You can't signal to rubes

The word 'signalling' is often used in Less Wrong, and often used wrongly. This post is intended to call out our community on its wrongful use, as well as serve as an introduction to the correct concept of signalling as contrast.

"We choose to go to the moon. We choose to go to the moon in this decade and do the other things, not because they are easy, but because they are hard.."

- John F. Kennedy

Why do peacocks grow such large, conspicuous tails? Why do people take degrees in subjects like Philosophy or Classics, despite these subjects having no obvious practical value? Why do people take pains to avoid splitting infinitives, even though everyone can understand split infinitives perfectly well?

These activities seem completely pointless, costly and difficult. Paradoxically, it is probably this very difficulty that serves to explain why they are done at all. Take the peacock’s tail. A peacock that has to struggle to survive while dragging around a conspicuous tail is clearly at a disadvantage. But if he can continue to survive, then clearly he must be pretty strong! So the peahens may choose to mate with him rather than the peacocks with less conspicuous tails, whose survival is thus a less impressive feat.

As for classics, getting a degree in classics may be pointless, but it’s also difficult. It requires one to read and memorize vast chunks of text, and to translate these texts between Greek, Latin and English precisely. So a person who has a degree in classics and got a good mark must be a person with a good memory who is able to execute tasks precisely. Qualities extremely useful in a civil servant, the occupation where many budding classicists find themselves. The rule that you mustn't split infinitives derives from Latin where splitting infinitives was impossible. So a person who doesn’t split infinitives is more likely to be a Latin scholar, with the qualities of class and intelligence that such a thing implies.

Even the decision to go to the moon might be explained in this way. Carl Sagan made the point that a rocket capable of going to the moon is certainly capable of reaching Moscow. And it’s clear why Kennedy in the middle of a Cold War would want to demonstrate such a thing.

When we explain a behaviour in this way, we say that the behaviour is signalling. The agent does not perform a task for its own sake, but to show others that they possess some important quality such as strength, a good memory, or military supremacy. The key features that a behaviour must possess for signalling to be a good explanation are as follows.

  1. The behaviour seems pointless. This of course, is a matter of perspective. Peacock’s tails are beautiful, classics is interesting, and the moon landing was a crowning moment of awesome from humanity. Peacock’s tails seem pointless when viewed as struggling to survive and reproduce, classics when viewed as a matter of finance, and the moon landing when viewed as a matter of military strategy. Priorities that genes, most people, and the American government might be expected to have.
  2. The behaviour requires a certain quality. Going to the moon requires a good rocket, surviving with a conspicuous tail requires strength and cunning, getting a good grade in classics requires a good memory. This is actually a stricter condition than necessary. All we need is for the quality to lower the cost of the behaviour. In the absence of superior rocket technology, a moon landing could be faked, but only by taking extraordinary and costly steps. This was argued by Messrs Mitchell and Webb here.
  3. You want others to believe you have this quality. Peacocks want to look strong and sexy, civil service applicants want to look intelligent and diligent, America wants to look innovative and powerful.
  4. Dishonest signalling isn’t worth it. By dishonest signalling, I mean enaging in a signalling behaviour when you don’t possess the quality you wish others to think you have. A weak peacock who grows a conspicuous tale will be eaten by predators. A stupid person who tries to get a degree in classics will fail his subjects. Faking the moon landing carries the risk that the conspiracy will be exposed and the American government would become a laughing stock. A good way of remembering this criterion is the slogan "You can't signal to rubes."


Unfortunately, not all proposed explanations involving the word "signalling" take care to establish these four properties. Our community seems especially guilty of this. The main misunderstanding is that it uses ‘signalling’ merely to denote behaviours that trick rubes in to thinking you’re good. This raises the question of why there are rubes to trick in the first place. Why haven’t more savvy competitors eaten their lunch? Here is an example of someone thinking that you can signal to rubes:

In other words, it's all about signaling, isn't it? Managers will take actions that actively harm the continued progress of the project if that action makes them look "decisive" and "in charge".  I've seen this on many projects I've been on, and it took me a while to realize that my managers weren't stupid or ignorant. It's just that the organization I was working in put a higher priority on process than on results. My managers, therefore quite rationally did things that maximized their apparent value in the eyes of their bosses, even if it meant that the project (and, as a result) the entire organization was hurt.

Here the rube is the managers bosses, why are they so stupid as to think that mismanagement is evidence of superior management qualities? Why haven’t these idiots been sacked? (This probably does occur in real life, but I don’t think "signalling" is the right term to describe it. I would describe it as "pandering to the prejudices of idiots".)

Another comment which falls in to the same trap:

Compare the skilled butcher, who, with no wasted movements, cuts his meat just where the joints are, and the flashy butcher, whose flourishes make for less skilful and efficient cutting but send a more impressive signal.
I agree that the flashy butcher could became engaged in his cutting and lose consciousness of the crowd and his impression on it without decreasing his signalling behaviour. If he did so, he might become more sincere, but his signalling behaviour would remain. For signaling is not a conscious addition to his art, which might strip away: skill at cutting and skill at signalling are woven confusedly together in it.

The trouble here is that it postulates stupid customers, just like the previous comment postulated stupid bosses. A much better test of butcher quality than flashiness is how good the meat tastes and how much is produced. An intelligent customer can probably test this fairly easily, and would not buy meat from the flashy butcher.

These uses of "signalling" at least have the advantage that they’re explanations along economic lines. The difference between signalling and pandering is the intelligence of your audience. What’s worse is that some people in our community use the word "signal" to mean "show" or "pretend".

An example:

I may have learned that signalling low status - to avoid intimidating outsiders - may be less of a good strategy than signalling that I know what I'm talking about.

A low status person that knows what they’re talking about? I suppose such things are possible... Seriously though, "signalling" is being used to mean "tricking people in to thinking that you are". Either you know what you’re talking about or you don’t. At least one of the two options given in the quote will result in you trying to trick someone. We’re signalling to rubes again.

