Essay: Heuristics


If we’re tasked to design the minds of our rabbits for maximal reproductive output, we’re subject to computational and cognitive constraints. Therefore, we need to make use of rules of thumb, or heuristics.

  • Why do we like the things we like? This falls out of a combination of “innate” heuristics and learning, mostly classical conditioning.
  • Tastes can be heuristics for our ultimate goal, reproduction. And tastes or heuristics can be developed for goals based on lower-level heuristics, so tastes can become “ungrounded” and arbitrarily complex.

Until he retired, my uncle worked as an engineer at Ford in Detroit. He had bought one of their new models (a 1997 Taurus), and said that although Ford’s model was more aerodynamic, Chevrolet’s competing model had higher sales because people “didn’t like the way the Ford model looked”. And the next model (the 2000 Taurus) ended up being redesigned with a more traditional –and less aerodynamic– shape.

Why wouldn’t people prefer the more aerodynamic model? Isn’t the goal to have the most “functional” car? If not, what else might people be trying to optimize?

One thing to note is that no one has a perfect wind tunnel simulator in their head. Instead, people have to rely on heuristics or tastes. There’s a way that people develop their tastes in cars. People see both the designs and the performance of fish, jet fighters, and cars. After many such examples, people develop an intuition for the kinds of features, such as sleekness, that make for a high performance thing-traveling-through-medium. People then associate these features with their performance. So, without ever having taken a course on aerodynamics, people can eyeball a car’s body design and tell you (with better than random accuracy) whether it will outperform some other body design. This is in contrast to the more analytical conceptual system of car design which is what the designers at Ford use when they apply equations and principles of aerodynamics to come up with a more exact answer for how aerodynamic the car is. In reality, most systems in our heads (and in the heads of the designers at Ford) are both intuitive and analytical.

This is the basic principle of how tastes are developed in general. Like peahens, our perceptual abilities are limited (where we can’t tell a car’s drag coefficient just by looking at it). So, we have to develop associations of features to “value” (speed or performance in the case of cars) so that we have an intuition to give us some idea of the value.

In stock market analysis, one can take a technical approach or a fundamental approach. A purely technical analysis of a company’s stock looks only at the “trends” in the stock’s selling price. Books are written about how to “predict” what a stock will do given only the history of the stock’s price. Generally, if a technical investor predicts that the stock’s price will raise, he’ll buy, and sell if the technical analysis predicts the stock will drop. In its most extreme form, a technical analysis wouldn’t even look at what the company does. A fundamental analysis is at the opposite extreme: it will ignore the history of the stock’s price, then make a prediction of the company’s earnings (and dividends) based on factors such as what capital the company owns, the “quality” of the people working for the company, and whether the current economic situation means that people will want to buy the company’s goods. You can then assign the stock a fundamental value based on the expected dividends and interest and inflation rates. (For example, if you expect a share of Bifislurf Inc. to yield $5 in dividends over the next year, and the inflation-adjusted interest rate is 5% per year, then a share of Bifislurf Inc. would have a fundamental value of about $100, because that’s how much money you’d need to put in the bank to get $5 of interest in a year.) Generally, a pure fundamental trader will buy if the stock’s current price is (significantly) less than its fundamental value and sell if its price is higher. An interesting thing is that if everyone invested solely on fundamentals, the plot of the Dow Jones Industrial Average would be much smoother, practically flat because it’d only reflect “real” changes in the companies’ values. (Tulip Mania also would’ve never happened if the tulip traders used only fundamental analysis. Tulips have little intrinsic value.) In reality, most traders use results from both fundamental and technical analyses.

Another example of fundamentals and technicals: Flugtag is a contest/event sponsored by the Red Bull company in which teams build non-motorized flying contraptions and launch them (carrying one of their members who is the “pilot”) off a 30 ft. high ramp into a pool of water. The teams are judged primarily by 2 criteria: 1. the distance flown before landing in the water, and 2. the “creativity” of the contraption’s design. If it weren’t for the 2nd rule (which I’ll call the “technical” rule), Flugtag would be a much more boring event. What would happen, I predict, is what happens with a myriad of other “purely fundamental” pursuits. Initially, there’d be a broad range of designs, but eventually one of the designs (or its basic principles, at least) would emerge as the “optimum”, and most entrants would be minor variations on this optimum. Take airplanes, for example. After the Wright Brothers’ success, there was a blossoming of all sorts of crazy designs. Take the Langley Flyer, for example. It actually predated the Wright Brothers’ Flyer by several years. Its full scale model was never fully capable of sustained flight, but the smaller models had some success. The interesting thing is that it looks completely unlike any airplane I’ve ever seen. By World War II, the basic body of the airplane (with a single aerofoil wing) converged to what’s still used by commercial jets. A similar process happened with locomotives, automobiles, and computers. Thus, Red Bull’s 2nd rule explicitly puts a limit on the kind of convergence that would otherwise happen.

