Nitpick, but I think music has at least as many dimensions as static visual art. We don't simply process music in real time but also "hear" the time signature and tempo. Music makes no sense to us without that information. Then, at any point in time you have various notes, timbres and amplitudes. We can display all of that information digitally in only 2-dimensions (which always blows my mind), but we hear the differences in the notes, the tone or timbre of the instruments and other acoustic or digital effects, changes in volume, the rhythm and the tempo.
Hi Hank! Thanks for the remark. I'd say it depends on how we want to look at it. It is certainly true that we normally use a higher-dimensional space to describe properties of music. When I say 1D, I am talking about the dimensionality of the raw input signal to the ear before processing to extract higher level properties (I should perhaps have said 2D, including the time values of the amplitude signals). It is true that once the brain processes it, it learns a whole bunch of different properties like timbre, tempo, rhythm etc. But when the brain hears music, it doesn't know time signature and tempo in advance - it does a computation to deduce it. These are exactly the kind of computations that I am referring to as "decoding" or "detecting of patterns". Thus, I want to know what the dimensionality of the input signal to the computation is.
We could do the same for visual arts. We could to a spatial Fourier transform and try to extract spatial periods (there will be a spectrum, rather than one) and get a list of x-BPMs and y-BPMs (beats per meter). Or extract various "visual" timbers. In this case we would get a number much higher than five. But I wanted to compare the dimensionalities of the raw input signals going into the computation.
Great read. This is probably basic, but it explains why we find lyrics and texts with rhymes satisfying, as our brains reward us for semi accurately predicting the sound that matches the pattern set up by the previous sounds, but it’s still novel (unless it’s the same word)
For whatever reason, I find the visual much more beautiful than the aural. I have a weird brain though.
Anyway, I’ve been wondering why humans like music. It makes little sense and all explanations I’ve gotten have fallen short. This explanation is the first that makes sense to me.
Been thinking the same thing! Have ended up in some lengthy arguments at dinner parties where I proposed that once you've fully 'learned' a piece of music–that is once you no longer notice anything new in it–it no longer gives you any enjoyment. A truth with modifications, I've later to come to think.
The predictive processing framework describes the fundamental computational building blocks of phenomenology, which necessarily encompasses beauty. It's an approach that doesn't let you distinguish beauty from 'liking', however, just like a scientific approach to color doesn't delineate where blue stops and green starts.
I agree with the premise, but find that a more potent way to think and talk about beauty is in a cultural sense: as an honorific term, the definition of which is always a matter of dispute and the function of which is to select which states to elevate and invest in. What should constitute beauty, and why, are questions people should be thinking way more about.
As for quantitative measures and dimensional descriptions of art forms, I think they miss some central points about conscious experience, namely that it is associative, multi-layered and multi-modal. The associative nature of experience means that input can allude to certain types of knowledge, memories and expectations, which mingle with the sensory pattern processing in producing specific sensations of beauty. By multilayered I mean that our conceptual understanding of the world is multiplicitous (think how you might appreciate a detail in a painting for its color, its shape, its contrast to the backgorund, its particular placement within the composition, its symbolical function etc.). The multi-modality of experience introduces the binding problem to aesthetics, and makes the phenomenological cascades of a sensory input even more complex.
In short, I think that a concept of beauty should be of a somewhat higher order than 'high rate of prediction error reduction', although it's a good anchor for discussion. I wrote way more than I intended to, hope I read you more or less correctly.
Thank you for taking the time to write up those thoughts, trbjrn! You definitely have some good points here. I do believe I can fit this into the framework however - at least I do not see immediate contradictions at this moment. Let me try to address point by point (in a slightly different order).
>>> As for quantitative measures and dimensional descriptions of art forms, I think they miss some central points about conscious experience, namely that it is associative, multi-layered and multi-modal. The associative nature of experience means that input can allude to certain types of knowledge, memories and expectations, which mingle with the sensory pattern processing in producing specific sensations of beauty. By multilayered I mean that our conceptual understanding of the world is multiplicitous (think how you might appreciate a detail in a painting for its color, its shape, its contrast to the backgorund, its particular placement within the composition, its symbolical function etc.). The multi-modality of experience introduces the binding problem to aesthetics, and makes the phenomenological cascades of a sensory input even more complex.
I definitely agree that conscious experience is associative, multi-layered and multi-modal. As far as I can understand at present, I think this is completely compatible, although it does of course introduce a lot of difficulty in empirically studying the issue. Let me focus on music for concreteness. Music contains a lot of information on both very low complexity and also very high complexity ideas and concepts. It communicates something as simple as a tempo, but perhaps also something as complex as statements on politics. During the subconscious processing of a musical signal, the brain is presumably looking for efficient ways to organize the information. Organization would be required to store information efficiently. Associations to other known concepts is probably a great way to organize and systematize the information, and so when I say ``pattern matching'' or ``organization'' of the information in the music, this higher order associative capability is almost certainly involved.
>>> Been thinking the same thing! Have ended up in some lengthy arguments at dinner parties where I proposed that once you've fully 'learned' a piece of music–that is once you no longer notice anything new in it–it no longer gives you any enjoyment. A truth with modifications, I've later to come to think.
Nice. I still think it is quite likely to be true if you actually could fully memorize every little detail of a piece of music. Thankfully, however, music contains a lot of microstructure with rich patterns. For example, subtleties in dynamics, timbre, tempo, etc. For this reason, a piece of music cannot ever truly be memorized (unless it is extremely banal), and so the beauty tends to be quite robust. There is microstructures to rediscover every time. This is a subject I am currently writing more about!
>>> The predictive processing framework describes the fundamental computational building blocks of phenomenology, which necessarily encompasses beauty. It's an approach that doesn't let you distinguish beauty from 'liking', however, just like a scientific approach to color doesn't delineate where blue stops and green starts.
I am not quite sure I agree on this point. At the very least, there seems to be certain categories of 'liking' the are very clearly distinct from the pleasure of beauty, under the assumption of my thesis. The 'liking' of food or the 'liking' of sex are pleasures triggered by very different mechanisms. In the first case the 'liking' is a response to the ingestion of certain special chemicals (fat, carbohydrates, and protein in high density), and in the second it is produced by certain very specific physical stimuli. Of course, in the latter case the aesthetic appreciation of your partners features might be mixed in, giving you a cocktail of likes.
>>> In short, I think that a concept of beauty should be of a somewhat higher order than 'high rate of prediction error reduction', although it's a good anchor for discussion. I wrote way more than I intended to, hope I read you more or less correctly.
I agree there are plausibly more pieces in play here! However, one thing I do find striking as a music composer/producer and as someone who spends quite a bit of time studying music composition, is that a very large fraction of all advice I come across in some way or another relates to aspects of manipulating predictability - sometimes explicitly and sometimes in slightly more cloaked ways. Another theme I am planning to write on more!
My overall point, which I failed to state clearly, is that while the input–that is the patterns/features of an aesthetic object–are constant, the processing is never quite the same, for the reasons I mentioned. You can fall in love with the same melody more than once, or feel indifferent to it at one time and enthusiastic at another, due to the context in which you hear it, your state of mind when you hear it etc. Microstructures I'm sure play a part in this, but I would think that the main driver of new experiences from 'old' input is associative. If association is also encompassed by pattern-matching, then I don't see why taste or orgasms wouldn't be.
Rather than understanding beauty as the reward administered when we decode effectively, I think it is better to conceive of it as a kind of conceptual category, a somewhat arbitrary set of states that we've some vague cultural consensus about (see Lisa Feldman Barrett on emotions). If the former description is to hold, it would have to include a wide range of experiences, potentially all pleasurable ones, since pattern processing seems to be a fundamental cognitive operation (whether this pattern is learned throughout your lifetime, like music, or encoded evolutionary as with an orgasm, is not necessarily relevant).
///
I think one of the most potent aspects of music is its invisibility, which makes it a good phenomenological 'container' for states, memories and associations. In general I tend to think that sensory novelty that falls within the perceptual proximal zone of development is interesting because phenomenological rewards can be coupled with ideas, values, behaviors, whatever; hence my emphasis on the value of cultural discussion about beauty. Perhaps more interesting than how the sensation of beauty arises is why it doesn't arise in so many contexts where we obviously have the perceptual ability required! Our capacity to feel beauty is highly regulated.
I agree that an orgasm isn't a case of entropic fine-tuning (the fine-tuning has already been carried out by evolution). I also contend, however, that my sense of beauty, when I'm touched by a great novel or when I'm attracted to a particular movement of a musical piece, is more than simply entropic fine-tuning, or that entropic fine-tuning becomes a too microscopic lens to talk meaningfully about it, just like saying "it's all just atoms" about a political event makes little sense. It's not a contradiction, but it's not all that enlightening either.
Or am I missing something? Again, I broadly agree, my difficulty lies in understanding what does and does not constitute pattern-decoding, or what kinds of pattern-decoding amount to beauty.
I agree with most of your first paragraph, and I am not saying that associations are not happening during food-based pleasure or orgasms. However, they do not seem very causally relevant to me. Also, with the latter two, there does not appear to be a comparably large amount of surprising (but structured) sensory information being processed and successfully organized/compressed. When we enjoy a burger or an orgasm, we to a large extent get precisely what we expect (where the expectation in the case of the sex was formed after intercourse started, but before climax).
The cultural aspect is where I disagree quite strongly. I certainly agree that culture strongly modulates what particular things we find beautiful, because culture shapes our brains; through experience it trains and biases us to be able to recognize and organize certain types of patterns over others. Through experience, it equips us with a repertoire of ideas and concepts we can then recognize as expressed in music and other art forms, basically providing us with a particular repertoire of culture-dependent templates we can use for pattern recognition, association and organization of information. This is especially potent when the artist shares our culture. However, stating that the pleasure of beauty itself is a primarily a cultural category so radically departs from my own conscious experience that it is very hard to even imagine what evidence would convince me; this is completely analogous to how you’d be very hard pressed to convince me that the joy of burgers and the joy of orgasms are fundamentally the same thing, but with boundary between them arbitrarily drawn by culture. It ought to be manifestly obvious that the latter two are distinct for non-cultural reasons, and I don’t see why beauty is any different. The analogy to food would be that culture affects what aromas, spices, and ingredients we prefer the most. But culture doesn’t determine the fact that we tend to feel pleasure from fats, carbohydrates and proteins, and that this pleasure is different from the pleasure of sex.
Let me also say that I certainly agree that our capacity to feel beauty is strongly regulated, and I cannot answer exactly when and why it gets regulated, beyond the obvious evolutionary (partial) arguments I already laid out above.
I’m trying to say two things: 1) pattern-matching/novelty-pleasure/entropic-fine-tuning, whatever you want to call it, is too broad a phenomenon to really pin down the concept of beauty. 2) If you were to insist on this however, it would be difficult to articulate why some kinds of pattern-matching along certain modalities qualify as beauty while others are simply pleasure. It’s a bit hard to talk about these two things separately, as I want to use 2 to argue for 1, and 1 as a jumping point for 2.
Perception is the organization of sensory information into experience, and ALL sensory information, to the extent that our brains can make sense of it, is patterned. It seems to me that you’re arguing that since some pattern preferences are pre-encoded by evolution, that is since they don't involve updating our expectations, that the pleasure associated with them do not qualify as beauty. Is getting a joke, then, the sensation of beauty? Is early childhood, when all we do is update our expectations, characterized by a constant sensation of beauty?
When I argue that beauty is a cultural category, I’m saying that our decision to label some pleasures as beautiful is a cultural ‘decision’–we decide that certain pleasures related to music, art, nature are experiences of ‘beauty’, while other experiences are pleasant or delicious. I think there are many good reasons for these groupings, but I don’t think they’re scientifically valid. As social animals we’re incredibly apt at imbuing things with status and sanctity, whether that be people, places or experiences.
I remember reading an article a few years ago about the neural bases of ‘liking’ and ‘wanting’, which, if I recall correctly, suggested that reward processing (liking) occurs in the same area of the brain for all rewards (‘wanting’ related to reward expectation, and was found to be independent). I’m not sure what the scientific consensus arouns that finding is (and strong neuroscientific claims are rarely appropriate), but it suggests that, in some sense, all liking is fundamentally the same.
If I understand you, you consider entropic fine-tuning to be ‘the’ marker of beauty; I think it is a necessary, but also quite insufficient, component rather than a definition of beauty, that it is a phenomenon that occurs in too many processes to be especially characteristic of it.
My apologies for the late response, life got in the way!
Indeed, clearly all sensory information is patterned. But most of it is not fine-tuned to lie in the band where we are operating on the edge of our sensory prediction capabilities (or where those capabilities are encountering a large gradient in time as we engage with the signals). I do not quite see that "it is a phenomenon that occurs in too many processes to be especially characteristic of it.". By definition, most sensory signals does not satisfy this property, and I have not been convinced that all signals associated with a pleasure satisfy this property. If I understand you correctly, you say that all pleasures are associated to sensory stimuli that are maximizing our capability for (gradients in) successful predictive processing, and I just don't see it :)
Consider for example the gradient in the number of bits/second that you accurately predict as you eat a burger. As you bite the burger, you perhaps accurately predict the presence of certain flavors, aromas, chemicals, yes. But its hard to see a gradient in time as you keep eating the burger, even though it keep tasting great. For the joke, as you get the punchline, there is perhaps some sense of beauty - there are certainly people making the case for comedy as art. But either way, the number of novel bits of information you successfully predicted during the joke, as compared to a piece of music or a painting, is probably very low. Of course, it is challenging to measure the information content in practice. I am not opposed to certain pleasures being a mixture though. Maybe eating a gourmet meal is part beauty-pleasure and part caloric-pleasure.
As for beauty in childhood - I certainly had frequent extreme experiences of beauty as I discovered my tastes in music during adolescence. I miss those days.
Finally, for the cultural aspects, I think I have raised my complaint in the last post and don't really see anything that addresses my above problems with this view. I could additionally add that we have neurological conditions like musical anhedonia, where some people are unable to feel pleasure from music, but they feel pleasure from things. Naively this seems to speak against the cultural view.
>>> If I understand you correctly, you say that all pleasures are associated to sensory stimuli that are maximizing our capability for (gradients in) successful predictive processing, and I just don't see it
I wouldn’t say that. I’m simply saying that all pleasure is reward for the brain receiving certain inputs (patterns). The range of patterns which may induce pleasure is highly variable according which sense/modality we’re speaking of course, but I don’t see why the principle of effective predictive processing should be relevant to (what you call) beauty and not orgasms or pleasant tastes.
If it boils down to the updating of models and gradients, I would at the very least submit culinary sensations as eligible for beauty-status: haven’t you ever tired of a flavor? Wasn’t it because you learned it?
...
Can you give a precise definition of ‘sensory prediction capability’? It is a fuzzy notion to me. Is it: the extent to which an agent is able to simulate the future iterations of a stimulus?
I can envision someone close to me dying, but I don’t know really know what it will feel like when I actually lose them. I wonder if this difference between envisioning (creating a low resolution model of) and predicting (high resolution simulation) confuses the discussion somewhat. Saying that you get ‘exactly what you expect’ when eating a hamburger or having an orgasm isn’t quite true in a predictive processing framework. Most of your hamburgers and orgasms are beyond your expectations, hopefully. When you can truly predict it, it hardly gives any pleasure–right?
Re: beauty in childhood
From what I gather you characterize beauty as a class of sensations related to learning to accurately predict stimuli. In early childhood you learn to decode slightly difficult patterns constantly, but I personally don't find experiences of beauty to be particularly evocative of childhood, not to the extent you'd expect anyway.
(I sense that we have different conceptions of culture, so it might not be productive to get into.)
I have often thought that the reason we like looking at water and clouds is because we're subconsciously trying to crack fluid dynamics
I'd say that's a great take!
