Through a hazy crystal ball

I was having trouble sleeping a few weeks ago, so I turned on the TV and caught a few minutes of a late-night infomercial touting a psychic hotline. The infomercial seemed like fairly harmless entertainment if you didn't spend too much money on the psychic services and didn't take the predictions seriously. I smugly thought that everyone I know is too sophisticated and knowledgeable to even begin to take the psychic predictions seriously.

But my thoughts changed the next day while I was watching a major news show. Perhaps I am limited in my perspective, but I believe that "news" is a report of what has already happened. However, this so-called news show did not report on past events at all. Instead, one reporter said that the Federal Reserve Board would meet later that day and that the board would make certain decisions. Then a purported expert talked about what the board might do, but the expert's report was mostly speculative. Many of the other purported news reports also included forward-looking expert commentaries.

Suddenly, a light bulb (or LED) turned on in my head, and I realized that this news program was just a sophisticated, slicker version of the psychic-hotline service I had seen earlier. Instead of reporting on recent events, the well-dressed, polished reporters explained what might happen today. Unfortunately, I wanted to hear about what had happened yesterday, but the reporters apparently thought their insight into the future was more interesting. I suspect that they know that the desire to see into the future is nearly universal.

Another problem with these slickly packaged predictions is that reporters are held virtually unaccountable for the information they report. Reporters make these innumerable predictions, which have varying degrees of accuracy, but fail to track the results of their predictions or prove the accountability of the information. Even if someone took the time to check the accuracy of the predictions, we know by the law of large numbers and the bell curve that these predictions sometimes do come true.

Why should you care about this problem? If you are a designer or a project leader, your colleagues may ask you to make predictions and offer a few of their own. Some of these predictions may be reasonably straightforward, such as the impact of doubling memory requirements or the consequences of switching to a faster processor. Other predictions may be less clear, such as the effect on application code size, complexity, and debug time when you have to add new user features. For such cases, it's wise to try to anticipate obstacles for your project, such as the likelihood of having to delay a project's competitor. And look for the inflection points or discontinuities that most predictions fail to include or even allude to. It's easy and logical to build a simple linear or well-characterized extrapolation of available trend lines, but reality doesn't always work out so neatly.

You can never predict the future, despite your strong desire to, as the recent tragic events of September 11 have too clearly shown. For most engineering situations and simulations (which are also a kind of prediction), we have the insight and tools to make some reasonable assessments. But always keep in mind that the data you have and the algorithms you use to analyze the data and extend the results contain many inaccuracies, assumptions, and implicit estimates. When you need to, factor these uncertainties into your predictions and make them explicit if appropriate.

Author Information

Contact me at bill.schweber@cahners.com.