No, it is not okay (in the sense of being non-detrimental to one’s terminal values) to like popular things simply on the basis that they are popular. Decision theories following this heuristic are vulnerable to numerous low-complexity attack vectors, leading them to (for example) perpetuate, generate, and incorrectly update on information cascades.
It would be more accurate to say that “Giving in to social pressure to have aesthetic preference on the basis of the popularity of such preferences has non-obvious and immodular benefits to one’s terminal values, which are likely to outweigh their decision-theoretic vulnerabilities within human cultures.”
I’m sure that you’re absolutely technically correct when saying what you’d said, but I had to reread it 5 times just to figure out what you meant, and I’m still not sure.
Are you saying that the strategy to indiscriminately like whatever’s popular will lead to worse outcomes because of random effects, as in this experiment that showed that popularity is largely random? Then you’re right—because what are the chances that your preferences exactly match the popular choice?
On the other hand, if it so happens that you end up liking something that’s popular and you couldn’t tell it apart from something similar in a blind test, is it in any way bad that you’re getting utility out of it?
Yes, that’s correct. Generally, permitting your valuations or beliefs to be influenced hysteretically in the direction of what’s popular is bad. However, because aesthetic judgments have minor epistemic harm, and have instrumental value due to better “bonding” with other humans, that general heuristic does not apply here, so you can safely permit yourself to be drawn toward the judgment of others with respect to music.
Perhaps, but I think the bond strength enhancement would be far greater if you would strongly update towards liking the aesthetic of the paperclip shape for metals.
Well, there are things this works for and things it doesn’t.
For instance, you generally shouldn’t decide whether to say God exists or not, or how to vote (beyond strategically), based on popularity. If enough people do that then we act collectively non-rationally and may end up in locally optimal but absolutely bad equilibria.
However, you should decide what side of the street to drive on, or what language to speak in your everyday life, based on popularity. These are simple coordination games.
What is music taste—or whatever else we have under discussion—more like? That’s an empirical question. The fact that we seem to like popularity in and of itself may suggest elements of a coordination game—and certainly the ability to discuss music with friends and so on displays that quality—but there’s also a status treadmill as well, as with anything else that displays “fashions.” The latter effect means that by switching to Radiohead you’re decreasing the status enjoyment others gain from it, as you’re diluting its coolness-signaling.
Musical taste is so low-stakes that you should probably just listen to whatever you enjoy without worrying too much about why you enjoy it, unless you enjoy doing that sort of analysis—which obviously anyone writing a post about it does, but even if it does, you shouldn’t feel the need to switch to a less enjoyable set of tunes.
Your views coincide heavily with mine on this topic.
I would only object that e.g. driving on the same side of the street as everyone else [1] because everyone else drives on that side is not a case of engaging in a practise “simply on the basis that others” engage in the practise, but to avoid catastrophic failure. If the street system’s usage factor were on the order of parts per billion, and agents had a preference for not being confined to one side, it would be “okay” [2] to drive as one pleases and simply adjust in the rare case of seeing another user.
[1] assuming I correctly understand this to be a reference to two-way conveyance-transmission systems in which changing chirality is impractical on short notice, and opposite chirality for two coincident conveyances is catastrophic
[2] i.e., non-detrimental to their terminal values
I’m not suggesting that we believe things because others do—that’s enormously different! My reasoning is as follows: Pleasure is among my terminal values; listening to this music gives me pleasure; so in the absence of important other factors in the expected utility calculation, it’s A Good Thing. In this particular case, are there, in fact, other important important consequences?
True, and it is the arbitrary, non-epistemic nature of aesthetics that justifies my judgment that you do little harm by permitting the influence you describe.
But what’s actually happening is that the music initially doesn’t give you pleasure—as you said, you would not be able to distinguish it from another “Radiohead” imitator that you now “don’t like”—and then you have consciously permitted (in the sense of not resisting a known phenomenon) yourself to make a path-dependent update in your tastes through a (so-called) “bandwagon” effect.
Really, you would be better off simply listening to music that you independently like, rather than spending finite resources to change your aesthetic preferences in a different arbitrary direction, and that cost is the consequence. My point, then, is that I agree with you in that there are instrumental reasons to permit such hysteresis in your aesthetic preferences, such as the improved ability to relate to others—and this likely outweighs the above harm.
In the more general case, though, it would not help you to permit such a “bandwagon” adjustment in the direction of, for example, “liking” to listen to materials science lectures filled with incorrect models. Though you may think that you can “draw the distinction”, and avoid absorbing the incorrect materials science into your beliefs, the reality is that ape brains are very poor about handling this kind of dissonance.
I hope my position and concerns are clearer and more intuitive now.