Nitpick, but I think music has at least as many dimensions as static visual art. We don't simply process music in real time but also "hear" the time signature and tempo. Music makes no sense to us without that information. Then, at any point in time you have various notes, timbres and amplitudes. We can display all of that information digitally in only 2-dimensions (which always blows my mind), but we hear the differences in the notes, the tone or timbre of the instruments and other acoustic or digital effects, changes in volume, the rhythm and the tempo.
Hi Hank! Thanks for the remark. I'd say it depends on how we want to look at it. It is certainly true that we normally use a higher-dimensional space to describe properties of music. When I say 1D, I am talking about the dimensionality of the raw input signal to the ear before processing to extract higher level properties (I should perhaps have said 2D, including the time values of the amplitude signals). It is true that once the brain processes it, it learns a whole bunch of different properties like timbre, tempo, rhythm etc. But when the brain hears music, it doesn't know time signature and tempo in advance - it does a computation to deduce it. These are exactly the kind of computations that I am referring to as "decoding" or "detecting of patterns". Thus, I want to know what the dimensionality of the input signal to the computation is.
We could do the same for visual arts. We could to a spatial Fourier transform and try to extract spatial periods (there will be a spectrum, rather than one) and get a list of x-BPMs and y-BPMs (beats per meter). Or extract various "visual" timbers. In this case we would get a number much higher than five. But I wanted to compare the dimensionalities of the raw input signals going into the computation.
Great read. This is probably basic, but it explains why we find lyrics and texts with rhymes satisfying, as our brains reward us for semi accurately predicting the sound that matches the pattern set up by the previous sounds, but it’s still novel (unless it’s the same word)
Absolutely! This very much fits within the framework.
For whatever reason, I find the visual much more beautiful than the aural. I have a weird brain though.
Anyway, I’ve been wondering why humans like music. It makes little sense and all explanations I’ve gotten have fallen short. This explanation is the first that makes sense to me.
Been thinking the same thing! Have ended up in some lengthy arguments at dinner parties where I proposed that once you've fully 'learned' a piece of music–that is once you no longer notice anything new in it–it no longer gives you any enjoyment. A truth with modifications, I've later to come to think.
The predictive processing framework describes the fundamental computational building blocks of phenomenology, which necessarily encompasses beauty. It's an approach that doesn't let you distinguish beauty from 'liking', however, just like a scientific approach to color doesn't delineate where blue stops and green starts.
I agree with the premise, but find that a more potent way to think and talk about beauty is in a cultural sense: as an honorific term, the definition of which is always a matter of dispute and the function of which is to select which states to elevate and invest in. What should constitute beauty, and why, are questions people should be thinking way more about.
As for quantitative measures and dimensional descriptions of art forms, I think they miss some central points about conscious experience, namely that it is associative, multi-layered and multi-modal. The associative nature of experience means that input can allude to certain types of knowledge, memories and expectations, which mingle with the sensory pattern processing in producing specific sensations of beauty. By multilayered I mean that our conceptual understanding of the world is multiplicitous (think how you might appreciate a detail in a painting for its color, its shape, its contrast to the backgorund, its particular placement within the composition, its symbolical function etc.). The multi-modality of experience introduces the binding problem to aesthetics, and makes the phenomenological cascades of a sensory input even more complex.
In short, I think that a concept of beauty should be of a somewhat higher order than 'high rate of prediction error reduction', although it's a good anchor for discussion. I wrote way more than I intended to, hope I read you more or less correctly.
Thank you for taking the time to write up those thoughts, trbjrn! You definitely have some good points here. I do believe I can fit this into the framework however - at least I do not see immediate contradictions at this moment. Let me try to address point by point (in a slightly different order).
>>> As for quantitative measures and dimensional descriptions of art forms, I think they miss some central points about conscious experience, namely that it is associative, multi-layered and multi-modal. The associative nature of experience means that input can allude to certain types of knowledge, memories and expectations, which mingle with the sensory pattern processing in producing specific sensations of beauty. By multilayered I mean that our conceptual understanding of the world is multiplicitous (think how you might appreciate a detail in a painting for its color, its shape, its contrast to the backgorund, its particular placement within the composition, its symbolical function etc.). The multi-modality of experience introduces the binding problem to aesthetics, and makes the phenomenological cascades of a sensory input even more complex.
