Nobody Tell Tyler Cowen

I believe there is an acquired tendency of behavioral psychologists, sociologists, economists, geneticists, etc. to draw broad conclusions that are not supported by evidence.

"Adhering to logical, rational principles of ideal economic choice may be biologically unnatural," says Colin F. Camerer, a professor of behavioral economics at Caltech.

It may be uncommon, but I cannot see how he can conclude that it is 'biologically unnatural.' Do we have genetic predispositions to illogical economic behaviors? Is there some complex of genes that produces such behavior? If so, where are the genes? Or, is it just that most people aren't very logical or rational in all their decision-making tasks?

I was talking to a neuroscientist the other day who explained his current working assumption about how the brain operates. He sees it as a sort of giant Monte Carlo system where different neural pathways predict different outcomes for a given set of inputs. As the game iterates, the ones with the best predictive value are favored and added to while the worse ones wither away.

If true, one can speculate how we might be full of patterns that served us well on the primordial savanna but don't fit the current world so well. However, my other takeaway from this very interesting conversation was that the structural complexity of the brain is astounding and our ability to model its processes is still really in its infancy. Our brains seem exquisitely evolved to solve problems like the Traveling Salesman that cause binary computers to bog down badly. His take is that we're just now starting to model insects with some precision, while rats are still some years off.

I wonder whether we will need to evolve new silicon architectures and software design paradigms to get there and beyond. Neurons are considerably more subtle than transistors, and we have a hundred billion or so of them. Parallelism is something we still have a heck of a time with in software and the brain is parallel beyond belief.

Long story short, though the field is terribly exciting, it seems a tad premature to expect too much explanation of higher-level behavior from it.

Dr. T wrote:

"I believe there is an acquired tendency of behavioral psychologists, sociologists, economists, geneticists, etc. to draw broad conclusions that are not supported by evidence."

Thank you for your hilarious example of drawing broad conclusions that are not supported by evidence.

From the article:

“Adhering to logical, rational principles of ideal economic choice may be biologically unnatural,” says Colin F. Camerer, a professor of behavioral economics at Caltech.

How does one determine what is "biologically unnatural?" Is a characteristic or behavior "unnatural" because it doesn't happen often?

Describing a human behavior as "unnatural" to me implies that the individual exhibiting that behavior made some effort to cultivate or enhance it. That effort could be supremely rational. Individuals could engage in altering or improving their behavior to signal to others that they are worthy credit risks, top notch investors, or good draft picks.

One more excerpt:

"Better insight into human psychology gleaned by neuroscientists holds the promise of changing forever our fundamental assumptions about the way entire economies function—and our understanding of the motivations of the individual participants therein, who buy homes or stocks and who have trouble judging whether a dollar is worth as much today as it was yesterday."

The behavioral economists aren't tilling new ground. They are trying to aggregate individual actions and decision making processes. To sneak around this fact they claim their insight may "change forever""our fundamental assumptions about the way entire economies function."

I've always thought that "Predictably Irrational" is a supremely rational title.

Cowen spends a great deal of time dispelling autism's societal stigma, arguing that mainstream society is reaping benefits from mimicking autistic cognitive strengths.

Tyler has a point, but I say equal time for obsessive-compulsives and transsexuals. OCD’s are ultra meticulous and organized, TS’s highly intelligent and creative. If we were to destigmatize the attributes associated with these disorders and mimic their cognitive strengths like we do those of autistics, wouldn’t this constitute a three-fold increase in societal benefits?

Snark,

"OCD's are ultra meticulous and organized"

That sounds like a broad conclusion that is not supported by evidence.

Character Strengths and Virtues (what the Obsessive-Compulsive type is proud of)

1. Industry, diligence.
2. Scrupulousness, conscientiousness, dutifulness, responsibility, idealism, highmindedness.
3. Deliberateness, judiciousness, rationality, logicalness, sensibility.
4. Having high standards; trying to be complete, perfect; radical, persistent, thorough, thoroughgoing.
5. Perseverance, tenacity, steadiness, firmness.
6. Orderliness, tidiness, cleanliness, meticulousness.
7. Prudence, self-control, self-restraint, carefulness, cautiousness, discipline.
8. Frugality, thriftiness, saving, conserving.

Most of these Character Strengths and Virtues were selected from a list of virtues compiled by Cawley, Martin, and Johnson in A Virtues Approach to Personality

Colin K wrote:

Parallelism is something we still have a heck of a time with in software and the brain is parallel beyond belief.

The hardware guys don't have much of a problem with parallelism. But most programmers, when they're faced with an inherently parallel problem, think, "I know! I'll use threads with shared state and monitors!" Now they have two problems. (With apologies to jwz et. ant.)

We have a heck of a time with parallelism in software because we won't give up state. It's like the old story (or possibly psych experiment) where the monkey reaches into the jar, grabs the fruit, then can't get his hand out.

And you don't have to go all the way to functional programming: Peter van Roy pushes what he calls "declarative" programming, in which you can have a kind of state, but each location can only be set once (kind of like single-assignment style). If you can adhere to that (and it's not much harder than any other programming paradigm), you get parallelism by simply throwing various chunks of the (initially serial) algorithm into separate threads.