Signaling games aren’t exclusively for information, though—at least not if by “information” we mean “factual objective pieces of external data such as a bank account balance or employer.” They are also for manipulating people’s emotions, hopefully in positive ways, for creating opportunities to connect with other people in conversation, and for directing attention to the information that actually matters (as opposed to conveying raw information directly).
So for example, let’s say I’m a teacher giving corrective feedback about how to do a task during a lesson to a child. The information I want to convey is how to do the task.
But I need signaling to do things like:
Make the kid feel encouraged about their progress with the task rather than frustrated with their shortcomings so far
Creating small opportunities to get to know one another better as people, beyond the main goal of mastering the task
Refocusing the kid on what actually matters, such as shifting their attention away from assessing “is my teacher frustrated because I keep messing up?” (I very well might be, but trying to hide it and not let it affect my behavior), and toward assessing “is my teacher going to keep being encouraging and give constructive feedback even when I mess up?” (hopefully yes, if I’m doing my job right).
To me, signaling is about shifting where the emphasis is, or what’s top of mind, rather than conveying information. If we’re playing a board game, signaling is the part where you are manipulating each other into keeping it fun. It’s an active process, not a passive information transfer process.
In most cases the other player benefits from using a filter on the information you didn’t provide.
For example, seeing company financial reporting the way “warren buffet does” has a higher expected value than seeing them through the “emphasis” by company financial officers.
Similarly, if a small child could query their implants or external system to filter information by the highest EV to the child, that’s far better than what a human teacher offers.
Signaling is a conflict of interest and is not trustworthy information.
It’s unaligned to the interests of the observer. My hypothetical AR system that queries records on the target at least can be built to filter things to the observer’s benefit.
(though, note, information may not be statistically relevant. If hypothetically (assume the counter factual) it turned out in reality that people previously convicted of assault charges were less likely to commit further assaults (in a counter-factual world where rehabilitation worked), AR systems or implants might still emphasize this because it increases the perceived value of the AR system.
I don’t think we are being specific enough to declare what proportion of information hidden by signaling would be best revealed for the interests of the person receiving the signals.
We can certainly agree important cases exist where the signaling is bad for/unaligned with the person interpreting those signals. Or some subset of their interests that they maybe care about most heavily at that time.
No. I am claiming the incentives are misaligned such that the information presented this way will always be biased and is untrustworthy. Same way you can’t trust infomercials or pharmaceutical advertisments.
This is a correct and fairly stable generality.
Note the exceptions don’t matter. If statistically you know an information stream is heavily manipulated you can’t use it even in situations when it happens to be an honest signal. (because you don’t know the adversarial other agent is telling the truth This Time)
You seem to define signaling as a way of transmitting information other than the signaling mechanism itself. For example, if I have a stern face and a raised voice, those signaling mechanisms convey an inner emotional state of anger, and the possibility of punishment. You seem to be defining “signaling” as the conveyance of information about anger and punishment possibility, but as not including the stern face and raised voice.
I define signaling as often being the concrete good we care about, intrinsically for its own sake. For example, if my girlfriend smiles and laughs at my joke, I enjoy those pleasant expressions, not only because they convey her emotions or let me make predictions about the future, but also because they are a viscerally pleasant spectacle in their own right.
If signals can be a viscerally pleasant (or unpleasant) thing in their own right, then they aren’t zero sum. They are a product. We can manufacture more of them, or less, and get certain signal types to the people who are eager to consume them. We can enter into relationships where we find ourselves exchanging pleasant signals with each other.
Perhaps all we’re doing is implicitly arguing semantics (whether a ‘signal’ is the abstract information content, or the physical object that may or may not also convey additional information content)?
In my case I am arguing the issue with easily faked signals makes their information content suspect. I wouldn’t argue about your relationship with your girlfriend but this is why for example many high tier employers pay attention to only the signals that cant be faked. Which are the degree, name of school, name of prior employers and yoe, and score on interview questions. Everything else just gets ignored and I have gotten quarter million TC offers without a haircut and wearing some random Walmart polos. I assume the HFT half mil offers are no different.
Yeah I think we are discussing two entirely separate issues then. Here, I’m interested in people treating conventionally information-carrying signals as objects of interest in their own right. A smile can convey information (i.e. about emotional state), but we can also be interested in the smiles themselves, as when we take aesthetic pleasure in viewing a smiling face. Just a different topic.
Yeah I am just talking about how cheap information carrying objects are not really any more trustworthy a signal than say the annual report of a Bahama based cryptocurrency firm.
It’s not even relevant whether the signal is an object or a digital file. Just how likely it is to be honest, and the signal itself.
Oh yeah I (almost) completely agree with you there. Basically if it costs people less to game the information-carrying signal than the benefit they’ll get from gaming it, then it will get gamed.
It’s just that the original discussion was about clothes and how wearing nice clothes can give you a nicer life, and I think the reason is that you look better, so people want to associate with you more because they find you aesthetically pleasing, and I don’t think that’s bullshit or even really primarily the kind of information-carrying signaling we’re talking about in this subthread. I also think that clothing choices give people something to talk about, and people enjoy the pleasure of that conversation, and that visceral enjoyment is a hard-to-fake signal of group bonding to which the clothing contributes but which is not directly conveyed by the clothing.
It may be there’s a component of why some people like associating with nicely dressed people that is about signaling in the more information-carrying sense (I dress in fancy clothes because I want people to know I’m popular so they’ll want to be friends with me, so I can make them do things to benefit me or add them to my crew of other cool-looking friends so that we can throw exclusive parties for hot people together). It just might be that the value of investing in a cool person wardrobe is either not worth the benefit for people who aren’t actually that popular, or that it’s harder to pull off that style of dress in practice because most people would come across as a poseur rather than an actual popular person.
But I don’t think that’s what the OP was really advocating, I think they were focused on the first part where dressing reasonably well makes you look OK to others, and people just enjoy (or at least don’t disenjoy) that raw fact about the way you look for its own sake.
Yeah, I don’t think the OP actually disambiguiated these two possibilities. They definitely have a ton of signaling stuff, but they also say:
“Because dressing nice makes your vibes better and people treat you better and are more willing to accommodate your requests. This is related to the way being attractive causes people to treat you better. While you can’t easily fix your face or age or weight (though you can try), you can easily change your clothes, and clothes go a long way towards making you attractive.”
And that to me is about the appeal of the clothes as beautification, rather than as an information-carrying (or trickery) device.
To me I see it the other way. People are treating you better because they are using “attractiveness” as a crude proxy for “what is my EV for interacting with that person”.
Hence why an AR device that let’s you know that ugly homeless man is a billionaire in disguise, and that good looking man in a suit has been convicted of running a pyramid scheme, would be examples of BETTER information than a crude proxy.
People are treating you better because they are using “attractiveness” as a crude proxy for “what is my EV for interacting with that person”.
I don’t really think that’s the whole story, I think that people are doing both. They associate with the well-dressed partly because they like the style, and partly because of other things they intuitively associate with the style. Insofar as they are happy about the fact that a guy wears a nice suit because they’re treating it as a crappy proxy for another piece of info they truly care about like his wealth, then yes by definition it would be nice to just know the thing rather than having to read the tea leaves.
But insofar as they are enjoying the suit because it’s pleasing to their eye, then that’s just kind of a raw fact of what they enjoy. There might be reasons that they have come to just intrinsically enjoy the look of nicely dressed people by the lights of their own culture, but it’s like we say about AI—we’re all trained to pursue/be rewarded by these proxies, and now it really is just about pursuing the proxies for their own sake because we’re just not that goal oriented.
It’s like house plants. What do they signal? That you can keep plants alive, that you aren’t so busy you can’t find time to water them (could be good or bad), that you have money to spend on plants, that you might be kind of a hippie, all sorts of things. But also, maybe I just happen to enjoy plant-filled environments, so if I’m friends with you, I get to hang out in your plant-filled house which is pleasant.
I think that second part is a big important part of why people choose to associate with people. For mysterious reasons they’ve been trained/endowed with instincts to feel pleasure when they behold certain attributes, and so they gravitate toward them without having a particular goal or caring much about the other forms of information the thing may convey.
Like if I had my lab mates over for a dinner party and my house was full of gorgeous art and plants (it is more like a monk’s cell in fact), I expect they’d just enjoy that a lot and find it more memorable, but wouldn’t necessarily start trying to form all kinds of new insights about my character and what I might be good for based on that. I think they’d just enjoy the plants and art and then go home with a nicer memory of the evening. Partly that’s because a lot of the information about me that they could pick up from those plants is information they also will be able to get from me in other ways. “Conscientious enough to keep plants alive” they can largely get from “has good work ethic in the lab,” “has enough money to spend on plants” they can get from “eats takeout lunch a couple days a week,” etc. The information signaling value of plants seems to me distinctly secondary to the visceral enjoyment aspect, and I think the same is true of clothes in many circumstances.
