You get what you select for

Selection is a stunningly powerful force. You start with a large and varied population, select a small percentage of that population based on some heritable characteristic, let the selected winners increase, and repeat with greater stringency. The results--whether it's natural selection, the force that Darwin invoked to generate all of life's diversity, or artificial selection, which humans unwittingly do to breed antibiotic-resistant superbugs--can amaze and confound.

When I was learning how to do genetic analysis, I was cautioned that "You get what you select for." This can be nice. To use an example from my postdoc, if you're interested in how bacteria "smell" their environment, you can create a large and varied population by taking trillions of bacteria and mutating the heck out of them. Then you can use selection: if you want bacteria that can no longer smell a particular substance, you only let the scent-blind live--all the rest you murder with antibiotics. Repeat this a couple of times, and in a week you'll get a large population of bacteria unable to smell. You can then study those guys and write a bunch of papers and become famous for understanding bacterial chemical perception. You get what you select for, right?

You absolutely get what you select for. In this case, you can tell if your bacteria aren't smelling something because they don't swim away from that substance. You kill the ones that do smell it with antibiotics. So, you have selected for non-smellers--AND non-swimmers, AND those that are spontaneously resistant to antibiotics. Worse, there's only one gene that you can mutate to eliminate the ability to smell that target substance--a very small target for mutagenesis to find. However, there's about 5 genes that can be mutated to give you antibiotic resistance, and dozens of genes that can be mutated to give you an inability to swim--a much larger target. So, you think you're selecting for inability to smell, but you're REALLY selecting for a bunch of different things, and you get what you select for--almost all the cells you select will have a swimming defect.

I'm reminded of this again and again, most recently by what's been happening with the public schools in Washington D.C. The schools were doing very poorly, with many students performing far below grade level. An aggressive superintendent came in, urgent to shake things up and demanding results. A perrenial question in education is how best to measure results--in this case, learning progress was measured by standardized testing. So, the new superintendent selected for improvement in standardized tests.

This was not exactly like genetic selection: teachers and administrators bringing in low test scores were not killed, but they were terminated. Those who brought in ever-increasing test scores kept their jobs and received bonuses. New teachers and administrators were brought in, all subjected to the same selection pressures. Over time, the population of teachers and administrators in the D.C. schools changed.

Well, you get what you select for. The superintendent did not select for improved learning. She selected for higher test scores, so that is what she got. It seems that teachers and administrators tinkered with the standardized tests--edited them after the students took them, changed lots of answers, corrected them, to produce the desired result. Others just gave their students the answers as they took the tests. Those who did were selected for with continued employment, awards and bonuses; those who used more ethical but less immediately effective means of changing test scores were selected against.

Selection is powerful--often more powerful than we can appreciate. The creativity of nature, when asked to provide a solution to a selective challenge, never fails to amaze and confound. We should not be surprised when the creativity of human nature comes up with similarly confounding solutions to selective challenges.