I definitely agree that conscious experience is associative, multi-layered and multi-modal. As far as I can understand at present, I think this is completely compatible, although it does of course introduce a lot of difficulty in empirically studying the issue. Let me focus on music for concreteness. Music contains a lot of information on both very low complexity and also very high complexity ideas and concepts. It communicates something as simple as a tempo, but perhaps also something as complex as statements on politics. During the subconscious processing of a musical signal, the brain is presumably looking for efficient ways to organize the information. Organization would be required to store information efficiently. Associations to other known concepts is probably a great way to organize and systematize the information, and so when I say ``pattern matching'' or ``organization'' of the information in the music, this higher order associative capability is almost certainly involved.
>>> Been thinking the same thing! Have ended up in some lengthy arguments at dinner parties where I proposed that once you've fully 'learned' a piece of music–that is once you no longer notice anything new in it–it no longer gives you any enjoyment. A truth with modifications, I've later to come to think.
Nice. I still think it is quite likely to be true if you actually could fully memorize every little detail of a piece of music. Thankfully, however, music contains a lot of microstructure with rich patterns. For example, subtleties in dynamics, timbre, tempo, etc. For this reason, a piece of music cannot ever truly be memorized (unless it is extremely banal), and so the beauty tends to be quite robust. There is microstructures to rediscover every time. This is a subject I am currently writing more about!
>>> The predictive processing framework describes the fundamental computational building blocks of phenomenology, which necessarily encompasses beauty. It's an approach that doesn't let you distinguish beauty from 'liking', however, just like a scientific approach to color doesn't delineate where blue stops and green starts.
I am not quite sure I agree on this point. At the very least, there seems to be certain categories of 'liking' the are very clearly distinct from the pleasure of beauty, under the assumption of my thesis. The 'liking' of food or the 'liking' of sex are pleasures triggered by very different mechanisms. In the first case the 'liking' is a response to the ingestion of certain special chemicals (fat, carbohydrates, and protein in high density), and in the second it is produced by certain very specific physical stimuli. Of course, in the latter case the aesthetic appreciation of your partners features might be mixed in, giving you a cocktail of likes.
>>> In short, I think that a concept of beauty should be of a somewhat higher order than 'high rate of prediction error reduction', although it's a good anchor for discussion. I wrote way more than I intended to, hope I read you more or less correctly.
I agree there are plausibly more pieces in play here! However, one thing I do find striking as a music composer/producer and as someone who spends quite a bit of time studying music composition, is that a very large fraction of all advice I come across in some way or another relates to aspects of manipulating predictability - sometimes explicitly and sometimes in slightly more cloaked ways. Another theme I am planning to write on more!
My overall point, which I failed to state clearly, is that while the input–that is the patterns/features of an aesthetic object–are constant, the processing is never quite the same, for the reasons I mentioned. You can fall in love with the same melody more than once, or feel indifferent to it at one time and enthusiastic at another, due to the context in which you hear it, your state of mind when you hear it etc. Microstructures I'm sure play a part in this, but I would think that the main driver of new experiences from 'old' input is associative. If association is also encompassed by pattern-matching, then I don't see why taste or orgasms wouldn't be.
Rather than understanding beauty as the reward administered when we decode effectively, I think it is better to conceive of it as a kind of conceptual category, a somewhat arbitrary set of states that we've some vague cultural consensus about (see Lisa Feldman Barrett on emotions). If the former description is to hold, it would have to include a wide range of experiences, potentially all pleasurable ones, since pattern processing seems to be a fundamental cognitive operation (whether this pattern is learned throughout your lifetime, like music, or encoded evolutionary as with an orgasm, is not necessarily relevant).