Signaling games aren’t exclusively for information, though—at least not if by “information” we mean “factual objective pieces of external data such as a bank account balance or employer.” They are also for manipulating people’s emotions, hopefully in positive ways, for creating opportunities to connect with other people in conversation, and for directing attention to the information that actually matters (as opposed to conveying raw information directly).
So for example, let’s say I’m a teacher giving corrective feedback about how to do a task during a lesson to a child. The information I want to convey is how to do the task.
But I need signaling to do things like:
Make the kid feel encouraged about their progress with the task rather than frustrated with their shortcomings so far
Creating small opportunities to get to know one another better as people, beyond the main goal of mastering the task
Refocusing the kid on what actually matters, such as shifting their attention away from assessing “is my teacher frustrated because I keep messing up?” (I very well might be, but trying to hide it and not let it affect my behavior), and toward assessing “is my teacher going to keep being encouraging and give constructive feedback even when I mess up?” (hopefully yes, if I’m doing my job right).
To me, signaling is about shifting where the emphasis is, or what’s top of mind, rather than conveying information. If we’re playing a board game, signaling is the part where you are manipulating each other into keeping it fun. It’s an active process, not a passive information transfer process.
In most cases the other player benefits from using a filter on the information you didn’t provide.
For example, seeing company financial reporting the way “warren buffet does” has a higher expected value than seeing them through the “emphasis” by company financial officers.
Similarly, if a small child could query their implants or external system to filter information by the highest EV to the child, that’s far better than what a human teacher offers.
Signaling is a conflict of interest and is not trustworthy information.
It’s unaligned to the interests of the observer. My hypothetical AR system that queries records on the target at least can be built to filter things to the observer’s benefit.
(though, note, information may not be statistically relevant. If hypothetically (assume the counter factual) it turned out in reality that people previously convicted of assault charges were less likely to commit further assaults (in a counter-factual world where rehabilitation worked), AR systems or implants might still emphasize this because it increases the perceived value of the AR system.
I don’t think we are being specific enough to declare what proportion of information hidden by signaling would be best revealed for the interests of the person receiving the signals.
We can certainly agree important cases exist where the signaling is bad for/unaligned with the person interpreting those signals. Or some subset of their interests that they maybe care about most heavily at that time.
No. I am claiming the incentives are misaligned such that the information presented this way will always be biased and is untrustworthy. Same way you can’t trust infomercials or pharmaceutical advertisments.
This is a correct and fairly stable generality.
Note the exceptions don’t matter. If statistically you know an information stream is heavily manipulated you can’t use it even in situations when it happens to be an honest signal. (because you don’t know the adversarial other agent is telling the truth This Time)
You seem to define signaling as a way of transmitting information other than the signaling mechanism itself. For example, if I have a stern face and a raised voice, those signaling mechanisms convey an inner emotional state of anger, and the possibility of punishment. You seem to be defining “signaling” as the conveyance of information about anger and punishment possibility, but as not including the stern face and raised voice.
I define signaling as often being the concrete good we care about, intrinsically for its own sake. For example, if my girlfriend smiles and laughs at my joke, I enjoy those pleasant expressions, not only because they convey her emotions or let me make predictions about the future, but also because they are a viscerally pleasant spectacle in their own right.
If signals can be a viscerally pleasant (or unpleasant) thing in their own right, then they aren’t zero sum. They are a product. We can manufacture more of them, or less, and get certain signal types to the people who are eager to consume them. We can enter into relationships where we find ourselves exchanging pleasant signals with each other.
Perhaps all we’re doing is implicitly arguing semantics (whether a ‘signal’ is the abstract information content, or the physical object that may or may not also convey additional information content)?
In my case I am arguing the issue with easily faked signals makes their information content suspect. I wouldn’t argue about your relationship with your girlfriend but this is why for example many high tier employers pay attention to only the signals that cant be faked. Which are the degree, name of school, name of prior employers and yoe, and score on interview questions. Everything else just gets ignored and I have gotten quarter million TC offers without a haircut and wearing some random Walmart polos. I assume the HFT half mil offers are no different.
Yeah I think we are discussing two entirely separate issues then. Here, I’m interested in people treating conventionally information-carrying signals as objects of interest in their own right. A smile can convey information (i.e. about emotional state), but we can also be interested in the smiles themselves, as when we take aesthetic pleasure in viewing a smiling face. Just a different topic.
Yeah I am just talking about how cheap information carrying objects are not really any more trustworthy a signal than say the annual report of a Bahama based cryptocurrency firm.
