Sunday, April 21, 2013

Excel in Historical Perspective

Much has been made recently of the embarrassing discovery that a fundamental proposition regarding the relationship between national debt and economic growth was the result of an error in computation. The fact that this error involved the faulty manipulation of a spreadsheet in Microsoft Excel just added an additional spice to an already pungent story, and today that spice inspired Irish comedian Colm O'Regan to write a piece for the BBC News Magazine Web site entitled "The mysterious powers of Microsoft Excel." The article has several bits that would probably play very well in a standup routine. My favorite is probably the following:
As much as oil and water, our lives are governed by Excel. As you read these lines somewhere in the world, your name is being dragged from cell C25 to D14 on a roster. Such a simple action, yet now you'll be asked to work on your day off. It is useless to protest. The spreadsheet has been printed - the word made mesh.
Reading this I realized that we now live in an age of Excel specialists who neither know nor care about this powerful system's origins. Those origins go back to what used to be a required subject in applied mathematics known as either "Numerical Analysis" or "Numerical Methods." This discipline tended to be held in distain by "pure" mathematicians, who believed that skill in manipulating symbols was always superior to that of manipulating numbers. As an aspiring undergraduate in pure mathematics, I remember seeing a door in the Mathematics Department at the Massachusetts Institute of Technology labeled "Numerical Analysis Laboratory." I used to wonder just what was in there. One day the door happened to be open, and I saw a room full of calculating machines whose most advanced feature was that they were powered by electricity, rather than a manual crank!

During the Second World War, there was an increased recognition of the need to get useful information out of highly complex mathematical formulas. Much of the Manhattan Project involved such efforts, as did the origins of the digital computer as we know it. I remember that the original ENIAC was promoted on its power to take radar data from the launch of a missile or torpedo and compute the entire trajectory before it hit the target. (Whether there was enough time to do anything with this information once you had it was beside the point.)

Thus, what made that Numerical Analysis Laboratory amusing was that, by my freshman year, which began in 1963, it was a room full of obsolete technology. As a result, Numerical Analysis was one of the first courses to migrate out of Mathematics Departments into Computer Science Departments. Knowing the methods was no longer sufficient. Knowing how to implement them in computer programs became more important.

Now, for many of my generation, the classic textbook in Numerical Analysis was Numerical Methods for Scientists and Engineers by Richard W. Hamming, a researcher at the Bell Telephone Laboratories. However, this book was most memorable for its epigraph following the title page, printed in large capital letters:
Since Hamming's book was first published in 1962, computing has gotten easier and easier; and insight seems to have suffered from smaller and smaller supply. The idea that economic policy may be at the mercy of as typographical error in an Excel formula is disheartening, but it is far from the worst abuse of insight we have encountered during the rise of mindless computing. I still think that a major nadir was hit during the Vietnam War at the time when military strategy was dependent on mathematical models of combat situations.

The way those models worked was that a research would define a military engagement in terms of some set of parameters. Those parameters would be the input to the model, and the output would provide quantitative information about the result of the engagement, things such as territory gained and lives lost. I believe it was Daniel Ellsberg who wrote about how every set of inputs to every imaginable model indicated that we were in a losing situation in Vietnam. However, when these results were presented to Pentagon "brass," all they did was ask for a set of parameters for a model that had us winning. So it was that Ellsberg changed from a detached researcher into a radicalized activist!

These days insight is in such short supply that most people no longer notice that it is missing. For them Excel is the perfect GIGO (garbage-in-garbage-out) engine. They deserve what they get, but do the rest of us deserve it?

No comments: