I’ve written about configuration spaces before. I plan to use the notion in some upcoming posts, so this seems like a good time for a refresher. (If you’re new to the idea, I recommend that you read at least the first post in the series. The third one might be a helpful read, too.)
Today I’ll talk about a configuration space where the axes consist of personal taste and objective quality. Which obviously implies there is such a thing as objective quality. I think there is, and I’ll try to make a case for it. (Certainly production quality offers objective metrics.)
Of course, as everyone knows, there is no accounting of personal taste.
That being the case, we’ll accept the simple fact that people have personal taste — they can either like something or dislike something, which gives us convenient respective positive and negative values. Likes or dislikes can be weak or strong, so those values (numbers) can be small or large.
People can also be indifferent to something, which, handily, is zero. All together that gives us exactly what we need for an axis.
Here it will be the X axis. It is entirely subjective. We like what we like and don’t like what we don’t like. (And don’t care about what we don’t care about.)
The Y axis is quality, which we take to be (at least somewhat) objective.
It involves such quantifiable things as originality, complexity, nuance, boldness, texture, beauty, engagement, universality, authenticity, and skill level. (To name a few.) There is a subjective aspect to these things (who is to say whether more or less complexity is better), but there are objective references as well.
There are fewer objective references for badness. Shows that are essentially voyeuristic, for example, might qualify. This obviously assumes voyeurism is bad, which might be arguable. But there are topics or ways of telling stories that most would define as objectively bad (on content grounds).
In any event, we end up with:
I haven’t mentioned what we’re judging the taste and quality of, but the example shows examples of one obvious domain: stories.
The idea is that for any given topic, there is a roughly objective quality to it, but whether people dislike or like it will vary considerably. So generally a given topic appears as a fuzzy horizontal cloud of dots (individual opinions).
For purposes of illustration, I’m assuming “Shakespeare” (by which I mean the stuff he wrote) is objectively good. That is certainly the general consensus. Some will find Shakespeare better than others, so it’s not a hard line — there’s some vertical volume to the cloud.
But what people think about Shakespeare is (literally) all over the map. Some dislike it for whatever reason, others love it. (We’d likely get slightly different clouds for each individual work of his. Some plays would get more likes.)
I’m equally assuming “Reality TV” is objectively bad, although again there is a spectrum of people who like it and dislike it. Many of our guilty pleasures probably fall on the objectively Bad side. Usually because they’re Bad for you.
In the middle of the space is a zone where something is neither good nor bad, and people really don’t care about it. (Note that both Shakespeare and Reality TV also have zones of “Don’t Care.”)
We would not expect to see something like this:
It’s almost incoherent if applied to a single topic. The implication is that people are all over the map in judging (supposedly objective) quality, but they’re largely unified in their personal opinion about it (in this case everyone likes it).
I’m not sure which is more incoherent, the idea of diverse judgements on quality or everyone having the same opinion of it.
I can think of things nearly everyone likes or dislikes, but not many. If we’re talking about stories, those are probably empty sets. You’d think maybe candy, pizza, or music, but there are people who don’t care for those. I think it’s pretty nearly impossible to find universal tastes.
As I mentioned above, different works of Shakespeare might get different quality judgements, so where we might see vertical spread would be in the case of considering many different works.
Having them all be (in this case) works one likes is still weird. Given a spread of different works of different quality, we’d expect the space to look more like this:
Which, in 2D, is reminiscent of how opinions of Neapolitan ice cream are likely to be. (The difference in that configuration space is all three axes are degrees of how much someone likes something, from zero to 10. That space is entirely subjective opinion, so naturally it’s filled with diversity.)
The point that I really want to make here is that the regions we identify in configuration space are fuzzy. This is due to the distribution of many individual data points.
In some cases, the points come from one source making multiple judgements — for instance one person evaluating a lot of different books. That would likely generate a plot similar to the one directly above. That person’s liking and judgement of quality would vary by book.
In other cases, the points come from multiple sources considering a single topic, such as in the case of Shakespeare or Reality TV. In these cases we expect the distribution of points to be (much) more horizontal than vertical, as in the first plot.
This was a very simple configuration space — only two dimensions, so it’s easy to diagram and visualize.
Three dimensional spaces (such as the Neapolitan ice cream one) are nearly as easy. But when the configuration space has lots of axes (as with Baskin & Robbins 31 Flavors) it’s much harder to visualize the space.
Some configuration spaces have a much larger number of dimensions (in a few cases, infinitely many). Not hundreds or thousands, but millions or billions. (All we can really do with such spaces is treat points as long lists of numbers.)
But simpler spaces that can be visualized, including a few more than three dimensions which we can intuit from 2D and 3D, can be helpful mental tools for thinking about something.
For instance, in this case, making it visually clear that taste and quality are orthogonal concepts — they are independent of each other.
But configurations spaces are also useful in dealing with complex, fuzzy concepts. We can understand such as clouds of points in a configuration space. That’ll be important in the next post.
Stay safe, my friends! Wear your masks — COVID-19 is airborne!