Computational Aesthetics

computational aesthetics
Computational methods applied to problems whose success is measured by natural (usually human) aesthetics. Also "automated aesthetics".

I made this term up to describe what I like to work on. Not only is it succinct and descriptive, but I figured that other people wanting to make up a term for it would probably pick the same one. That turns out to be true. I recently did a search and discovered that lots of other people do use the same term, though many of them use it to describe algorithmically-generated art, which is a narrower view than mine.

Negativland sums it up perfectly in the intro to Escape From Noise. Of course, they were being facetious, but it's perfect anyway.

Modeling aesthetics is interesting because understanding people is interesting, but it's also useful when trying to make predictions. That's what models are usually for, and this area is no exception. Predicting what someone will like is tremendously valuable, and I think that improving the accuracy and reducing the cost of predictions for individuals will be increasingly valuable in the coming decades as we have more personal data available. That could mean "your data available to you and agents you trust" or it could mean "your data available to everyone and beyond your control", depending on people's priorities and how well they express them through commerce and regulation. I'm not interested in spying on people, but I'd like to permit my computer to spy on me and behave in useful and appropriate ways.

PizzaZone is an example of something that uses computational aesthetics, though it currently doesn't learn except through explicit instruction. My MSc Dissertation is another example.