///
I think one of the most potent aspects of music is its invisibility, which makes it a good phenomenological 'container' for states, memories and associations. In general I tend to think that sensory novelty that falls within the perceptual proximal zone of development is interesting because phenomenological rewards can be coupled with ideas, values, behaviors, whatever; hence my emphasis on the value of cultural discussion about beauty. Perhaps more interesting than how the sensation of beauty arises is why it doesn't arise in so many contexts where we obviously have the perceptual ability required! Our capacity to feel beauty is highly regulated.
I agree that an orgasm isn't a case of entropic fine-tuning (the fine-tuning has already been carried out by evolution). I also contend, however, that my sense of beauty, when I'm touched by a great novel or when I'm attracted to a particular movement of a musical piece, is more than simply entropic fine-tuning, or that entropic fine-tuning becomes a too microscopic lens to talk meaningfully about it, just like saying "it's all just atoms" about a political event makes little sense. It's not a contradiction, but it's not all that enlightening either.
Or am I missing something? Again, I broadly agree, my difficulty lies in understanding what does and does not constitute pattern-decoding, or what kinds of pattern-decoding amount to beauty.
I agree with most of your first paragraph, and I am not saying that associations are not happening during food-based pleasure or orgasms. However, they do not seem very causally relevant to me. Also, with the latter two, there does not appear to be a comparably large amount of surprising (but structured) sensory information being processed and successfully organized/compressed. When we enjoy a burger or an orgasm, we to a large extent get precisely what we expect (where the expectation in the case of the sex was formed after intercourse started, but before climax).
The cultural aspect is where I disagree quite strongly. I certainly agree that culture strongly modulates what particular things we find beautiful, because culture shapes our brains; through experience it trains and biases us to be able to recognize and organize certain types of patterns over others. Through experience, it equips us with a repertoire of ideas and concepts we can then recognize as expressed in music and other art forms, basically providing us with a particular repertoire of culture-dependent templates we can use for pattern recognition, association and organization of information. This is especially potent when the artist shares our culture. However, stating that the pleasure of beauty itself is a primarily a cultural category so radically departs from my own conscious experience that it is very hard to even imagine what evidence would convince me; this is completely analogous to how you’d be very hard pressed to convince me that the joy of burgers and the joy of orgasms are fundamentally the same thing, but with boundary between them arbitrarily drawn by culture. It ought to be manifestly obvious that the latter two are distinct for non-cultural reasons, and I don’t see why beauty is any different. The analogy to food would be that culture affects what aromas, spices, and ingredients we prefer the most. But culture doesn’t determine the fact that we tend to feel pleasure from fats, carbohydrates and proteins, and that this pleasure is different from the pleasure of sex.
Let me also say that I certainly agree that our capacity to feel beauty is strongly regulated, and I cannot answer exactly when and why it gets regulated, beyond the obvious evolutionary (partial) arguments I already laid out above.
I’m trying to say two things: 1) pattern-matching/novelty-pleasure/entropic-fine-tuning, whatever you want to call it, is too broad a phenomenon to really pin down the concept of beauty. 2) If you were to insist on this however, it would be difficult to articulate why some kinds of pattern-matching along certain modalities qualify as beauty while others are simply pleasure. It’s a bit hard to talk about these two things separately, as I want to use 2 to argue for 1, and 1 as a jumping point for 2.
Perception is the organization of sensory information into experience, and ALL sensory information, to the extent that our brains can make sense of it, is patterned. It seems to me that you’re arguing that since some pattern preferences are pre-encoded by evolution, that is since they don't involve updating our expectations, that the pleasure associated with them do not qualify as beauty. Is getting a joke, then, the sensation of beauty? Is early childhood, when all we do is update our expectations, characterized by a constant sensation of beauty?
When I argue that beauty is a cultural category, I’m saying that our decision to label some pleasures as beautiful is a cultural ‘decision’–we decide that certain pleasures related to music, art, nature are experiences of ‘beauty’, while other experiences are pleasant or delicious. I think there are many good reasons for these groupings, but I don’t think they’re scientifically valid. As social animals we’re incredibly apt at imbuing things with status and sanctity, whether that be people, places or experiences.
