Monday, February 11, 2008

Inherent Error

My father used to tell the story of a meeting where, as an advocate for children with learning disabilities, he confronted a politician with evidence that a major law affecting the treatment of those children in the public schools wasn't working. The politician's solution was simply to tell the schools to follow the law. “How will you make sure they're following the law?” my father asked, incredulous. “I'll just tell them to,” the politician responded. For many years afterward, this exchange was used by my father as a prime example of how politicians, like many managers in industry, were more interested in appearing to do something than achieving verifiable results. In my first career as a test engineer, I came to refer to this phenomenon as “the appearance of functionality.”

In the personality cult that has become the Republican Party (to which I once proudly belonged with my father), I see the same dynamic at work regarding the assessment of new laws that grant the president vast new powers, the most recent example being the updated FISA law. If the president's intentions are good and the cause is right, which many Republicans in Congress seem to believe, then a law can be approved without any consideration of oversight. These politicians are unwilling or unable to accept any challenge to their beliefs, especially if the challenge comes from people of a different political persuasion. Any obvious red flags are ignored, such as the new FISA law's provision for retroactive immunity to telecom companies for unspecified actions.

Suppose, as in my father's example, the politicians who pass a law have honorable intentions; and fully intend to follow its provisions. Is oversight really necessary? The answer, I would argue, is still a resounding “Yes!”

As a test engineer, I saw over and over again how even with the strictest controls and most up-to-date technology, physics and humans (like it or not, just another part of Nature) could never get repeatedly closer than ten percent to any given goal. The obvious and best way to deal with this fact of life is to set one's goals at least ten percent higher than what is required (conversely, if someone loosens a specification rather than tightening it, then the error becomes ten percent of a larger number). I also noticed that quality tended to slip when the amount of testing was reduced, often in response to tight schedules; so I walked away with another important lesson: Don't take ANYTHING for granted.

Faith, I've concluded, is dangerous when applied to anything important. With the strictest oversight, a law, like a product specification, can be expected to be properly followed 90 percent of the time. Without oversight, that fraction may be as low as 50 percent. When evaluating a new law or a change to an existing one, we and our representatives must ask ourselves if we can live with the consequences of the inherent error, and adjust the law accordingly.

No comments: