I recently finished Kazuo Ishiguro’s 2021 novel Klara and the Sun. It was pretty good overall: the future developments don’t seem that realistic1, but I liked how most of the characters and themes were handled. It’s strange to have a robot who can tutor physics but think the sun sets in a barn, and while humans have similarly conflicting rational vs. religious intuitions, I wasn’t really satisfied with how the book dealt with this.
The above are my own thoughts, but after thinking about it for a bit on my own I couldn’t help reading some online discussion of the book. It was interesting to hear the interpretations people had, some of which were neat clarifications of ideas I had sort of picked up on and others which I didn’t think of or didn’t agree with but which were still fun to consider. I also read an interview with Ishiguro where he talks about the book, which shed some light on his thinking.
I generally enjoy seeing what other people (including the authors) have to say about a given book/movie/TV show/other piece of media I know about, but I also have a strong impulse not to and instead come up with my own thoughts about a given work. Just like writing, thinking about why you like certain things and don’t like other things can help develop and articulate your tastes. Even if some of your interpretations are wrong, they could be wrong in interesting ways that reveal interesting aspects of both the work you are reflecting on as well as you yourself.
After doing this, I realized that a lot of differences in personal taste are reducible to factors that are distinct from the qualities of the works themselves. There are some aspects of taste that most people hold in common, which is what many reviews focus on: having a coherent and compelling plot, well-rounded and dynamic characters, good storytelling, good prose/effects, and more. An essay from someone who likes a book might explain that the characters are three-dimensional and therefore it is good, while someone who disliked the book might emphasize the plot holes. But already you can see how just having (or not having) the “required” elements of a given medium isn’t sufficient on its own to like (or dislike) an example in that medium.
Another interesting axis is novelty vs familiarity. This is related to the explore-exploit dilemma in media, which I have written about before and which Gwern covers in more depth. Generally we tend to like things we are more familiar with, which is a known bias that plays into things like brand recognition. But working in the opposite direction is an element of novelty seeking: Iron Man (2008) was genuinely groundbreaking but the 20th Marvel movie in the last two decades gets old fast. So not only do the qualities of a work matter, but so does its context.
Finally, even if something does all of the above just right, you still might just not like it. There might be a free jazz or death metal song that connoisseurs of the genre agree is amazing, but I just don’t like those genres of music. Music is a great example of the problem of taste, since the qualities of “good” vs “bad” songs are even harder to pin down. Even if you like a given genre or broader style, it is hard to find someone with the exact same tastes as you.
This poses a fundamental limit to any description of the objective qualities of a work: you can describe a work in any arbitrary detail, but to then say that some set of those qualities makes a work “good” is a normative claim that doesn’t universally hold. There can be any combination of tropes, their subversion, or their absence, but whether you enjoy the resulting product is more a function of your own tastes than those tropes. You can argue that people should like some features more than others, but this is again a normative claim rather than an objective one.
Just like moral claims don’t automatically fall out of amoral language, claims about the quality of a work don’t automatically fall out of descriptors of the work itself. Not only do there have to be some links between descriptors and quality (which you can disagree with), but these links don’t always hold. If they did, you could just apply a formula to generate a work with arbitrary levels of quality. Since there is intuitively no (as-yet-discovered) way to automatically produce media with a given level of quality, this suggests that people have a level of reflexive novelty-seeking that precludes not just a quantifiable mapping from description to quality but a prediction of how that mapping will change with time. Even if you could approximate this with some accuracy, it would probably chaotically evolve like weather systems, doing so even more quickly if people become aware of this system.
Given this complexity, it is all the more important to think about what works you like and why you like them. Not only can you seek out more of what you like, but you can find others who share your tastes, and even create your own works as that process becomes easier.
Footnotes
-
Spoilers: Regarding technology, I don’t think post-fertilization genetic enhancement will be as effective as pre-fertilization selection/enhancement or brain-computer interfaces (of which keyboards/mice/touchscreens are preliminary versions), and AI will develop differently, but for being published pre-ChatGPT it does well. ↩