (e.g. if I think there’s a tail-risk that the listener might find my preference very hurtful, then maybe it’s noticeably better if the listener is only 95% rather than 100% confident that I actually have that preference? Maybe it helps everybody cope?). But by and large, I don’t think plausible deniability is the main reason that people use indirect speech, nor the main reason that guess culture exists (which I’ll get to in §1.5 below).

In fact, I’ll make a stronger claim that ask-vs-guess is basically orthogonal to plausible deniability:

Diagram illustrating my claim that plausible deniability is mostly orthogonal to ask-versus-guess

I’ll elaborate with two sub-claims:

A. “Guess culture” utterances are often intended to be unambiguous, not plausibly-deniable. I already gave an example above (§1.4.2): “…Mr. Chen was quite certain he had communicated that he absolutely could not come because he was going to be celebrating his daughter’s birthday with his family.” Meyer’s book has several other examples along the same lines, including this one (where she’s quoting a Mexican manager at an international pharmaceutical company):

…Following a miscommunication between one of my Mexican team members and his Saudi colleague, I spoke with each of them about what had happened. The Mexican told me, “I made it known, so he could see it if he wanted to see it.”…

These people were not intending to leave the listener genuinely confused and uncertain in order to maintain plausible deniability. They were trying to communicate politely but unambiguously. Hence the common expression of exasperation: “Jeez, can’t you take a hint?!”

B. Plausible deniability is used in “ask culture” too. Suppose that Beth says “I had pizza yesterday, but I’m happy to have it again”. In ask culture, she is probably expressing a desire not to eat pizza while maintaining plausible deniability.^[10] Whereas the same utterance in a guess-culture context would probably lack plausible deniability—everyone would get the message loud and clear that Beth is voting against pizza. It would be common knowledge to everyone who hears her words; no guess-culture listener could deny it with a straight face. (See the figure above for how the same utterance may have plausible deniability in ask-culture but lack it in guess-culture.)

1.4.5 Tug-of-war analogy

I find that a tug-of-war is a nice mental image when thinking about zero-sum interactions. In particular:

If both sides of the tug-of-war are pulling the rope extremely gently, then someone can get extra rope onto their side by pulling with merely moderate force.
If both sides of the tug-of-war are pulling the rope quite hard, then you need to pull out all the stops and tug with all your might, to get extra rope.

I claim that this is basically the spectrum from “guess culture” to “ask culture”:

In a guess-culture context, the normal expectation is that both people will be excruciatingly polite, deferential, and indirect (in the sense of hinting at their desires rather than stating them outright). If you’re trying to “lead” in such a context, or to express an unusually strong desire, then you would do that by acting slightly less polite, deferential, and indirect than normal.
In an ask-culture context, the normal expectation is that both people will be direct and explicit about what they want, even for weak desires. If you’re trying to “lead” in such a context, or to express an unusually strong desire, then you would do that by making your demand emphatically, or perhaps even raising your voice.

1.5 Explaining cross-cultural differences

1.5.1 Another toy model: Communication cultures may result from “arms races”

In friendly tug-of-war, if each side wants to be pulling a bit more strongly than the other, then we rapidly get to a place where both sides are pulling with all their might. Conversely, if each side wants to be pulling a bit more weakly than the other, then we rapidly get to a place where neither side is pulling at all.

By the same token, if Alice is trying to be A% leader, and Beth is trying to be B% leader, then, if $A + B > 100 %$ , then the two of them will arms-race towards “ask culture”, and if $A + B < 100 %$ , then the two of them will arms-race towards “guess culture”. And if this happens a lot, the result will congeal into a widespread cultural norm—part of our unspoken expectations of what “leading” and “following” are supposed to look like in different circumstances.^[11]

I guess both directions of arms-races can happen. As an example of an arms-race towards “ask culture”, I tend to think of a boisterous group of fraternity brothers trying to shout over each other about which movie to watch.

But I think the more common pattern in culture-at-large is an arms-race towards “guess culture”, thanks to the fact that (at least in many cultures / contexts) people want to err on the side of leading too little, because leading too much can have serious negative consequences (you can hurt people’s feelings, get a reputation as a jerk, etc.), whereas leading too little is not such a big deal.

These kinds of arms races presumably continue until they hit their limits:

The extreme ask-culture limit occurs when you burn through all the most emphatic words in your language (cf. “semantic bleaching” of words like “literally” or “awesome”), and tire of raising your voice.
The extreme guess-culture limit occurs when miscommunication becomes intolerably frequent—hence the observation that “conversation deliberately skirts the border of incomprehensibility”.^[12] The latter limit would result from people trying to make their desires known in such an extremely subtle way that it flies right over the head of the person they’re talking to, in which case, they might repeat their request in a slightly less subtle and more pushy way.^[13]

In The Culture Map, Erin Meyer helpfully ranks 26 countries on the ask-to-guess scale, based on her experience as a business consultant specializing in resolving culture clashes within international business ventures. (Obvious caveat: these are broad trends, and nobody is denying that there is lots of variation across smaller-scale regions, groups, people, and situations.)