It’s not even relevant whether the signal is an object or a digital file. Just how likely it is to be honest, and the signal itself.
Oh yeah I (almost) completely agree with you there. Basically if it costs people less to game the information-carrying signal than the benefit they’ll get from gaming it, then it will get gamed.
It’s just that the original discussion was about clothes and how wearing nice clothes can give you a nicer life, and I think the reason is that you look better, so people want to associate with you more because they find you aesthetically pleasing, and I don’t think that’s bullshit or even really primarily the kind of information-carrying signaling we’re talking about in this subthread. I also think that clothing choices give people something to talk about, and people enjoy the pleasure of that conversation, and that visceral enjoyment is a hard-to-fake signal of group bonding to which the clothing contributes but which is not directly conveyed by the clothing.
It may be there’s a component of why some people like associating with nicely dressed people that is about signaling in the more information-carrying sense (I dress in fancy clothes because I want people to know I’m popular so they’ll want to be friends with me, so I can make them do things to benefit me or add them to my crew of other cool-looking friends so that we can throw exclusive parties for hot people together). It just might be that the value of investing in a cool person wardrobe is either not worth the benefit for people who aren’t actually that popular, or that it’s harder to pull off that style of dress in practice because most people would come across as a poseur rather than an actual popular person.
But I don’t think that’s what the OP was really advocating, I think they were focused on the first part where dressing reasonably well makes you look OK to others, and people just enjoy (or at least don’t disenjoy) that raw fact about the way you look for its own sake.
I felt the OP was saying “it’s bullshit primate signaling to wear nice clothes”.
BUT “being a popular primate is well worth the low cost of good looking threads”.
Yeah, I don’t think the OP actually disambiguiated these two possibilities. They definitely have a ton of signaling stuff, but they also say:
“Because dressing nice makes your vibes better and people treat you better and are more willing to accommodate your requests. This is related to the way being attractive causes people to treat you better. While you can’t easily fix your face or age or weight (though you can try), you can easily change your clothes, and clothes go a long way towards making you attractive.”
And that to me is about the appeal of the clothes as beautification, rather than as an information-carrying (or trickery) device.
To me I see it the other way. People are treating you better because they are using “attractiveness” as a crude proxy for “what is my EV for interacting with that person”.
Hence why an AR device that let’s you know that ugly homeless man is a billionaire in disguise, and that good looking man in a suit has been convicted of running a pyramid scheme, would be examples of BETTER information than a crude proxy.
I don’t really think that’s the whole story, I think that people are doing both. They associate with the well-dressed partly because they like the style, and partly because of other things they intuitively associate with the style. Insofar as they are happy about the fact that a guy wears a nice suit because they’re treating it as a crappy proxy for another piece of info they truly care about like his wealth, then yes by definition it would be nice to just know the thing rather than having to read the tea leaves.
But insofar as they are enjoying the suit because it’s pleasing to their eye, then that’s just kind of a raw fact of what they enjoy. There might be reasons that they have come to just intrinsically enjoy the look of nicely dressed people by the lights of their own culture, but it’s like we say about AI—we’re all trained to pursue/be rewarded by these proxies, and now it really is just about pursuing the proxies for their own sake because we’re just not that goal oriented.
It’s like house plants. What do they signal? That you can keep plants alive, that you aren’t so busy you can’t find time to water them (could be good or bad), that you have money to spend on plants, that you might be kind of a hippie, all sorts of things. But also, maybe I just happen to enjoy plant-filled environments, so if I’m friends with you, I get to hang out in your plant-filled house which is pleasant.
I think that second part is a big important part of why people choose to associate with people. For mysterious reasons they’ve been trained/endowed with instincts to feel pleasure when they behold certain attributes, and so they gravitate toward them without having a particular goal or caring much about the other forms of information the thing may convey.
Like if I had my lab mates over for a dinner party and my house was full of gorgeous art and plants (it is more like a monk’s cell in fact), I expect they’d just enjoy that a lot and find it more memorable, but wouldn’t necessarily start trying to form all kinds of new insights about my character and what I might be good for based on that. I think they’d just enjoy the plants and art and then go home with a nicer memory of the evening. Partly that’s because a lot of the information about me that they could pick up from those plants is information they also will be able to get from me in other ways. “Conscientious enough to keep plants alive” they can largely get from “has good work ethic in the lab,” “has enough money to spend on plants” they can get from “eats takeout lunch a couple days a week,” etc. The information signaling value of plants seems to me distinctly secondary to the visceral enjoyment aspect, and I think the same is true of clothes in many circumstances.