Worst of all, some people use "signalling" as a version of ad hominem. "You just say that to signal." A comment to Overcoming Bias’s controversial post "Gentle, Silent Rape" reads:

“I'm sad but not at all surprised that so many of these comments are to the effect of "HOW DARE YOU EVEN THINK ABOUT THIS TOPIC?!!?!?"
I can only imagine how much more frustrating it must be for Professor Hanson have to deal with increasingly harsh anti-mind climate that envelopes the Western world. The all-encompassing ideology of political correctness is truly a fearsome juggernaut.
I think the status signaling arguments are right on the money, and the rest of the comments serve as the proof.”

Let’s go through the criteria again:

1. The behaviour seems pointless.

This clearly doesn’t apply. The behaviour is easily explicable. Comments might be made out of genuine disagreement, or (more cynically) to intimidate Hanson and others away from making arguments like these in the future.

2.The behaviour requires a certain quality.

The quality proposed was "status", but outrage is cheap. Any fool can be outraged at a blog post mentioning rape. It doesn’t require exceptional intelligence, charisma, wealth, or feminist credentials. You could be homeless and leave an outraged comment just by going to a public library. You don’t even have to read the post.

3. You want others to believe you have this quality.

Well this seemingly applies. People do want to be thought of as being against rape, and high status. The only trouble is that many of the comments are left anonymously.

4. Dishonest signalling isn’t worth it.

This does not apply at all. Even a convicted rapist could leave an outraged comment.

Clear thinking requires making distinctions. Using the word "signalling" to mean "pandering", "tricking people", "showing", or "toeing the party line" does nothing but lead to confusion and muddle. If you’re going to use jargon, use it in its precise sense. That’s what is jargon is for, communicating precisely. Next time you feel like using the word "signalling", ask yourself whether the four criteria apply. Remember: You can’t signal to rubes.

115 comments, sorted by
magical algorithm
Highlighting new comments since Today at 7:06 PM
Select new highlight date

You've basically come up with four criteria that describe the use of the word "signal" in a highly specific context - traits that exist for pure signalling purposes in evolution or game theory - and then decided, arbitrarily, that this is the one true meaning of "signal." I do not think you have provided adequate evidence or argument to back this claim up.

If everyone around me is a Republican and I am not, it might make sense that I would do things that would signal that I am a Republican, even if these are very cheap and have obvious positive returns. Your definition would not allow this - if it is cheap and has obvious positive returns, it is not "signaling" to you. What you're saying is that if I send a birthday card to a coworker I hate, then I am not "signaling" that I like that person because it's too cheap to send the card.

It may make sense to speak of weak or strong signals, or reliable or unreliable or misleading signals. But you've arbitrarily said that the word applies only when a certain arbitrary threshold is crossed (your 2 and 4).

Incidentally, your theory might actually work if 4 were eliminated and 2 read "the behaviour is more likely to occur if you possess a certain characteristic than if you do not." This would cover my birthday card example - it's cheap, but I'm more likely to do it if I like the person, so it does signal liking the person. But this change would also fix the counter-productive manager. She's doing things that she is more likely to do if she is decisive and in charge. Since she's being evaluated on those criteria, and not "good manager-ness" - which is not generally observable - it would make sense that she would choose to give those signals rather than not. But revising the theory appropriately seems to nullify most or all of your objections.

I do think that cost asymmetry is a defining feature of signalling. To me, signalling is a way of getting around the problem of cheap talk. To me, a "cheap signal" is like an "unenforceable pre-commitment". It defeats the point. (Of course, many people talk about pre-commitments without actually discussing the mechanics of enforcement. I view this as a grievous omission.)

I probably was too absolutist in my criteria, they should probably be read with an invisible "ceteris paribus" attached to them. I'm happy to talk of weak and strong signals.

I want to keep 4. because I make the assumption that the audience does not be deceived. Employers do not wish to hire lazy workers, and it's in every worker's interest to say that they aren't lazy.

Regarding your birthday card example, I'd classify that as a white lie. Your coworker probably doesn't care too much if he's genuinely liked or not. Same thing with Republicanism. We could also call it a "cover story".

Re: Good managerness, what you're talking about is not signalling, but Gresham's law. Decisiveness is meant as a proxy for good-managerness. Of course good manager-ness isn't observable, that's why we would tempted to invoke signalling in the first place! "How can I show that I'm a good manager? I know, I'll act decisively!".

I do agree I was being too absolutist, I do not agree that I should modify the theory. It seems to me that once we do that, we're no longer talking about signalling as it was originally conceived. I don't know how to argue for that other than to gesture at the economics literature, which talks about deceitful employees and employers without the werewithal to sort the wheat from the chaff.

Annendum: Katja Grace discusses signalling here and uses "costly signal" and "signal" as synonymous with my version of "signal". (The previous version of this comment falsely attributed it to Robin Hanson, mental note: Always check the byline)

It may be possible to rescue the word "signal", but it's going to take an equally evocative word that covers what people think they mean by "signal". "Stealing associations" isn't going to work because it's not one word. Robin covers a lot of mileage with "affiliate" but many times when people say "signal" they don't mean "costly-signal" or "affiliate".

It may be possible to rescue the word "signal", but it's going to take an equally evocative word that covers what people think they mean by "signal".

The biological term is "mimicry", and it fits quite well. A mimic is a "false signaler" that relies on the fact that most signalers in its environment aren't false or that signal verification is costly, and therefore signal receivers are statistically better off trusting a signal than they are attempting to verify it. Mimicry only works within an environment where there is an honest signaller (which is called the mimic's "model"), since the strategy causes the signal to become more noisy.

Note that biologists still consider mimicry to be a form of signaling, so they don't seem to share the idea that a signal has to be honest.

But if we want to continue to use the term "signaling" and be specific about what kind of signaling we're talking about, the biological metaphor seems particularly fertile. Accurate phrasing would look something like:

Managers will take actions that actively harm the continued progress of the project if that action makes them look "decisive" and "in charge". I've seen this on many projects I've been on, and it took me a while to realize that my managers weren't stupid or ignorant. It's just that the organization I was working in put a higher priority on process than on results. My managers, therefore quite rationally did things that maximized their apparent value in the eyes of their bosses, even if it meant that the project (and, as a result) the entire organization was hurt.

"Managers who are unnecessarily hostile to their subordinates are attempting to mimic a recognized signal (dominance) of good leadership."