Suppose I want to evaluate a piece of clothing. For instance, I have fundamental criteria such as how comfortable a shirt will be, whether it’ll keep me warm on chilly days, and how difficult it’ll be to wash it. I also might be concerned with how I’ll look in it, and I won’t be looking at myself much. So I’m really concerned with associations other people will make of me when I’m wearing the shirt.

Now, there’s one more factor of taste: associations made of people based on their appearance. One can take a Holmesian approach and use logical reasoning to deduce things about a person from their appearance, but we don’t have enough time/brainpower to do this with every person we meet, so we have to rely on our intuitions (or associations developed through experiences), just as we might in determining how “functional” a car will be just from eyeballing it.

Some associations are formed from fundamentals. If I see a picture of a person wearing a heavy coat and hat, I’ll “guess” that the person’s somewhere cold. This is because a heavy coat keeps you warm regardless of the coat’s social context. If a person’s wearing glasses, they probably don’t have perfect vision without them. Fundamentals are pretty easy, at least compared to technicals. These don’t have to do so much with the intrinsic properties of things, but more with what the thing means for other people. With clothes, technicals would be “ornamentation”, such as printed designs, the particulars of the cut of the fabric, and “flair”.

I’d guess fashion designers spend 99% of their thought on technicals. Ornamentation has its own abstract (and often intuitive) “rule system” as well. Like the peacock, you can’t stray too far from the status quo. The “fundamental” problems of clothing are pretty straightforward. Like the stock market, if clothes were designed only by fundamentals, fashions would hardly change at all. (Changes would only be with innovations such as with materials and manufacturing techniques, and with changes in what people use the clothes for. For example, a much lower percentage of Americans farm than was the case a century ago, so the average American doesn’t need clothes designed for farm work.)

So, how do technicals and intuitions formed about technicals work to influence fashions? Take bell bottom pants, for example. The theory is that attractive people originally started wearing bell bottoms. (The reason for doing so may have been to distinguish themselves from “the masses”, or perhaps for the same reason very fit gazelles will flaunt their fitness in the face of approaching lions as if to say “Don’t waste your time chasing me, look how fit I am.”.) Then people began to associate bell bottoms with being attractive. Then, (knowing about this association (or having the association themselves), and maybe it’s not explicit) less attractive people began wearing the bell bottoms so the association would be transferred to them. Eventually, so many unattractive people started wearing them, that after a lag, the association became extinguished, and there was little reason to wear bell bottoms.

The book Freakonomics talks about similar trends with babies’ names: people of high socio-economic status start naming their kids with a particular group of names. Then, the trend catches on because other people see that “The Beautiful People” are called by these names. Thus, people (from lower classes) form the association from the name (a nearly arbitrary symbol, practically) to the person’s status, and name their own kids with that name. Eventually, the name becomes “common”, the association is extinguished, and the upper classes find new names.

This association from technicals to fundamentals might be why people have certain tastes in food. Oncologists have long known that you can cause people to develop a strong aversion for almost any kind of food simply by putting their chemotherapy medicine in it a number of times. My theory is that you can similarly cause people to like just about any flavor by creating a fatty food with that flavor: I remember the first time I had the Greek candy halvah. My friend, Charles’s mom, Zabia (who’s father was Greek), offered it to me. Despite its taste, which I would’ve described as awkward but not bad, I ate it out of curiosity and politeness. I had halvah several times after that, and soon developed a taste for it. Once, I found some commercially packaged halvah, and I read the label: halvah’s made from crushed sesame seeds, and is about 20% fat and 60% sugars by weight. My theory is that fundamental “yumminess” is mostly fat and sugar, and that eventually we (or our taste buds and the associated brain areas, to be precise) associate the flavor with the fat and sugar content.

I’d hazard to guess that virtually anyone can be conditioned to like almost anything. (Well, anything that a large group of people also enjoy. There are probably few people who could be conditioned to enjoy having their toenails removed.) Through conditioning, almost anyone could probably be made to enjoy the flavor of chocolate and dislike the flavor of vanilla, or vice versa. Similarly, barring physical disability, associations and reinforcement could be used to cause almost anyone to be made to enjoy mountain climbing, painting, or knitting. To do this, we’d just need to cause the association of the features of these activities with more fundamental rewards.

This raises the question of what the fundamental rewards for people are? For our rabbits, they should be those that, when coupled with the rabbits’ cognitive systems, causes them to survive and reproduce. Some rewards might be innate even though they could be learned from more fundamental rewards. For example, it might make sense to install in our rabbits an innate desire to not fall from high places (as a form of bootstrapping), even though this desire is a “corollary” of the more fundamental desire to not break bones. So, some of our rabbits’ desires could be redundant or even contradictory.