Posted Thursday, October 23, 2008 by
Damon Richards

A story about United Airlines’ 2002 bankruptcy emerged on the Internet on September 8, 2008, looking like a new story. UAL’s stock dropped 76 percent that day before trading was temporarily halted. Three days later, the stock had climbed back to $10.50. This demonstrates the amazing power of the Internet on real things in the world, and the incredible amount of influence an Internet posting can have.
What happened was that a Google search bot found mention of the bankruptcy on a Florida Sun-Sentinel page that had a date of September 7, 2008. The story was listed in a summary called “Popular Stories: Business.” If you or I had come across that page, we would immediately have known that it referenced the 2002 bankruptcy, but Google’s bot isn’t as smart as you or me. It dutifully summarized the story with the date it found on the page. Next, the story ran on the Google News service, where it was picked up by Income Security Advisors and posted to the Bloomberg service. From there, more people searching for details made the story more popular, which moved it higher in search results which led to more people seeing it. And UAL lost 76 percent of its value.
Lots of little things went wrong to make this happen. Any one of them could have been prevented by having a human involved. Unfortunately, there is too much information that we deem important, but usually not critical, to be able to afford a human watch it. Processes get automated and computers sometimes make mistakes because they have no judgment. Similar errors happen every day in the routine monitoring of a small business’s computer system.
At our
Indianapolis computer consulting company, we use automated tools to monitor many routine functions of our small business customers’ computer networks. We rely on these tools to be “smart” enough to alert us when we need to turn some human attention to something. It doesn’t always happen the way we plan, so sometimes we miss something. We can usually resolve the problem without causing a 76 percent loss in value. When one of our automated processes goes awry, we tighten up that process to avoid that happening again. We have plenty of new mistakes available to us. Why repeat an old one?