On September 8, 2007, a reporter for an obscure news company called Income Securities Advisors typed "bankruptcy 2008" into a Google search box. Google News service instantly alerted the reporter to an article from the a newspaper called the South Florida Sun-Sentinel announcing that UAL, the parent company of United Airlines, had filed for bankruptcy protection. The reporter, who worked for a company that feeds stories to the powerful Bloomberg news service, posted a simple alert with no story or background attached: "United Airlines files for Ch. 11 to cut costs." This alert went out to thousands of influential readers of Bloomberg's financial news network.
There were three big problems with this series of events. First, the Sun-Sentinel archive did not display a date on the story, allowing Google News to scrape it up and put it in a list of recent or current stories. Second, Google's computers placed a new date on the link to the article: September 6, 2008 - the day Google's Web crawling software found and indexed the article. But the UAL bankruptcy filing had occurred in 2002. The company had emerged successfully from protection and reorganization in 2006. And third, the reporter did not check out the veracity of the report. The reporter was apparently unfamiliar with the travails of UAL and of the potential for Google's automatic processes to amplify mistakes that others make in the creation of Web content.
When the Nasdeq market opened the morning of September 8, 2008, UAL stock was trading at $12.17 per share. Once the alert zoomed around Bloomberg at approximately 11 a.m. Eastern Standard Time, panicked sellers unloaded 15 million shares of UAL and shoved the price per share to a meager $3. By 11:16 a.m. Bloomberg had issued an alert denying that UAL had filed for bankruptcy. Once the trading day concluded, and word had spread that the bankruptcy alert was false, the stock recovered a bit. But it still finished down $1.38 from its opening at $10.92 per share. This simple glitch cost UAL shareholders - including most of its employees - 11.2 percent of its market value. In addition, the panic sent shares of two other airlines, Continental Airlines and AMR (parent of American Airlines) down as well. The airlines had done nothing wrong. They had released no bad news. Yet they were all worth less at the end of the day than at the beginning because Google's Web crawlers found a mislabeled story in an open newspaper archive.
This anecdote teaches us some valuable lessons about our alarming dependence on Google. Certainly, had those responsible for posting the Sun-Sentinel article used proper metadata - the elements around a file that tell us its context, such as a date of origin - Google's computers would not have pushed the story in front of the Income Securities Advisors reporter. And had that reporter understood a bit more about the story itself and been a more critical and less credulous reader, no one would have ever heard about the mistake. And if any one in this story understood that aggregators of information like Google News are only as good as their sources, no one would have overreacted. And if Bloomberg were set up to enhance understanding rather than induce gut reactions, someone could have put the breaks on the error. And if traders and investors around the world read more than headlines and tickers before making huge decisions that potentially cost innocent people fortunes and jobs, the errors that preceded the sell-off might not have mattered at all.
But that is not the world we live in. We live in a world flooded with data. We live lives devoted to maximum speed and dexterity, rather than deliberation and wisdom. Many of our systems, not least electronic journalism, are biased toward the new and the now. And even after living intimately with networked computers for almost two decades we lack a widespread understanding of what such systems can and cannot do, or even how they work. We trust them with far too much that is dear to us and fail to master their limits and problems.
Despite all the loud accusations of fault that flew between Google and those responsible for the journalistic errors, it's clear that Google itself did nothing wrong. It's hard to expect that Google's programmers would consider the possibility of the basic metadata error that the Sun-Sentinel made. Nor should we expect them to predict the utter stupidity of the rest of the humans involved in the chain reaction that followed. So the chief lesson here is not that Google is the problem. The lesson is that we are imperfect. When we choose to rely blindly on a pervasive, powerful gatekeeper that we do not understand, we are destined to make monumental mistakes.
RSS Feed
Comments (5)
Siva, Like the article, but you misspelt nasdaq (Nasdeq).
I always wondered who made the money out of this by realizing earlier than others that the story was wrong. How did they come to that conclusion? Were they simply more cool-headed, did they rely on multiple feeds, or were they close to the company so they could knew the story was bogus?
There are a couple of ways to think about this:
This is a story about Google or... this is a story about investing in stocks.
If this is a story about stocks, there are two things you can take home from this:
1. There was no net loss of wealth. Yes, that day, for the investors who bought or sold while the trading range was between the day's low and the day's closing price there was no net destruction of wealth. That is, if you total all the losses and all the gains for all the investors who traded in that range the net loss was zero. Economically, in the aggregate, there was no harm done.
Some like to worry about the losers in this particular transaction. In every stock trade there is a winner and a loser. It is an unavoidable artifact of the system.
2. If this story scared you, then don't invest in stocks. They are really risky. Especially, airline stocks.
"So the chief lesson here is not that Google is the problem. The lesson is that we are imperfect. When we choose to rely blindly on a pervasive, powerful gatekeeper that we do not understand, we are destined to make monumental mistakes."
Which is to say, Google is a tool. What is important is whether we are trained to use these tools properly.
Understanding comes when the education system catches up with our media environment.
Looking forward to the book.
Like the fact that you are focusing on the different actors on the system and how they interconnect and interplay. You might very well be doing this, so this is just a broad-brush comment. What would be useful is to focus on the interconnected web of information systems, what kind of "normal accidents" (Charles Perrow's book -- i absolutely love that phrasing) this system is prone to, and where Google fits in this picture. That way, you are not overly focused / obsessed with Google (even though I understand the book is about Google) and you don't have to then go around explicitly stating where they are or are not a problem. Readers might be able to come to some of their own (and different) conclusions.
"So the chief lesson here is not that Google is the problem. The lesson is that we are imperfect. When we choose to rely blindly on a pervasive, powerful gatekeeper that we do not understand, we are destined to make monumental mistakes."
This paragraph is a bit weak. That we are imperfect tells us nothing new, or at least should not be -- it should be the starting point of the analysis (refer broad comment earlier). You might very well be doing this -- just your phrasing of it reads a little odd.
Look forward to see what you come up with.