I think we agree on the “broad brush” aspects of the problem (in particular that it’s really about the specifics here).
Thank you for the example of fashionable materials science lectures, I now agree on that point.
The specifics here are interesting, though, so I’ll take it further.
Discounting values for path-dependency has an appealing elegance, but as a general principle it is, I think, unsound. Any value that saturates has such a path-dependence: I enjoy eating, then I become full, then I no longer enjoy eating. I enjoy learning about linear algebra, then I master it, then I no longer enjoy learning about linear algebra. If you appeal to the underlying value of “not being hungry” or “learning new mathematics” then I can equally appeal to “enjoying fashionable music”.
That ignores the specific social component of bandwagons, though, which I have already said that I agree with you on.
Really, you would be better off simply listening to music that you independently like, rather than spending finite resources to change your aesthetic preferences in a different arbitrary direction, and that cost is the consequence.
This assumes that the cost of finding music that appeals to one’s innate preferences is lower than the cost of changing those preferences to match what’s commonly available. This is not necessarily the case—in fact, I generally find it not to be. (Pandora music service’s AI comes close, but it can still only play music that’s actually been produced and entered into its database, and even in that case it takes a fair amount of effort to teach the AI what I like in each general category of music that I like, plus I suspect it’s incapable of learning certain nuances like ‘I only like this instrument when it’s paired with one of these other instruments’.)
Although there may be people who never heard a piece of music that they actually liked, while still having the ability to like some music, I thought such cases are so rare that they can be disregarded. I generally dislike most of the modern music (where “modern” means something like “less than 50 years old”), and have no problem to find something which fits my preferences. In purely financial costs, non-fashionable music is probably much cheaper.
Setting up a status-effect-free “clean room” from which to evaluate new music would be far more costly than just letting the mostly-autonomous bandwagon effect produce the same amount of aesthetic satisfaction.
That may hold for a certain class of human situations. However, User:alexflint’s story indicates that User:alexflint did endure a gross cost [1] greater than what would have resulted from not applying filters for popularity. The alternative to allowing one to be influenced in this matter is not to impose an impermeable “clean room”, but to simply avoid avoidable expenditures when the source of the bias is noticed.
[1] though as I’ve made clear, not necessarily exceeding the benefits to terminal values
I think alexflint’s point is something along the lines of “it’s okay to like popular things just because they’re popular.”
No, it is not okay (in the sense of being non-detrimental to one’s terminal values) to like popular things simply on the basis that they are popular. Decision theories following this heuristic are vulnerable to numerous low-complexity attack vectors, leading them to (for example) perpetuate, generate, and incorrectly update on information cascades.
It would be more accurate to say that “Giving in to social pressure to have aesthetic preference on the basis of the popularity of such preferences has non-obvious and immodular benefits to one’s terminal values, which are likely to outweigh their decision-theoretic vulnerabilities within human cultures.”
I’m sure that you’re absolutely technically correct when saying what you’d said, but I had to reread it 5 times just to figure out what you meant, and I’m still not sure.
Are you saying that the strategy to indiscriminately like whatever’s popular will lead to worse outcomes because of random effects, as in this experiment that showed that popularity is largely random? Then you’re right—because what are the chances that your preferences exactly match the popular choice?
On the other hand, if it so happens that you end up liking something that’s popular and you couldn’t tell it apart from something similar in a blind test, is it in any way bad that you’re getting utility out of it?
Yes, that’s correct. Generally, permitting your valuations or beliefs to be influenced hysteretically in the direction of what’s popular is bad. However, because aesthetic judgments have minor epistemic harm, and have instrumental value due to better “bonding” with other humans, that general heuristic does not apply here, so you can safely permit yourself to be drawn toward the judgment of others with respect to music.
I shouldn’t, but it’s safe for humans.
Clippy, I think we would bond better if you would strongly update towards liking the aesthetic of Radiohead.
Perhaps, but I think the bond strength enhancement would be far greater if you would strongly update towards liking the aesthetic of the paperclip shape for metals.
This is almost undoubtedly true; shared beliefs forge stronger bonds when they are not widely held.
Well, there are things this works for and things it doesn’t.
For instance, you generally shouldn’t decide whether to say God exists or not, or how to vote (beyond strategically), based on popularity. If enough people do that then we act collectively non-rationally and may end up in locally optimal but absolutely bad equilibria.
However, you should decide what side of the street to drive on, or what language to speak in your everyday life, based on popularity. These are simple coordination games.