Ultimately, what we seem to be talking about is what happens when a signal gets de-coupled from what it signifies, and an optimization process begins to exploit that decoupling. If your current priors have come to associate signal "a" with fact A, and I want you to believe A about me, it's in my interest to send signal "a" whether or not A happens to be true about me.

[edited for clarity]

"Managers who are unnecessarily hostile to their subordinates are attempting to mimic a recognized dominance signal."

That isn't true (or at least doesn't follow). It isn't mimicry to actually dominate when that domination hurts oneself (or a particular goal). It's just a signalling behavior that happens to be counterproductive to some goal.

Successful dominance isn't what's being signaled; "successful dominance" IS the signal. "Good at leadership" is what's being signalled.

Successful dominance isn't what's being signaled; "successful dominance" IS the signal. "Good at leadership" is what's being signalled.

Saying that instead of what you actually said would have saved me a whole lot of trouble.

I thought that's what I did say; I apologize if I was unclear. My brain isn't what it used to be, so clear communication is sometimes a struggle.

The more I think about this, the less sure I am that I understand the distinction you're trying to draw here between "attempting to mimic a recognized dominance signal" and "a signalling behavior". Can you expand?

The more I think about this, the less sure I am that I understand the distinction you're trying to draw here between "attempting to mimic a recognized dominance signal" and "a signalling behavior". Can you expand?

  • "Dominance signal" corresponds to actions which indicate to others (be they lower in status, equals or those higher in status) that the signaler that they are in charge. Dominance is almost entirely mediated by signalling.
  • "Attempting to mimic" in the context introduced by ialdabaoth indicates that the motive of the behavior is "signal to a particular target audience of superiors that one is dominant over subordinates despite in fact not being dominant over subordinates".
  • The distinction is between on one hand any "attempting to mimic a recognized dominance signal" and on the other the set of all signalling behaviors that is any one of "not about dominance", "not motivated to deceive some class of observer" or "actually indicates or causes dominance within the relevant hierarchy".

If mimicry absolutely must be dragged in to describe an aspect of the scenario then it could be said that "Managers who are unnecesarily hostile to their subordinates are attempting to mimic a recognized signal of competent leadership" or something similar. It would still be wrong as a point of human psychology but at least it wouldn't be fundamentally muddled thinking about signalling, mimicry and dominance.

I'm not sure this follows. I have had managers who displayed unnecessary hostility towards me when in front of their own superiors, but had no proper dominance or control over me to speak of.

In one case, I held said manager in the palm of my hands and tolerated the hostility in light of the fact that his display of dominance to his superior increased said superiors' allotment of resources to our team, which in turn made my own life easier and overall more enjoyable even after factoring out the hostility itself.

(for context, I had an indirect veto power over that manager: I could credibly threaten to quit, since I wasn't dependent on the job and had a guarantee of having a new one secured for me within a week, while they stood to lose some productivity by having to hire and train a new employee up to my skill level - they had a lot more to lose than me and the costs of hiring and training stood to be far higher than the costs of adjusting for minor demands on my part)

On the other hand, I'm slightly more inclined than usual to believe I'm misinterpreting you, given my current state of mind and (lack of) awareness.

I'm not sure this follows.

Note that my own assertion was (with emphasis added):

That isn't true (or at least doesn't follow).

I am not claiming that it is not ever possible to mimic dominance displays in some context without actually having dominance. I am claiming that using behaviors that actually successfully dominate is not mimicking dominance. Said dominance may in turn be done for the purpose of mimicking something else.

I am not claiming that it is not ever possible to mimic dominance displays in some context without actually having dominance. I am claiming that using behaviors that actually successfully dominate is not mimicking dominance. Said dominance may in turn be done for the purpose of mimicking something else.

Ah, thanks. Oops.

"Attempting to mimic" in the context introduced by ialdabaoth indicates that the motive of the behavior is "signal to a particular target audience of superiors that one is dominant over subordinates despite in fact not being dominant over subordinates".

No, the motive of the behavior is "signal to a particular target audience of superiors that one is effective at leadership despite in fact not being effective at leadership. The specific signal that is interpreted as effective leadership happens to be dominance over subordinates, therefore someone who wishes to signal effective leadership will dominate their subordinates."

It's telling that our culture so intertwines the two (effective leadership and dominance over subordinates) that no one even remembers that the actual trait we're trying to signal is "effective leadership"; the mistake that everyone seems to keep making is that dominating subordinates is itself the desired trait.

No, the motive of the behavior is

Note that my replies were to the unedited version and would not make sense as reply to the current claim.

It's telling that our culture so intertwines the two (effective leadership and dominance over subordinates) that no one even remembers that the actual trait we're trying to signal is "effective leadership"; the mistake that everyone seems to keep making is that dominating subordinates is itself the desired trait.

I don't believe the premised claim that active attempts to signal effective leadership are the predominate motive or cause of dominance of subordinates.

I don't believe the premised claim that active attempts to signal effective leadership are the predominate motive or cause of dominance of subordinates.

But it was the original premise of the quoted post that Patrick was objecting to, so whether we believe that premise or not is immaterial. This discussion is about whether the word "signaling" was used correctly in that post, not about whether that post was factually correct. A discussion about whether the predominate motive for dominating subordinates is to signal effective leadership would be interesting, but seems like a bit of a distraction when we're trying to have a discussion about semantics and word choice.

I appear to be confused - I don't understand why the previous post was downvoted. What am I saying that is incorrect or inappropriate?

I wouldn't worry about a single downvote or two.

That said, while I didn't downvote (or, indeed, previously read) the comment, I would say that discussions purely about semantics and word choice are rarely worth the energy to have.

Well, now everything I've said over the past few days has been downvoted.

Am I not the right sort of person to be posting on lesswrong?

It's generally a better idea to criticize actions rather than people: question your actions before you question yourself (the latter leads to a mindset which is more resistant to changing actions). But in any case, the most likely hypothesis, given that many of your posts have only one downvote, is that someone is serially downvoting you.

