Friday, April 21, 2006

In Search of Good Metaphors   posted by Matt McIntosh @ 4/21/2006 06:15:00 PM
Share/Bookmark

When trying to get across the gist of complex technical subjects to a layperson, sometimes a good metaphor can do more pound-for-pound explanatory work than hurling large amounts of jargon-laden details.

Fore example, Gary Marcus provides a lucid explanation of gene expression by analogizing each gene to a conditional statement in a software program. Each gene has a set of conditions under which it's expressed (IF/WHILE), and a specific protein it generates when activated (THEN/DO). While this is most readily intelligible to people with experience in computer progamming, it can be understood by intelligent non-programmers with a little coaxing and can clear up several confusions and help order one's thoughts:
  • So-called "junk" DNA is just dead code -- code whose execution conditions are never met. Piles of this stuff can accumulate during the development of sufficiently large and complex software programs, and that's with intelligent programmers watching over it. Small wonder if orders of magnitude more happen to accumulate under the guidance of a completely blind process.

  • The oft-quoted trope that we're 98% similar to chimpanzees at the genetic level loses significance; even a 2% change in the source code of a program, particularly on the conditionals, can have massive effects on how the program executes.

  • The sillyness of opposing genes versus environment or "innateness" versus "plasticity" becomes evident -- these things are orthogonal to eachother rather than in conflict. A program can be specified to execute multiple different ways given different inputs, and even if the brain starts with a highly specified innate structure, it could be programmed to rewire itself based on environmental inputs.

And so forth. Another good example for the numerate is the definition of race that got Steve Hsu's comments deleted by Brad DeLong: represent each individual's genome as a point in a space of extremely high dimension, and define a race as a set of points whose distance from each other is less than some radius. These clusters map onto intuitive self-identified race with a very high degree of accuracy.

A third example (which I first encountered from Henry Harpending [PDF]) is analogizing intelligence to size. Many opponents of IQ testing start from the correct point that there is no single definition, measurement or task that completely captures everything we consider "intelligence," but then slide from there into saying that IQ is meaningless and intelligence can't be measured.

If you substitute "size" for "intelligence" it becomes apparent just how silly the argument is. Height is one dimension of a person's size, as is shoulder breadth, weight, the length and circumference of one's limbs and so forth. There are short guys who are built like tanks and tall guys who'd blow away in a strong wind, but nobody takes exception to the concept of "size" on their account or argues that size can't be measured.

This cuts both ways: it also suggests that attempting to reify g is a category mistake. Like the concept of a center of gravity, g is an abstractum, a theorist's fiction -- but one that is well-behaved and has a causal "reality" to it all the same. We can predict that someone with a high g will perform well on all kinds of cognitively challenging tasks just as we can predict that a chair tipped on its back legs past a certain angle will fall over.

Since GNXPers are used to dealing in complex subjects which are very misunderstood and commonly regarded with perplexity (if not suspicion or outright hostility) by laymen, I thought it might be worthwhile to solicit whatever other metaphors y'all have found useful in both understanding and explaining technical concepts. They don't have to be biology-related, though I expect most will be. Discuss amongst yourselves.

Update: John Wilkins responds thoughtfully to the brief treatment of race here, but I think he and PZ Myers, both professed Lewontinites, ultimately miss the boat. John gives it away thusly: "So, do I think there are races in biology as well as culture? No."

Asking whether race exists in this way is a category mistake, albeit an all too common one. Race is another abstractum, like the general intelligence factor. How we define it will be wholly a matter of convention, though not totally arbitrary because some definitions obviously have more utility than others. Returning to the initial representation I used, you could shrink the race-radius right down to the point where each individual (plus his twin, if he had one) was his own distinct "race" if you wanted to, but this wouldn't be interesting. If all that one means by "race is a social construct" is that one can can twiddle the granularity of racial categories virtually however one likes, then this is perfectly true and perfectly beside the point.

Because at bottom, all this abstraction and definition is based on a real molecular substrate. Piles of rock and dirt don't change their height based on whether one choses to call them hills or mountains. Hypertension doesn't suddenly stop being more frequent in African men based on how one decides to classify them. Genetic variation between populations doesn't become any less real. Eppur si muove.

Update the 2nd: GMTA. Razib goes on at greater length and depth, as usual.