Image from *The Culture Map* by Erin Meyer. I added the red text suggesting that her anthropological terminology “low-context versus high-context” is similar to “ask versus guess”, or at least related, I think.

For example, she labels the USA as the most ask-culture, and East Asian countries as among the most guess-culture. We might wonder about the root cause: why do different countries fall where they do on the scale?

Meyer suggests an explanation based on the miscommunication issue above. East Asian countries have relatively high cultural homogeneity, and long shared history, and thus they can make requests in an extremely subtle way before hitting the limit of miscommunication. By contrast, anglophone countries tend to have much higher rates of immigration and of non-native speakers, and therefore extremely subtle and indirect communication is often impossible to pull off in the first place.

I think Meyer makes a good case, and that what she says is probably part of the story. I think another part of the story is that violating norms by trying to lead too much (a.k.a. being rude and pushy) is generally seen as an extremely grave error in East Asian countries, more so than in the USA. Therefore, East Asians would have a stronger incentive to err on the side of following / being less pushy, even at the risk of miscommunication.^[14]

1.5.2 Combat-vs-nurture is different from ask-vs-guess

Continuing with Meyer’s book, she draws a strong distinction between ask-vs-guess on the one hand, and the style of negative feedback and confrontation on the other hand.^[15] I am calling the latter “combat culture versus nurture culture” following this 2019 blog post by Ruby.

Sometimes these two axes go together: for example, German and Dutch cultures tend to be direct and explicit when asking for favors and when giving negative feedback; and conversely, East Asian cultures tend to be indirect and understated in both situations.

But in other cases, they come apart. For example, compare the USA on the one hand, to Israel, France, and Russia on the other hand. If I understand correctly, the latter three cultures tend to be comparatively subtle and indirect when asking a friend or colleague for a favor, but among the world’s most direct and explicit in negative feedback. And conversely, USA culture leads the world in being direct and explicit when asking for favors, but tends to be comparatively indirect, understated, and averse-to-confrontation when criticizing.

(She offers an amusing anecdote of a French employee receiving feedback from her USA boss. The boss kept trying to tell her that she was doing poorly, but she interpreted his words as gushing praise!^[16])

Anyway, I think there are two quite different dynamics here. If you overdo negative criticism, it can be interpreted as blame. If you overdo “leading”, it can be interpreted as belittlement. These are both unpleasant, but they are not the same! It’s perfectly possible to think someone ought to be mostly following, while bearing them no ill-will whatsoever—think of a parent-toddler relationship.

In fact, in a perfect world, nothing I’ve written in this post would have anything to do with disagreements over how to make progress towards a shared goal. (Such a goal might be “we figure out the correct answer”, “the client presentation goes well”, etc.). After all, in such disagreements, both parties would have the exact same object-level desires (i.e., the same utility function $U_{A} = U_{B}$ ): namely, the desire for the shared goal to actually happen. So there isn’t necessarily the same zero-sum aspect to the interaction that there is in the context of negotiating favors.

Instead, the relevant issue is that I might judge someone’s beliefs or actions as not only mistaken, but so mistaken that “they should have known better” and thus I now think much less of their judgment and competence than I did before. There is an arbitrary degree-of-freedom in how I would communicate that this threshold has been crossed. Here’s a diagram to illustrate:

Obviously, there is really a smooth continuum between “combat culture” and “nurture culture”, as the gray wedge on the left moves up and down.

So again, there’s a degree of freedom in what disagreement-related words and phrases are sufficiently strong as to imply judgment and blame, and this degree of freedom is evidently set quite differently in different cultures, and moreover this is not the same degree of freedom from the one I was discussing above that translates between pushiness and leading.

As above, there’s an interesting question regarding why different cultures wind up at different places on the combat-versus-nurture spectrum. As mentioned, the USA is in the middle, much more directly critical than places like China and Japan, but much less directly critical than places like Israel and continental Europe.

I’d guess that the East Asia story is similar to the discussion above, i.e. a combination of (1) severe social censure for being perceived to blame / judge someone whom it’s not your proper role to blame / judge, plus (2) enough cultural and linguistic history and homogeneity to enable people to pull off subtle communication without rampant misunderstandings.

But in the other direction, I don’t know why, say, a group of friends or coworkers in the USA will typically disagree with each other in a much more indirect and understated way than a group of friends or coworkers in Israel or continental Europe. Of course, there might not be any grand elegant structural explanation; it might be somewhat random and contingent.

1.6 Conclusion

Hopefully the above discussion is helpful in thinking more clearly about everyday “negotiations” over favors, joint decisions, delegation-of-responsibilities, and so on. I’m very interested in feedback—the comments section is below. Among other things, I’m especially worried about typical mind fallacy, and the parts where I speak in generalities about other cultures around the world that I have no personal experience with.

As mentioned at the top, everything in this post is centered on object-level preferences. The next post will move to people’s preferences about leading and following itself, which gets us into dominance, prestige, blame, offense, “status” itself (finally), and more.

(Thanks Rafael Harth, Seth Herd, Linda Linsefors, Justis Mills, Charlie Steiner, and Rasmus Herlo for critical comments on earlier drafts. Thanks Rafael Harth for his criticism of my first attempt to talk about social status, which I found sufficiently compelling that I basically wound up restarting from scratch.)

^
Some context for how I came to write these posts: While I often write about neuroscience and brain algorithms, these two posts have essentially none of that. They’re just about systematizing everyday behavior and folk psychology, and I hope they will be generally useful as such. As it happens, my own larger project is to understand the neuroscience underlying social status behaviors (as part of this even larger project related to AI alignment). But I have no hope of figuring out the neuroscience underlying social status behaviors, if I don’t understand social status behaviors in the first place. Hence these posts.
I previously attempted to talk about social status a couple months ago here. I still think I was pointing towards something important and true in that old post, but it was just one little piece of the puzzle, and I described it very poorly because I was confused about the bigger picture. Anyway, I neither expect nor recommend that you read that; these two posts will hopefully be self-contained.
^
Why am I switching terminology? Two reasons. The less important reason is that mine is slightly more concise, which is nice given that I will be using these terms approximately 10 gazillion times in these two posts. The more important reason is that I find “follow / lead” to have less misleading connotations for my purposes, specifically because it feels a bit more agnostic about what the person is trying to do. In particular, “make yourself small” kinda connotes a deliberate act, but I want to apply the term to lots of other situations too. For example, it is possible for a person to “mostly follow” by getting ignored and steamrolled despite their best efforts. In that case, they’re not really “making themselves small”; rather they’re being made small by the other person. Granted, my preferred terms don’t have perfect connotations either, but I think they’re marginally better.
^
Needless to say, everything in this post applies to people of all genders, but I gave the characters real names to make the scenarios a bit easier to visualize.
^
Note that “Beth is paying attention to Alice” is different from “Beth is visibly watching Alice”. For example, it’s possible to pay attention to someone without directly looking at them, by stealing surreptitious glances. If Beth is in fact visibly watching Alice, especially with eye contact, then that act itself would constitute a communicative social signal from Beth to Alice—a signal which Beth might or might not want to be sending right now.
^
I guess technically, the way I defined it, it’s “constant-sum” rather than “zero-sum”, but that’s not an important difference.
^
There’s an edge-case where both people in a two-person interaction can “mostly follow” simultaneously—namely, when both people are trying so hard to be deferential, that they end up doing something neither had any desire to do, each thinking incorrectly that they are going along with the other person’s object-level preference. This kind of catastrophic miscommunication is pretty rare in my experience, but it does happen.
^
I think a somewhat-more-elegant toy model might look something like the following: Alice’s object-level preferences are $U_{A}$ , and Beth’s are $U_{B}$ . Alice’s all-things-considered preferences are $U_{A} + α U_{B}^{'}$ , and Beth’s are $U_{B} + β U_{A}^{'}$ . Here, $U_{A}^{'}$ & $U_{B}^{'}$ represent Beth’s current beliefs about Alice’s desires and vice-versa, and the $α, β$ parameters represent how much Alice cares about Beth’s object-level desires and vice-versa. The latter could arise from admiration of the other person, fear of pissing them off, or various other considerations discussed in the next post.
If Alice states a preference, she is really requesting / demanding that Beth increase her $α U_{B}^{'}$ —i.e., Alice is saying that either Beth should be more motivated to accommodate Alice’s preferences on this topic (i.e., Beth should increase $α$ ), or Beth is mistaken about how strong Alice’s desires are here (i.e., Beth should increase this entry of $U_{B}^{'}$ ), or some combination of both. And the more that Alice’s preference-statement is “pushy” (see §1.4 below), the more forceful Alice’s request / demand becomes.
Interested readers can try to sort out the rest of this story, assuming it hangs together, or fix it up if it doesn’t. And then please tell me how it turns out. For my part, I think the simple models I’m using in this post are good enough for the things I want to say.
^
Here’s an example scenario: Alice says to Carl, “I wish you would leave”. Then Carl replies “OK” but doesn’t leave. Then Alice says to Carl “Go away now!” Then Carl says “Sure! Why didn’t you just say so?” Carl walks away, muttering to himself, “Jeez, why does Alice have to be so guess-culture?!”
As ridiculous as this might sound, I will now try to defend Carl’s point-of-view in this interaction. We imagine that Carl is extraordinarily oblivious to conversational implicatures, but understands the literal definition of words. When interpreted literally, “I wish you would leave” is informing the other person of a fact about your object-level preferences. You are not explicitly telling them to do something. Think about it: It is not remotely true that, in general, if Person X has an object-level preference for Person Y to leave, then it is always rude for Person Y to stay. There are tons of situations to the contrary—I’m probably in a situation like that multiple times each day. It’s just that, under normal circumstances, none of those situations would involve Person X voicing this preference out loud.
Thus, everyone (except Carl) is aware of a widespread conversational implicature that the act of saying out loud “I wish you would leave” is a strongly pushy speech act—it’s not a mere “FYI” about an objectively-very-common circumstance that’s happening in Alice’s brain, but rather it is to be interpreted as Alice making a strong demand upon Carl.
^
I use the word “socially oblivious” in various places in this post. The term “oblivious” has a negative connotation, but that is not my intention. I hope it goes without saying that some people are much less attentive to subtle conversational implicatures and cues through no fault of their own, including people with autism, and people who are not native to the language and culture. We should all try to be understanding and helpful. For what it’s worth, I myself am pretty “socially oblivious”—I dunno, maybe 90th percentile.
^
Guess-culture readers might not believe me when I say that “I had pizza yesterday, but I’m happy to have it again” gives plausible deniability to Beth expressing a preference, if she’s in an ask-culture context. To spell it out, the two interpretations would be “I had pizza yesterday, so obviously I’m not too jazzed about having it again, but I can deal”, versus, “I had pizza yesterday. For other people, that fact would imply that they don’t want to have it again. But not for me! That’s just not how my taste buds work—I don’t tire of food like that. I am just as happy to have pizza today as I would be on any other day.”
^
I think a more realistic picture would be: in any given individual interaction, there is a mutually-understood background cultural norm for the appropriate pushiness to display in different situations. Then if an arms-race happens, it will probably not escalate too far away from that norm—nobody wants to act too weird. But even if each individual arms-race is somewhat constrained by the gravitational pull of current cultural norms, they can still collectively move the cultural norm, bit by bit, insofar as the millions of everyday arms-races systematically tend to pull in a consistent direction.
^
The “conversation deliberately skirts the border of incomprehensibility” maxim arises from both arms-race-to-guess-culture and plausible deniability (§1.4.4). Think of it as two stages. In the first stage, a culture develops its conversational conventions—i.e., a way to translate from an utterance (including words, tone-of-voice, etc.) to a strength-of-preference, in a particular context. Here, an arms-race-to-guess-culture will result in those conversational conventions being as subtle as possible while still allowing most speakers to clearly communicate their preferences. Separately, whatever the conversational conventions are, people will sometimes dance around the boundary between what that convention declares to be “expressing a very weak preference” versus “not expressing any preference at all”. These two factors work together to make life difficult for non-native speakers, socially-oblivious people, etc.
^
As a somewhat-socially-oblivious person in the most ask-culture country on Earth, I think I frequently miss subtle communication, and thus people who talk to me need to be more pushy, or perhaps just wind up thinking I’m a jerk. I can only speculate that there must be entire universes of communication subtlety that I cannot even fathom, dancing underneath my feeble powers of perception, much like the UV coloration of flowers that only bees can see.
^
Related: this article has a suggestion that the tendency for “southern culture” in the USA to be more “polite” (guess-culture) is connected to the greater tendency for people to react violently to perceived disrespect. H/t this old comment by Stefan Schubert.
^
I’m taking liberties with Meyer’s claims in a couple ways. First, in the book, she uses the terms “low-context” versus “high-context” instead of ask-versus-guess; as mentioned previously, I think they’re either synonymous or closely related. Second, she has two different disagreement-related categorization schemes: one that she calls “direct versus indirect negative feedback”, and the other which she calls “confrontational versus avoids confrontation”. But then she ranks every country almost identically along those two scales. So I think those are basically pointing to the same thing, and I am merging them together for the purposes of this post.
^
Similarly, see the humorous “Anglo-Dutch translation guide”. But note that this “guide” is a mix of combat-vs-nurture and ask-versus-guess, since the Netherlands is apparently more direct than the UK along both those axes.

What links here?