You've done nothing wrong (that I've seen), either someone is misusing the downvote as a retaliation tactic or LW is just bad at voting (or both). I've been there. It clearly hasn't affected you over a long period of time, since your karma is in the positive hundreds (as opposed to mine, which hasn't hit the 20 shreshold). But to be sure, try to avoid dubious assertions/axioms (or if you have to use one, affix the phrase "according to my beliefs" or something similar).

By my reasoning (or more accurately, by the reasoning of my inner Death Eater), if someone is serially downvoting me as a retaliation tactic, then that means I've done something wrong (or more accurately, that I'm BEING something wrong).

No one post you've said (regardless of how bad it could've been) would ever justify downvoting everything in retaliation, because that's not what the downvote is meant to be used for. I can understand why you, as a rationalist, place doubt upon yourself by default, but there's no need to defend your (theoretical) aggressor. Well-kept gardens die by pacifism, some die by moderation.

Yes, but what I'm saying is, I think that the retaliatory downvoting is an indication that I'm one of the weeds. (I'm used to this from other fora, so it's not unreasonable to assume that I'm a weed here, too.)

I understand what you're saying, and I can't falsify your (claimed) experiences from other fora. Nonetheless, even if you are a "weed", staying on lesswrong has to benefit someone. Law of bivalence: either you're irrational or you're rational. If you're rational, others will benefit from your points you'll make. If you're irrational, you'll benefit from the rational points others make. Either way, rationality increases. Q.E.D. (please no constructionism objections)

My objections aren't constructionist; like I said, I have an inner Death-Eater that keeps screaming "Weaklings should suffer!".

Oh no, I wasn't referring to your earlier objections. I was just pre-empting the rejection of the law of bivalence that some bystander may invoke. On a semi-interesting side note, these recent comments haven't been downvoted, so your aggressor has either stopped or temporarily left (hopefully the latter).

Well, one thing I learned from my childhood - the more viciously self-deprecating and self-abusive I can be, the more likely it is that my aggressors will leave me alone and just watch.

Ah, nice bit of wisdom there. You evade all of your abusers except for one: yourself. Ironically we've worked our way back to the topic of the article, because that can be useful as a signal; "don't bother dominating me, I already have that covered".

That was partially deliberate; I derive wry amusement from taking things meta and then bringing them back around.

This is one of those "words are slippery" moments.

Signaling is about displaying traits that can be interpreted by an outside agent as correlating strongly with another desired or undesired trait.

In this case, the signaller (the manager) is attempting to signal a desired trait (leadership ability) by displaying a behavior that our evolutionary ancestry has primed us to see as correlated with it (domination of subordinates). If we accept the premise that we need to separate this kind of signaling from the kind of signal that someone with actual leadership ability would perform, then the closest analogy to what we're seeing here is mimicry (since we have a specimen that does not possess the leadership ability, but still possesses the traits that produce the domination signal).

If signals had zero possibility of error (i.e. no one ever falsely signaled), I suppose the word "prove" would be an appropriate replacement for "signal" (actual meaning). If it's non-zero, I guess "strongly support" or some close one-word equivalent could work. Is it better to rescue "signal" and find a substitute for the false meaning of signal, or to find a substitute for the true meaning of signal and let the word "signal" be used falsely?

We got the word signal from its technical meaning in economics and evolutionary biology. We should strongly avoid giving standard technical words non-standard meanings, or using a non-standard word where a standard one exists.

mental note: Always check the byline

If you never make a proofreading mistake you are doing too much proofreading.

To me, signalling is a way of getting around the problem of cheap talk.

To me, signalling through cheap talk is like spam: it rarely works, but the costs are low and there's a lot of it about.

Alternatively, I can think of two other variations.

  1. Talk is cheap, but not free in some cases. As a subset of this, you can signal for the benefit of one audience (possibly a minority of the overall population, or your own ingroup, by saying things that will meet with moderate or minor social disapproval from the rest of the population
  2. People sometimes expect costs to exist even when they don't.

I was thinking more of people who make boasts that 99% of the audience would consider ridiculous. If they are already low status, the cost is low and they can still get an RoI on the 1%.

Robin Hanson discusses signalling here

That particular post is by Katja Grace.

we're no longer talking about signalling as it was originally conceived.

The word "signal" dates back to the 14th century. The use of the word as a verb dates back to at least the 17th century. The specific meaning you are trying to use seems to have started in the mid-to-late 20th century. That's the issue. Signaling means what you say it means, but it also has a broader meaning. If I do an action that I wouldn't do were it not for the fact that others observe me doing it, it seems very likely that part of my motivation is signaling. The manager clearly qualifies for this, as she would not be "acting decisively" but for the fact that she is being observed. (I also think that Gresham's Law is the wrong one, or that you need a bit of an explanation to tie it into this behaviour, but that's besides the point; the fact that there is a more precise name for a problem does not make that the only name for the problem.).

Unenforceable pre-commitments are still precommitments. If I promise never to cheat on my spouse again, despite a long history of cheating, I've made a commitment. It's not a very credible commitment, but it still belongs in the set labeled, "commitments." If you define "commitment" to only count "credible commitment," you've essentially created a new word.

As with any debate over definitions, this can get circular rather quickly. My point is this: if you want people to use the word "signal" to mean something very specific, and to abandon the conventional use of the word, you need to provide a viable alternative definition, and you need to explain why it would be more productive to abandon the conventional use of the term. I do not think your definition is viable, because it necessarily involves an arbitrary cost threshold. Even if your definition were viable, I don't see how you've shown that there is a problem with the conventional use of the term. Yes, there are different types of signals that differ in important ways, but I don't see why this warrants completely changing how we use the term, rather than specifying weak vs. strong signals.

I think it is worth preserving a distinction between the specific kind of signaling Patrick describes and a weaker definition, because "true signaling" explains a specific phenomenon: in equilibrium, there seems to be too much effort expended on something, but everyone is acting in their own best interest. "High-quality" people do something to prove they are high quality, and "low-quality" people imitate this behavior. If education is a signal, people seem to get "too much" education for what their jobs require.

As in an exam problem I recently heard about: Female bullfrogs prefer large male bullfrogs. Large bullfrogs croak louder. In the dark, small bullfrogs croak loudly to appear large. To signal that they are the true large frogs, large ones croak even louder. When everyone is croaking as loudly as they can, croaking quietly makes a frog look incapable of croaking loudly and therefore small. Result: swamps are really noisy at night.

Or, according to this paper, people "expect a high-quality firm to undertake ambitious investments". Investment is a signal of quality: low-quality firms invest more ambitiously to look high-quality. Then high-quality firms invest more to prove they are the true high-quality firms. Result: firms over-invest.

In this sense, you can also signal that you are serious about a friendship, job, or significant other, but only where your resources are limited. An expensive engagement ring is a good signal of your seriousness -- hence, expensive diamond engagement rings instead cubic zirconium. Or, applying to college and sending a video of yourself singing the college's fight song is a good signal that you will attend if admitted, and writing a gushing essay is a cheap imitation signal of that devotion. Hence, high school seniors look like they spend way too much effort telling colleges how devoted they are.

So you might use signaling to explain why "too many" people get "useless" degrees studying classics, or why swamps are "too loud", or engagement rings are "too expensive". I don't think it's true that too many people pretend to be Republicans, or too many birthday cards or sent.

I think it is worth preserving a distinction between the specific kind of signaling Patrick describes and a weaker definition

What Patrick refers to is called costly signaling in evolutionary psychology, and I believe in general.

You've basically come up with four criteria that describe the use of the word "signal" in a highly specific context - traits that exist for pure signalling purposes in evolution or game theory - and then decided, arbitrarily, that this is the one true meaning of "signal." I do not think you have provided adequate evidence or argument to back this claim up.

Aren't you arguing over definitions here? If he had defined a wholly new term as having that meaning, would your objections still apply?

Aren't you arguing over definitions here?

You did realize that the OP is an argument about definitions, and thus a response that continues that argument is spot-on, right?

I hadn't actually read the OP, but it seems like proposing a formal definition of a term being used is rather different to attacking an opponents use of a term on the basis that they haven't provided "adequate evidence or argument" that their definition is the "one true meaning of 'signal.'"

My objections would indeed not apply if a new term were used. You can define a new term however you like; that's the point of making a new term. You can't just declare that a commonly used term has a specific meaning without providing some justification for abandoning its other existing meanings.

If I wanted to argue that the definition of "bachelor" is "an unmarried man," I could do so rather easily, by citing this for example. If I were arguing over what counts as "theft," I could offer an argument as to why a particular act should or should not fit under the general definition. An argument like the OP's could theoretically include evidence (of common usage, of confusion, etc.) or argumentation, but the OP's post does not really seem to do this. It declares, "The definition should be X" and then rejects certain usages as not fitting the definition. If you're using an extremely common word like "signaling," you don't get to arbitrarily redefine it.

If you're using an extremely common word like "signaling," you don't get to arbitrarily redefine it.

I'm not totally against redefining, or introducing a technical meaning for a common word that is used in some discipline, but doing that and then complaining about how other people are misusing the word is too much.

It's a word. It means whatever we agree it means. If he wants to introduce a new, ore precise set of criteria for calling something a "signal" then he's welcome to. However, reading the OP, he seems to think his definition is the only acceptable one, which is clearly nonsense.

EDIT: ninja'd by HalMorris

You're describing costly signaling. Contrary to your opening statement,

The word 'signalling' is often used in Less Wrong, and often used wrongly.

people on LessWrong are usually using the term "signalling" consistently with its standard meaning in economics and evolutionary biology. From Wikipedia,

In economics, more precisely in contract theory, signalling is the idea that one party credibly conveys some information about itself to another party

Within evolutionary biology, signalling theory is a body of theoretical work examining communication between individuals. The central question is when organisms with conflicting interests should be expected to communicate "honestly".

In particular, the ev bio article even includes a section on dishonest signalling, which seems to be what you're complaining about here:

Seriously though, "signalling" is being used to mean "tricking people in to thinking that you are".

This post is still interesting as a highlight reel of different examples of signalling, and shows that the term is, in its standard usage, rather non-specific. It's just not an illustration that people here are using it wrongly.

Well I'm happy to use "costly signalling". I was under the impression that costly signalling was signalling. If it isn't costly, at least for potential fakes, then I'm not sure how it can serve as an explanation for behavior. Why should I signal when the fakes can signal just as easily? What is there to gain? I think at the very least, there has to be some mechanism for keeping out cheats, even if it's rarity. From the wikipedia article on signalling theory:

" If many animals in a group send too many dishonest signals, then their entire signalling system will collapse, leading to much poorer fitness of the group as a whole. Every dishonest signal weakens the integrity of the signalling system, and thus weakens the fitness of the group."

But what am I? Some kind of prescriptivist? Evidently my understanding of the term is a minority, and people far cleverer than I don't use it my way. I'll stick to "costly signal" in future.

“No! I must resolve the muddle” he shouted

The radio said “No, Patrick. You are the muddled one”

And then Patrick was a zombie.

From the wikipedia article on signalling theory:

" If many animals in a group send too many dishonest signals, then their entire signalling system will collapse, leading to much poorer fitness of the group as a whole. Every dishonest signal weakens the integrity of the signalling system, and thus weakens the fitness of the group."

Did you just use the appeal 'weakens the fitness of the group' to predict or describe the signalling behaviors of individuals?

A lot of signalling is bad for the group, whether honest or dishonest. When it happens to be good for the group that is, well, good for the group but not something one should necessarily expect from an individual.

You're accusing me of group selectionism? We might disagree on a point of terminology, but come on, I'm not a completely nutter. Anyway, my point in quoting the wikipedia article is that too much dishonest signalling makes signalling completely pointless ('weakens the integrity of the signalling system'), so for signalling to work you need some way of keeping out the cheats. I'm not proposing anything as daft as "groups without cheats will prosper". Indeed, that's why I was making such a big deal about criterion 4 and cost asymmetry, because the analysis of signalling has to work on an individual basis, including the individuals that might be tempted to cheat.

In my limited imagination, the only way I could think of for keeping out the cheats was having an asymmetric cost structure for honest signalling compared to dishonest signalling. Thus cheating wouldn't be worth it. I now realize this is not the only way. ialdaboth called my attention to Batesian Mimicry, where cheaters are "kept out" simply by the fact that mimics are comparatively rare. Doubtless other ways could be invented.