I remember reading an article a few years ago about the neural bases of ‘liking’ and ‘wanting’, which, if I recall correctly, suggested that reward processing (liking) occurs in the same area of the brain for all rewards (‘wanting’ related to reward expectation, and was found to be independent). I’m not sure what the scientific consensus arouns that finding is (and strong neuroscientific claims are rarely appropriate), but it suggests that, in some sense, all liking is fundamentally the same.
If I understand you, you consider entropic fine-tuning to be ‘the’ marker of beauty; I think it is a necessary, but also quite insufficient, component rather than a definition of beauty, that it is a phenomenon that occurs in too many processes to be especially characteristic of it.
My apologies for the late response, life got in the way!
Indeed, clearly all sensory information is patterned. But most of it is not fine-tuned to lie in the band where we are operating on the edge of our sensory prediction capabilities (or where those capabilities are encountering a large gradient in time as we engage with the signals). I do not quite see that "it is a phenomenon that occurs in too many processes to be especially characteristic of it.". By definition, most sensory signals does not satisfy this property, and I have not been convinced that all signals associated with a pleasure satisfy this property. If I understand you correctly, you say that all pleasures are associated to sensory stimuli that are maximizing our capability for (gradients in) successful predictive processing, and I just don't see it :)
Consider for example the gradient in the number of bits/second that you accurately predict as you eat a burger. As you bite the burger, you perhaps accurately predict the presence of certain flavors, aromas, chemicals, yes. But its hard to see a gradient in time as you keep eating the burger, even though it keep tasting great. For the joke, as you get the punchline, there is perhaps some sense of beauty - there are certainly people making the case for comedy as art. But either way, the number of novel bits of information you successfully predicted during the joke, as compared to a piece of music or a painting, is probably very low. Of course, it is challenging to measure the information content in practice. I am not opposed to certain pleasures being a mixture though. Maybe eating a gourmet meal is part beauty-pleasure and part caloric-pleasure.
As for beauty in childhood - I certainly had frequent extreme experiences of beauty as I discovered my tastes in music during adolescence. I miss those days.
Finally, for the cultural aspects, I think I have raised my complaint in the last post and don't really see anything that addresses my above problems with this view. I could additionally add that we have neurological conditions like musical anhedonia, where some people are unable to feel pleasure from music, but they feel pleasure from things. Naively this seems to speak against the cultural view.
Thanks again for the comments!
>>> If I understand you correctly, you say that all pleasures are associated to sensory stimuli that are maximizing our capability for (gradients in) successful predictive processing, and I just don't see it
I wouldn’t say that. I’m simply saying that all pleasure is reward for the brain receiving certain inputs (patterns). The range of patterns which may induce pleasure is highly variable according which sense/modality we’re speaking of course, but I don’t see why the principle of effective predictive processing should be relevant to (what you call) beauty and not orgasms or pleasant tastes.
If it boils down to the updating of models and gradients, I would at the very least submit culinary sensations as eligible for beauty-status: haven’t you ever tired of a flavor? Wasn’t it because you learned it?
...
Can you give a precise definition of ‘sensory prediction capability’? It is a fuzzy notion to me. Is it: the extent to which an agent is able to simulate the future iterations of a stimulus?
I can envision someone close to me dying, but I don’t know really know what it will feel like when I actually lose them. I wonder if this difference between envisioning (creating a low resolution model of) and predicting (high resolution simulation) confuses the discussion somewhat. Saying that you get ‘exactly what you expect’ when eating a hamburger or having an orgasm isn’t quite true in a predictive processing framework. Most of your hamburgers and orgasms are beyond your expectations, hopefully. When you can truly predict it, it hardly gives any pleasure–right?
Re: beauty in childhood
From what I gather you characterize beauty as a class of sensations related to learning to accurately predict stimuli. In early childhood you learn to decode slightly difficult patterns constantly, but I personally don't find experiences of beauty to be particularly evocative of childhood, not to the extent you'd expect anyway.
(I sense that we have different conceptions of culture, so it might not be productive to get into.)