What is music taste—or whatever else we have under discussion—more like? That’s an empirical question. The fact that we seem to like popularity in and of itself may suggest elements of a coordination game—and certainly the ability to discuss music with friends and so on displays that quality—but there’s also a status treadmill as well, as with anything else that displays “fashions.” The latter effect means that by switching to Radiohead you’re decreasing the status enjoyment others gain from it, as you’re diluting its coolness-signaling.
Musical taste is so low-stakes that you should probably just listen to whatever you enjoy without worrying too much about why you enjoy it, unless you enjoy doing that sort of analysis—which obviously anyone writing a post about it does, but even if it does, you shouldn’t feel the need to switch to a less enjoyable set of tunes.
Your views coincide heavily with mine on this topic.
I would only object that e.g. driving on the same side of the street as everyone else [1] because everyone else drives on that side is not a case of engaging in a practise “simply on the basis that others” engage in the practise, but to avoid catastrophic failure. If the street system’s usage factor were on the order of parts per billion, and agents had a preference for not being confined to one side, it would be “okay” [2] to drive as one pleases and simply adjust in the rare case of seeing another user.
[1] assuming I correctly understand this to be a reference to two-way conveyance-transmission systems in which changing chirality is impractical on short notice, and opposite chirality for two coincident conveyances is catastrophic
[2] i.e., non-detrimental to their terminal values
I’m not suggesting that we believe things because others do—that’s enormously different! My reasoning is as follows: Pleasure is among my terminal values; listening to this music gives me pleasure; so in the absence of important other factors in the expected utility calculation, it’s A Good Thing. In this particular case, are there, in fact, other important important consequences?
True, and it is the arbitrary, non-epistemic nature of aesthetics that justifies my judgment that you do little harm by permitting the influence you describe.
But what’s actually happening is that the music initially doesn’t give you pleasure—as you said, you would not be able to distinguish it from another “Radiohead” imitator that you now “don’t like”—and then you have consciously permitted (in the sense of not resisting a known phenomenon) yourself to make a path-dependent update in your tastes through a (so-called) “bandwagon” effect.
Really, you would be better off simply listening to music that you independently like, rather than spending finite resources to change your aesthetic preferences in a different arbitrary direction, and that cost is the consequence. My point, then, is that I agree with you in that there are instrumental reasons to permit such hysteresis in your aesthetic preferences, such as the improved ability to relate to others—and this likely outweighs the above harm.
In the more general case, though, it would not help you to permit such a “bandwagon” adjustment in the direction of, for example, “liking” to listen to materials science lectures filled with incorrect models. Though you may think that you can “draw the distinction”, and avoid absorbing the incorrect materials science into your beliefs, the reality is that ape brains are very poor about handling this kind of dissonance.
I hope my position and concerns are clearer and more intuitive now.
I think we agree on the “broad brush” aspects of the problem (in particular that it’s really about the specifics here).
Thank you for the example of fashionable materials science lectures, I now agree on that point.
The specifics here are interesting, though, so I’ll take it further.
Discounting values for path-dependency has an appealing elegance, but as a general principle it is, I think, unsound. Any value that saturates has such a path-dependence: I enjoy eating, then I become full, then I no longer enjoy eating. I enjoy learning about linear algebra, then I master it, then I no longer enjoy learning about linear algebra. If you appeal to the underlying value of “not being hungry” or “learning new mathematics” then I can equally appeal to “enjoying fashionable music”.
That ignores the specific social component of bandwagons, though, which I have already said that I agree with you on.
This assumes that the cost of finding music that appeals to one’s innate preferences is lower than the cost of changing those preferences to match what’s commonly available. This is not necessarily the case—in fact, I generally find it not to be. (Pandora music service’s AI comes close, but it can still only play music that’s actually been produced and entered into its database, and even in that case it takes a fair amount of effort to teach the AI what I like in each general category of music that I like, plus I suspect it’s incapable of learning certain nuances like ‘I only like this instrument when it’s paired with one of these other instruments’.)
Although there may be people who never heard a piece of music that they actually liked, while still having the ability to like some music, I thought such cases are so rare that they can be disregarded. I generally dislike most of the modern music (where “modern” means something like “less than 50 years old”), and have no problem to find something which fits my preferences. In purely financial costs, non-fashionable music is probably much cheaper.
Setting up a status-effect-free “clean room” from which to evaluate new music would be far more costly than just letting the mostly-autonomous bandwagon effect produce the same amount of aesthetic satisfaction.
That may hold for a certain class of human situations. However, User:alexflint’s story indicates that User:alexflint did endure a gross cost [1] greater than what would have resulted from not applying filters for popularity. The alternative to allowing one to be influenced in this matter is not to impose an impermeable “clean room”, but to simply avoid avoidable expenditures when the source of the bias is noticed.
[1] though as I’ve made clear, not necessarily exceeding the benefits to terminal values