I think I prefer MagnetoHydroDynamics definition of signalling, and would reserve my criteria for describing costly signalling.

You're accusing me of group selectionism?

No, I carefully avoided that particular charge because it doesn't strictly apply even to the author that you quote---at least not without additional context.

Nevertheless, thankyou for elaborating on which part of the quote you intended to emphasize. You are indeed a non-nutter.

" If many animals in a group send too many dishonest signals, then their entire signalling system will collapse, leading to much poorer fitness of the group as a whole. Every dishonest signal weakens the integrity of the signalling system, and thus weakens the fitness of the group."

Do you conclude from that tha lying is extremely rare in human society?

No. I think that because lying is common in human society, a credible signal must be costly to liars.

Huh? Liars are lying non-credibly? Liars can afford credibility? What?

people on LessWrong are usually using the term "signalling" consistently with its standard meaning in economics and evolutionary biology.

And psychology.

Both status and signaling as concepts have at least in some circles pervaded common language. This will inevitably cause new members to use these words imprecisely. I don't recall the exact quote, but Daniel Dennett has said about consciousness that everyone feels like they're an expert on it. I think this applies to signalling and status too once one learns about them, since they're such a constant and seemingly direct part of our experience.

I too now feel guilty of not studying these concepts more, and seem to have quite incompetently participated the discussion.

I looked at http://en.wikipedia.org/wiki/Signalling_%28economics%29 Huh? That isn't what "signalling" means! If that article is correct, it looks like a case of confusing terminology.

On this site, rube generally means 'red cube', and I had to look up the word to figure out what you meant here. Though this still makes a bit of sense--you can't signal to red cubes either.

To those wondering, the first definition for rube I found was "an unsophisticated person from a rural area; a hick."

I thought the same thing, but I guessed it was too silly for me to write it in a comment. Judging from the karma score of the parent, it seems like I was wrong!

A peacock that has to struggle to survive while dragging around a conspicuous tail is clearly at a disadvantage. But if he can continue to survive, then clearly he must be pretty strong! So the peahens may choose to mate with him rather than the peacocks with less conspicuous tails, whose survival is thus a less impressive feat.

I'm not sure that's true. Wikipedia lists several competing theories:

  • Sexual selection, first proposed by Darwin himself in this case: he thought peahens preferred peacocks with bigger tails, and so bigger tails evolved, regardless of expense in fitness.
  • Costly signalling, proposed by Zahavi, as you describe.
  • Some studies listed by Wikipedia claim that peacocks with smaller tails suffer more predation (mechanism unknown), and that females prefer males with certain tail patterns (many 'eyes') rather than size.
  • Some other studies confuse everything even further by claiming big, erect, flashy tails serve to intimidate predators and evolved through ordinary natural selection.

Now, how would one test this? Especially considering that some theories mostly differ in their explanations of the evolutionary history leading to the present; they all match the present birds. I do know that real-life biologists sometime claim peacocks as archetypical evidence for sexual selection, and others claim it as evidence for costly signalling of fitness.

Sexual selection, first proposed by Darwin himself in this case: he thought peahens preferred peacocks with bigger tails, and so bigger tails evolved, regardless of expense in fitness.

This isn't an explanation until it can explain why peahens prefer peacocks with bigger tails. That's what the signalling hypothesis is for. I don't see why these are considered competing hypotheses.

Do you suppose a cat's hiss tells another cat something about strength and fitness via lung-power?

That could explain their fear of vacuum cleaners and blow-driers.

I don't suppose that, and I assume you don't expect me to suppose that, but I don't know what you're trying to tell me.

Sorry, should have made that an independent comment, not really an answer to anything in this thread. Just a pet theory brought to mind by all this talk of peacocks and signalling. Cats threaten each other with a toneless sound of rushing air, and and are afraid of high-powered mechanical devices that produce a version of the same effect.

A surplus of resources is one thing that can be signaled, and resistance to parasites is another. They both involve signalling, but these ideas are a bit different from each other - and from sexual selection, which can magnify a wide range of "fashionable" traits.

Of course those things are different from sexual selection. Comparing a trait that can be signalled, to a mechanism by which traits can be magnified, is a type error.

I'm not saying that "sexual selection" and "costly signalling" are the same hypothesis, I'm just saying they aren't competing. One attempts to explain how a trait gets magnified, the other attempts to explain why.

It looks as though the post you were responding to was wrong to treat these as incompatible hypotheses.

No doubt, peackock tails are magnified by sexual selection, costly and illustrating parasite resistance. However, though compatible, these explanations do compete with each other a little - for example, when explaining particular features of the tails.

Some studies listed by Wikipedia claim that peacocks with smaller tails suffer more predation (mechanism unknown),

A predator which relies on vision for target identification might survivably assume that anything which appears large and flashy actually is physically large, with muscles and defenses to match, thus out of it's league, and look elsewhere for prey. A wide flat tail with a colorful pattern is probably the most cost-effective way to maximize your apparent size, at least from the front and back.

The word 'signalling' is often used in Less Wrong, and often used wrongly.

I recommend that people do not use this post as a guide to how to use the term 'signalling'. It presents false claims with authority and confidence.

These activities seem completely pointless, costly and difficult. Paradoxically, it is probably this very difficulty that serves to explain why they are done at all. Take the peacock’s tail. A peacock that has to struggle to survive while dragging around a conspicuous tail is clearly at a disadvantage. But if he can continue to survive, then clearly he must be pretty strong! So the peahens may choose to mate with him rather than the peacocks with less conspicuous tails, whose survival is thus a less impressive feat.

This is misleading. It idealizes and trivializes the signal given conveying the impression that it is 'strength' or (non-inclusive) fitness that must be signaled. However the peackock's tail is the go to example of Fisherian runaway. Even without considering strength-signalling implications the tails are signalling "if you mate with me your male offspring (and later generation male descendants) will likely also have enormous tails and be chosen as a mate by other hens". When it comes to sexual signalling, sexiness itself is one of the most powerful signals that can be sent. We don't need to rationalize that out into some objective signal of strength, health or virtue other than ability to successfully signal. In this (and some other social situations) signalling the ability to signal well is the primary point.

