People fusion

“Fusion” is an important concept in nearly all artistic fields — music, cooking, painting, what have you. Each endeavor tends to feature multiple strongly-identified styles — Mexican food, Japanese food, Ethiopian food…; Jazz, Classical, Rock…; Impressionism, Abstract Expressionism, Pop… — and it is fun to mix and match to obtain exciting new combinations. Here in Chicago, you can visit SushiSamba for Latin/Japanese cuisine, then head over to the Empty Bottle to listen to some jazz/rock hybrid.

Despite the excitement, however, genres do not just blend together to form one homogeneous goop. While styles evolve, they are typically held together by distinct aspects setting them apart from other approaches. (Latin/Japanese fusion will never grow in popularity to displace the authentic cuisines of either region.) This can be informally understood through the concept of a fitness landscape, a function of the underlying variables that describes how successful a certain approach ultimately is. A typical fitness landscape has peaks and valleys, indicating that particular combinations work well together, much better than a random mish-mash. Imagine setting aside our delicate sensibilities for a moment to contemplate a giant “food machine” (or “music machine” or whatever) that can create any dish we want, simply by adjusting the position of a large set of dials. (A thought experiment, okay?) There’s a dial for the amount of jalapeno peppers, another dial for how long the food should be cooked, and so forth. There is a region of dial settings that corresponds roughly to “Japanese food” and another that corresponds to “Latin food.” If we simply adjust the dial setting to be a linear combination of the Japanese and Latin settings, the likely outcome is — some sort of horrifying mush. We would find ourselves in a valley of the fitness landscape where nobody would want to live. In other words, different approaches to cuisine (or other artistic endeavors) tend to cohere into sensible and distinct groupings, and random mixtures between them are unlikely to be an improvement; successful fusion is a delicate art.

Interestingly, however, there is a well-known counterexample to the peaks-and-valleys structure of aesthetic fitness landscapes: human faces. It’s been known for a while now that if you take a selection of randomly-chosen people, and construct a picture by averaging their features together, the result is typically a more attractive person. The phenomenon is extremely easy to notice in examples. Alex Tabarrok at Marginal Revolution writes about an experiment that tested this conclusion. He extracted this figure from Judith Rich Harris’s book No Two Alike. From top to bottom, it shows some actual faces, then the average of two faces, four faces, and so on up to 32 faces.

Averaging People

For an even more striking demonstration, see The Face of Tomorrow, a photography project that takes portraits of random people in certain cities around the world and blends them together (city by city), with the idea that these composite portraits will resemble future citizens in an increasingly interconnected world. (Alas, real heredity doesn’t work that way — stupid discretized genetic code.) Go to The Faces and click on a city to see the composites decomposed into the individual people. The future, I have to say, looks pretty hot.

This entry was posted in Miscellany. Bookmark the permalink.

8 Responses to People fusion

  1. citrine says:

    Great post, Clifford!

    Nancy Etcoff goes into the People Fusion issue in her book “Survival of the Prettiest” (about humans being hard-wired to respond to beauty – which we subconciously equate to fertility). She says that throughout most of human evolution, traits such as asymmetric features were generally due to ill-health or some kind of physical weakness and therefore we learnt to correlate healthy stock with average (as in not too deviant) physical traits. Thus features that are statistical averages became embedded in our subconscious as appealing.

  2. citrine says:

    Oops!! I read several posts by Clifford and then came to this.

    Great post, Sean. 🙂

  3. anonymou says:

    Alas, the discrete nature of genetic inheritance means that future people are unlikely to look more “average” than people today. Darwin puzzled over this as well, wondering why natural variations weren’t “averaged out” over succeeding variations. Mendel put it straight.

  4. jay says:

    I always wondered about this. When we have a large number of something differing in some physical property, we usually see a distribution more or less peaked around the average value of the physical property. But, why isn’t it the case with the people’s faces? The averaged face, the most aesthetically pleasing one, is so rare that you got to be extremely lucky to have one close to the average. Does anyone know the answer?

  5. Jack says:

    I’m afraid that the answer in this case may be rather mundane: the averaging process appears to “airbrush out” facial blemishes! Thus the average face seems to have perfect skin! The superposition principle: on average, you can expect, at any given point, that a bump in the skin will cancel out that nasty acne pockmark!

  6. jb says:

    I’d like to see averages of different pictures of the same person. It seems possible to me that anyone would look better when averaged with several pictures of themself. If those pictures look just as good as averages of different people, then it would seem like a bad sign for the people-fusion theory.

  7. Cynthia says:

    Human aesthetics encompasses a haphazard route towards a nondescript landscape of flatness. In contrast, the culinary and musical arts entail an orderly trajectory through a picturesque field of peaks and valleys. Therefore, a landscape explorer needs a minimal degree of navigational skills to traverse the meandering path towards human beauty. On the other hand, a landscape explorer needs a maximal degree of navigational skills to charter the finely-tuned course towards great tasting food and great sounding music!

  8. Pingback: Arbitrary Chronological Signifiers | Cosmic Variance