Is there a "hard problem of aesthetics?"

Question

As a society, we already spend a lot of time and energy predicting what somebody will find enjoyable; programmers and psychologists have spent decades honing recommendation algorithms to predict what somebody will enjoy watching.

What if we imagine a "perfected" version of this: an algorithm that, given enough details about me, can predict how beautiful, funny, sublime, disgusting, etc I will find any particular work of art to be. (Say, it can accurately predict how I'd rate it on a scale of 1 to 10, or for any two works of art, which one I'd choose as "more beautiful" or "more funny" etc.)

Does this algorithm "solve" aesthetics, or is is there an explanatory gap/something else that remains unresolved about the nature of beauty/taste/aesthetics? In particular, have any contemporary (or historical) philosophers written about what remains if this part of aesthetics is "solved"?

Answer 1

Short Answer

Yes. There is far more to aesthetics than a preference algorithm. Developing a good preference prediction algorithm is a step toward understanding Aesthetics. But there is a lot more to understand about WHY we have preferences, and what they do to us.

Aside on Consciousness

The experience of aesthetics is a key feature of consciousness, but this is not a new problem. It would just be a subset of the hard problem of consciousness.

Answer 2

One of the most insightful books about this is Fluid Concepts and Creative Analogies (1995) in which Douglas Hofstadter and collaborators investigate core elements of this question, approaching it both from the side of our aesthetic perception (which includes an appreciation for surprising recursion) and from the side of how to create programs that mirror those perceptions.

Another one is Analogy-making as Perception, by Melanie Mitchell (originally a student of Hofstadter), which explains in more detail how the CopyCat project worked (a project that is also described in the first book).

The computer programs described in those books are very very different from recommendation systems. They are also not geared towards the same tasks, but towards certain "toy domains", such as the domain of letter analogies. A typical CopyCat problem is "if abc goes to abd, what does ijk go to?". (Ok, that was easy. But how about "if abc => abd, then iijjkk => ?" Now things start to get more interesting, but the answer most humans come up with is still pretty predictably iijjll. Ok, then what about abc => abd, then xyz => ?). It turns out that even in such a toy domain lots of our aesthetic sense comes into play.

Do those programs shed light on how our aesthetic sense works? Do they explain why we might find one solution "ugly and dumb" (even though it's defendable and possible) and another solution "elegant and perhaps surprising"? A little bit, since in these programs we sometimes can tweak some parameters and have the program output either the "ugly" or the "elegant" solution. In this sense they do hint at some of the mechanisms that could also underly our aesthetic sense which is deeply entangled with any kind of perception and with seeing analogies. In a more general sense, they also give a primitive sketch of how something like consciousness perception (meaning) could arise.

As an example of a smart solution to abc => abd then xyz => ? you could give xyy, but a more elegant one might be xya. Note that what is 'elegant' or 'smart' would not be builtin, it's an emergent aspect (something that appears to be so in our eyes), and it can only remain a somewhat volatile or 'subjective' aspect. (Note also, that the program does not have a concept of a circular alphabet where 'a' follows 'z', so if it comes up with xya as solution that is surely 'surprising', an unforeseen, 'creative' leap.)

The kind of programs developed by Hofstadter and his group are very different from current large recommendation systems (or from LLMs). Those large systems are based on ridiculously simple algorithms: It's not the algorithms that encode the preferences or dispositions that are being predicted, but the models (the matrices of parameters). And those models are basically black boxes. We don't understand those models (and are unable to ever understand them as such) because they are high-dimensional vector spaces. They do give us concrete objects to study however. So, by studying those black boxes we may be able to also come to a better understanding of aesthetic perceptions. (But the main caveat, I think, is that current architectures are very different from biological systems, so the way the model perceices something may be rather different from the way humans do so. Some of the general principles may still be the same, of course.)

Answer 3

Well, if we are drawing a connection between aesthetics and consciousness, it could be that the "problem of aesthetics", whatever that may be, directly informs the hard problem of consciousness. How so? Remember, the hard problem of consciousness is asking a fundamentally related question because aesthetics simply studies preferences manifested during consciousness, particularly of art and beauty and is therefore axiological in nature. From WP:

In the philosophy of mind, the hard problem of consciousness is to explain why and how humans and other organisms have qualia, phenomenal consciousness, or subjective experience. (emphasis mine) It is contrasted with the "easy problems" of explaining why and how physical systems give a (healthy) human being the ability to discriminate, to integrate information, and to perform behavioral functions such as watching, listening, speaking (including generating an utterance that appears to refer to personal behaviour or belief), and so forth. The easy problems are amenable to functional explanation—that is, explanations that are mechanistic or behavioral—since each physical system can be explained (at least in principle) purely by reference to the "structure and dynamics" that underpin the phenomenon.

Taking a page from cognitive science, where consciousness is often understood as a control system for the organism, aesthetics can be understood through a similar lens, as put forward by Aesthetics and Cognitive Science (SEP). Whatever evolution has moved consciousness forward for, it certainly related to both intentionality, which is the study of how the mind seems to manifest aboutness and produce representations, and providing preferences for those representations. For instance, physical beauty is certainly interrelated with sexual reproduction and rearing offspring. Also, certain type of gardening and landscaping styles may be rooted in our evolutionary ancestry and preferences in the natural environment.

Lastly, our emotions and their phenomenological manifestations also seem tied to questions around aesthetics. Emotions and feelings are deep and powerful conscious experiences, and far more influential than mathematical logic and reason to our day-to-day lives. Certainly, aesthetics plays a role in resolving questions about why our conscious experience is as it is. From the SEP article:

A good deal of aesthetic thinking has been taken up with arguing about the limits to its capacity to appeal to the emotions. Two issues are particularly notable, one descriptive and the other, in part, normative. What, first of all, are the facts about our (apparently) emotional responses to fictions? Do we really admire, despise or pity people we know do not exist? This sounds like an empirical question, perhaps one to be resolved by studies of the brains of people engaged by fictional work.

You ask:

Is there a "hard problem of aesthetics?"

I would say yes, but it would seem to be an extension of the hard problem of consciousness. Insofar as we must have an accounting for the origins of the experience of consciousness, that is, reconciling our subjective, first-person experiences with the third-person quality of the external world around us, aesthetics seems to answer deep and fundamental questions about qualia and human will and action. Perhaps it is through this line of questioning where the first- and third-person are ultimately blurred to provide a convincing argument for the hard problem of consciousness (presuming you believe the problem exists to begin with).

Answer 4

Note: answering your question as phrased specifically:

"an algorithm that, given enough details about me, can predict how beautiful, funny, sublime, disgusting, etc I will find any particular work of art ..."

Note that that would be a device (like a speedometer) which, given a work of art will (amazingly) know your 1-10 rating, and it will be correct.

However, this has no connection at all to knowing how to create an art work that results in a given rating; which is what you are getting at with "aesthetics solved".

Just as speedometers are wonderful devices (for knowing the speed of a vehicle) but are of literally no help in "making vehicles fast".