Clear thinking requires making distinctions. Using the word "signalling" to mean "pandering", "tricking people", "showing", or "toeing the party line" does nothing but lead to confusion and muddle.

To various degrees every one of those can be correctly described in terms of the signalling implications.

If you’re going to use jargon, use it in its precise sense. That’s what is jargon is for, communicating precisely.

Yes. And if you are going to preach about the usage of jargon it becomes even more important to get it right.

(I endorse Academian's description.)

A good way of remembering this criterion is the slogan "You can't signal to rubes."

This doesn't seem like it describes your fourth criterion; it reads more like "Rubes can't signal."

I do think that the claim "you can't signal to rubes" is a good one to keep in mind, but like Psychohistorian points out, that limits you to a narrow component of signalling, which I might characterize as spending resources on authority, which is only worthwhile if the audience can recognize authority. If a moviemaker decides to actually use an Bald Eagle cry instead of a Red-tailed Hawk cry, most movie-goers will think (TVTropes link) they got it wrong, and only ornithologists will notice it's correct.

A rube is a sucker, someone easily deceived.The slogan means that potential signalling explanations shouldn't assume that the receiver of the signals is stupid.

I have a hard time telling whether you're trying to say that 'signaling' models are inaccurate, or just that calling them 'signaling' is misleading. I agree with the latter insofar as 'signaling' means this specific economic model, because the behaviors in question aren't directed at economically rational agents. I also can't tell if you dislike models that postulate stupidity (the strong status connotations of the word "rube" make me suspicious).

If you mean the former: I think you greatly overestimate median rationality in your take on the manager and butcher examples. All positive traits get conflated with each other by default. People can and do override their affective impressions with explicit reasoning, but more often than not they don't, especially when evaluating performance is difficult — and it's almost always more difficult than evaluating "does this person look like a winner?".

I also used to think that simple non-costly signaling couldn't possibly stably work, but experience (often with my own irrationality) changed my mind. This is less confusing if I think of it as social-primate (rather than general-intelligence) behavior; liking things/people other people like is socially useful. (This would likely be significant in the manager example in real life, e.g., I'll look better to my superiors if I make similar evaluations of my subordinates to them.)

The quality proposed was "status", but outrage is cheap. Any fool can be outraged at a blog post mentioning rape.

Now, status signaling is overused as an explanation. If the "HOW DARE YOU" comments are signaling (or 'signaling') anything, the obvious thing is alignment with the perceived-as-socially-powerful (implicit-Schelling-point-)faction condemning Robin, not status.

I just mean the latter. I think explanations involving pandering can work. The trouble I have with models that postulate stupidity, is that they need people to be stupid in a convenient direction. Stupidity is a much larger target than intelligence after all. I think explanation involving pandering work if you can explain (like you did with the affect hueristic) why these tricks will work on people.

Out of curiosity, what are the connotations of the word "rube" that make you suspicious?

The trouble I have with models that postulate stupidity,

But I am not convinced that your examples actualy do that.

Here the rube is the managers bosses, why are they so stupid as to think that mismanagement is evidence of superior management qualities? Why haven’t these idiots been sacked?

The idiots are where they are because they have Won -- they have been playing the games of Climb The Coroporate Ladder and Look After Number One But Don't Make It Obvious quite succesfuly. It/s a lesswrongian prejjudice that the only game anyone would want to play is Highly Competent But Criminally Underappreciated Backroom Boffin. They don;t get sacked because their superiors are playing the same game according to the same rules.

You could object that companies where dick-swinging is appreciated more than achieving goals and targets won't have a long term future. Well, if there is someone in the chain who is playing Build A Company with a Lasting Future, then they're being stupid. But rationality is achiveing your goals. They've achieved theirs.

It/s a lesswrongian prejjudice that the only game anyone would want to play is Highly Competent But Criminally Underappreciated Backroom Boffin.

Yes. The general case of this prejudice is probably something like 'behavior morally should be evaluated according to its stated far-mode purpose; other purposes are possible and important, but dirty'. Of course, this has the large upside of making us seriously evaluate things according to their stated purpose at all....

Out of curiosity, what are the connotations of the word "rube" that make you suspicious?

Low status, contemptibility, etc. I expect making status hierarchies salient to make people less rational (hence fully generic suspicion), and I had the specific hypothesis that you might see people using 'signaling' models as judging others as contemptible and be offended by this.

Relatedly, I dislike calling the behavior in question "pandering", since I expect using condemnatory terms for phenomena to make them aversive to look at closely, and to lead to bias in attribution (against seeing them in oneself/'good' people and towards seeing them in 'bad' people, as well as towards seeing people who unambiguously exhibit them as 'bad').

Now that you mention it, I think this does occur, although I think most of the judgement is directed at the 'signaller' (or in my language 'panderer') for being vain or duplicitous, although I don't like saying I'm offended by it ("Offense is a sign of a weak and bourgeois mind" says my inner Dali.)

I think that 'pandering' does carry the connotations of how 'signalling' is used, but I'm happy to accept alternatives. One I can think of right away is "appealing to", and I'd be happy to switch from 'pandering' to 'appealing' if you like.

"Influencing" is pretty neutral, if not very specific. "Exploiting the halo effect" is too long, but precise.

Isn't signalling just the practice of having non-obvious characteristic A and then exhibiting obvious characteristic B because P(A|B) > P(A) && P(B|A) > P(B) ? Which means that B gives information of A.

Intelligence is non-obvious, a degree in classics is. Peacock strength and cunning is non-obvious to peahens, but big tails are. Having intercontinental missiles is classified information, but moon landings are sensational news.

It seems logically rude to create your own definition, declare it the one "true" definition, and then complain that others are using their own terms. The term "signalling" has been used consistently, judging by your examples of so-called misuse, and it is you who is refusing to interpret them correctly.

Furthermore, is it really that hard to postulate that people may associate certain traits with signals that do not require said traits to produce?

Well, that doesn't take into account definition drift. Like it sounds like 'costly signalling' got abbreviated to 'signalling' and then spread out while still sorta-referring specifically to costly signalling.

