Tea Party Technologists
November 18, 2011
"String theory - How's that working out for you?"
When you work in
industrial research, you work with a lot of different people, not all of whom are devotees of science. Many are just there for the money. There are the technicians, some of whom seem to spend a large fraction of their time comparing notes on motorcycle maintenance;[1] the managers, who use their spreadsheets only for financial calculations; and the rest of us, who try to do precise measurements with decade-old voltmeters, since we're not allowed the purchase of new capital equipment.
Industrial research has been afflicted with various process philosophies over the years. We suffered through Total Quality Management, Six Sigma and similar business processes as we were carried along on the tidal wave of these movements in other areas of our corporations.
Although many precepts of these business processes did cause increased efficiency among our pencil-pushing colleagues, most of them didn't make sense in a laboratory environment. They were merely unwanted distractions. There was, however, one process that was exclusive to industrial research labs. That was Design of Industrial Experiments, sometimes just called Design of Experiments (DOE).
I took a short course on DOE many years ago. The first thing that I learned was that I had been practicing its core principle, linear regression, for many years. Crystal growth is a complicated process, so much so that its practitioners were said to practice a black art, and not science. There are so many variables that it often appears accidental when you achieve your desired result; namely, a large enough crystal for an experiment.
In the early days of bubble memory technology, a group of scientists from Bell Labs defined a set of parameters that were useful in describing a liquid phase epitaxy process for the growth of garnet crystals. I would track certain properties as a function of these parameters to zero-in on a desired result. The best mathematical technique for doing this is linear regression, and I developed a series of computer programs to facilitate such an analysis.
As a first step in the DOE methodology, you list all the variables of a process. Then, you decide what would be a low and high (and sometime middle) value for each variable. After that, there's software available that tells you what experiments you should do. For example, the first experiment could be done when all variables are high, you do the experiment and measure the result. The second experiments could have all the variables high, except for the second variable, and so on.
If there are just two, or three, variables, you can try all combinations of low and high variables. If there are many variables, you would be allowed to skip some experiments. In the end, you plug all your data into some software, and you see which variables are the most important, and where the variables should be set to get your desired result. For more control, you can narrow the range of variables and do another DOE.
The instructor of this short course was quite enthusiastic about DOE, to the point of claiming that it could solve all the world's problems; or, at least, any problem we might encounter. At the end of the course, each participant was asked to present a DOE for a problem he would like to solve. I used this as an opportunity to make a valid point. My DOE was on winning the New Jersey State Lottery. The variables were things like where I bought my ticket, the time of day I bought my ticket, and so on. This was a counterexample to the idea that all problems can be solved with DOE.
One feature of the DOE methodology is that it's theory-free. In fact, you're instructed to ignore all prior beliefs and crank though all the experiments, even those that seem absurd. It's the theory-free part that's always troubled me. DOE is advocated by corporate managers who distrust theory or don't take the time to discover how theory can guide their process. Worse yet, at least for scientists, is that they consider funding for scientific research to be wasted money. I call these people the Tea Party Technologists. They want their technology, but they want it unencumbered by science.
The Tea Party is a loose political movement in the US with varied and nebulous aims. Members of this group are always looking for the "literal" interpretation of everything from the US Constitution to the Bible, and they are likely to be fundamentalist and conservative. Conservatives and fundamentalists are generally anti-intellectual and anti-science. They don't believe in evolution or global warming.
"Americans Throwing the Cargoes of the Tea ships into the River, at Boston," an engraving from a 1789 book on the history of North America. This was known afterwards as the Boston Tea Party. (Via Wikimedia Commons).
There was a time when scientists and intellectuals were revered in this country. Such reverence is quite apparent in the newspaper and magazine stories about Albert Einstein that were published during his lifetime; or, in the articles about "rocket scientists" during the early years of space exploration. Now, the motives of scientists are suspect, as if we are really trying to foist a bogus threat of global warming on the world as part of some evil scientist's plan you might find in a James Bond movie.
Unfortunately, the realization that policy-making needs scientific advice may only come in an "I told you so" moment when global warming has forced people to flee their seaside dwellings and move to high ground. Somehow, scientists convinced people to stop ozone depletion, so maybe there's still hope to stop global warming.
Reference:
- The Robert M. Pirsig novel, "Zen and the Art of Motorcycle Maintenance," was a cult classic while I was in graduate school.