Upvoted for opening discussion on a concept way too carelessly used. Another such concept I would like to see scrutinized here is social status.

Thanks, I enjoyed and upvoted them both. The wiki page didn't have any posts on the concept of status. I tried to add these, but couldn't since for some reason couldn't verify my email account.

Batesian mimicry is a form of signaling in which dishonest signaling is, in fact, worth it - and it is absolutely still considered a signal.

Dishonest signalling isn’t worth it. By dishonest signalling, I mean enaging in a signalling behaviour when you don’t possess the quality you wish others to think you have. A weak peacock who grows a conspicuous tale will be eaten by predators. A stupid person who tries to get a degree in classics will fail his subjects. Faking the moon landing carries the risk that the conspiracy will be exposed and the American government would become a laughing stock. A good way of remembering this criterion is the slogan "You can't signal to rubes."

This seems false as a generalization.

Suppose that a double agent's loyalty is in question, so they choose to turn a sympathizer of their own country in to the government. Turning in a sympathizer of their country is certainly something that a person loyal to that country would have reason to be reluctant to do. It's likely to decrease suspicion that they're a double agent. But if it allows them to continue passing valuable information to their country, and to avoid being caught and punished themself, then it's probably worth it.

I don't know what it would make sense to call such behavior if not signalling, and dishonest signalling at that.

Why do peacocks grow such large, conspicuous tails? Why do people take degrees in subjects like Philosophy or Classics, despite these subjects having no obvious practical value? Why do people take pains to avoid splitting infinitives, even though everyone can understand split infinitives perfectly well?

The second example seems to fit in worse than the two others, given that plenty of people take degrees in such subjects due to a love of learning and due to being genuinely interested in them.

I think the thesis usually is that signaling is a product of evolution, so this shouldn't be a problem. The peacock genuinely doesn't have a choice growing its tail either and the females don't have a choice genuinely liking those tails.

IMO one had better avoid making it unclear which side of the evolutionary-cognitive boundary one is talking about.

I agree. Thanks for the relevant link. Now I just wish you would've made this comment directly on the op.

ISTM that the OP actually meant that cognitively rather than just evolutionarily; see this comment.

Why did he use the peacock as an example then? Does it make sense to use the word signaling on both sides of the boundary, especially if you're trying to constrain the concept?

I meant the peacock example evolutionarily. I got it from The Selfish Gene.

I thought so. I think it might be useful to have a different word for signaling on each side of the e-c boundary, since signaling clearly can't mean the same thing on both sides.

Why not? Can't we regard evolutionary signalling as completely analogous to cognitive signalling, just as played by genes over a much longer time scale?

But going by the evolutionary origin of behaviors makes the whole concept of signaling uselessly broad, which is the exact opposite of what Patrick was trying to do in this article. E.g. this article suggests that all of these behaviors could have their evolutionary origins in signaling:

We do many things which are not in a narrow sense instrumentally useful: we dance, joke, write poetry, throw parties, go on vacations, dress up in expensive fashionable clothing, watch and participate in sports, and so on.

I think some of those behaviours are still (i.e., cognitively) signalling (according to the OP's definition) for a large part of the people who engage in them. (How many people actually terminally value dressing up in expensive fashionable clothing? Hint: Do they also do that at home?)

I agree that the concept is uselessly broad, and probably could be ditched altogether without too much information lost. All of the behaviours in that quote could reasonably fit Patrick's criteria, which I guess goes to show that they're pretty weak. I appreciate the effort come up with such criteria, though.

If we add to the criteria that signaling should be of evolutionary origin, shouldn't it just constrain the concept even more?

I agree that the concept is broad and the phenomenon pervasive, and these siggest to me that the concept needs breaking down: we need words for costly-signalling, spam-signalling, practical-signalling and so on.

I agree. I think you can use signalling to explain this decision, but I wouldn't say that it's otherwise inexplicable. I guess I was being too cheeky.

The rule that you mustn't split infinitives derives from Latin where splitting infinitives was impossible. So a person who doesn’t split infinitives is more likely to be a Latin scholar

Are you being serious?

I don't understand the link between intelligence and signalling. It implies that only human beings signal, which contradicts the purpose of using the peacock as an example of real signalling.

Or maybe evolution really is smarter than us...

So, what message are you sending by failing to capitalize America? :P

A weak peacock who grows a conspicuous tale

This should say "tail", not "tale".

4 Dishonest signalling isn’t worth it.

I seem to recall seeing some game theoretic signaling models with wasteful pooling equilibria, where everyone sends the same costly signal and the receiver can't distinguish between the senders of different qualities. Can't find a specific model now, but here is a slide that mentions this fact on page 20.

One of the original tag-based cooperation papers features a scenario where signal-based cooperation persists - despite regular forgery of the signals used. See:

-Edit it looks like Academian said most of this already, I need to stop going afk mid-comment. I see several problems with this article. The most obvious is that it is arguing about definitions, and doing so in a situation were the majority of the participants seem to understand each other well enough that I don't think the argument is necessary for clarifying communications. The next is that I feel like the phrase "you can't signal to rubes" is confusing, and certainly not a good slogan. Lastly, to extent its meaningful to argue about the meaning of words what you say is wrong or at least not accepted in all disciplines. Based on my psychology classes and how I've seen the words used off of LW, what your four features loosely pinpoints is costly signaling. This is the type of signaling that people commonly are talking about when they say signaling, but it is just a subset of signaling which very much includes what you call "cheap talk" in one of the earlier comments. If I'm understanding wikipedia correctly this is also the case in biology. I'm not sure about economics or other disciplines. Even then most definitions only require 3 & 4. If trait A is something you want others to think you posses and doing behavior B without trait A is more costly than people thinking you have trait A is worthwhile, then B is a costly signal of A. Even if B is possible without A and/or it is worthwhile for people with A to do B even if nobody gets the signal.

Dishonest signalling isn’t worth it.

There are career criminals, so dishinesty can work. Dishonesty is not guaranteed to work, but "worth it" is statistical. BTW, many nations have been caught out in all kinds of bullcrap -- the Zinoviev Letter, Iran-Contra, etc.. It tends to get forgotten about. It's not like people have high expectations of politicians.