Features

net.wars: Slow news

It took a confluence of several different factors for a six-year-old news story to knock 75 percent off the price of United Airlines shares in under an hour earlier this week.

The story said that United Airlines was filing for bankruptcy, and of course was true – in 2002. Several media owners are still squabbling about whose fault it was. Trading was halted after that first hour by the systems put in place after the 1987 crash, but even so the company's shares closed 10 percent down on the day.

Long-term it shouldn't matter in this case, but given a little more organisation and professionalism that sort of drop provides plenty of opportunities for securities fraud. The factor the companies involved can't sue: human psychology. Any time you encounter a story online you make a quick assessment of its credibility by considering:1) the source; 2) its likelihood; 3) how many other outlets are saying the same thing.

The paranormal investigator and magician James Randi likes to sum this up by saying that if you claimed you had a horse in your back yard he might want a neighbour's confirmation for proof, but if you said you had a unicorn in your back yard he'd also want video footage, samples of the horn, close-up photographs, and so on.

The more extraordinary the claim, the more extraordinary the necessary proof. The converse is also true: the less extraordinary the claim and the better the source, the more likely we are to take the story on faith and not bother to check.

Like a lot of other people, I saw the United story on Google News on Monday. There's nothing particularly shocking these days about an airline filing for bankruptcy protection, so the reaction was limited to "What? Again? I thought they were doing better now" and a glance underneath the headline to check the source. Bloomberg.

Must be true. Back to reading about the final in prospect between Andy Murray and Roger Federer at the US Open. That was a perfectly fine approach in the days when all content was screened by humans and media were slow to publish.

Even then there were mistakes, like the famous 1993 incident when a shift worker at Sky News saw an internal rehearsal for the Queen Mother's death on a monitor and mentioned it on the phone to his mother in Australia, who in turn passed it on to the media, which took it up and ran with it.

But now in the time that thought process takes daytraders have clicked in and out of positions and automated media systems have begun republishing the story. It was the interaction of several independently owned automated systems made what ought to have been a small mistake into one that hit a real company's real financial standing – with that effect, too, compounded by automated systems.

Logically, we should expect to see many more such incidents, because all over the Web 2.0 we are building systems that talk to each other without human intervention or oversight.

A lot of the Net's display choices are based on automated popularity contests: on-the-fly generated lists of the current top ten most viewed stories, Amazon book rankings, Google's page rank algorithm that bumps to the top sites with the most inbound links for a given set of search terms. That's no different from other media: Jacqueline Kennedy and Princess Diana were beloved of magazine covers for the most obvious sale-boosting reasons.

What's different is that on the Net these measurements are made and acted upon instantaneously, and sometimes from very small samples, which is why in a very slow news hour on a small site a single click on a 2002 story seems to have bumped it up to the top, where Google spotted it and automatically inserted it into its feed. The big issue, really – leaving aside the squabble between the Tribune and Google over whether Google should have been crawling its site at all – is the lack of reliable dates.

It's always a wonder to me how many Web sites fail to anchor their information in time: the date a story is posted or a page is last updated should always be present. (I long, in fact, for a browser feature that would display at the top of a page the last date a page's main content was modified.) Because there's another phenomenon that's insufficiently remarked upon: on the Internet, nothing ever fully dies. Every hour someone discovers an old piece of information for the first time and thinks it's new

Most of the time, it doesn't matter: Dave Barry's exploding whale is hilariously entertaining no matter how many times you've read it or seen the TV clip. But Web 2.0 will make new money for endless recycling part of our infrastructure rather than a rare occurrence.

In 1998 I wrote that crude hacker defacement of Web sites was nothing to worry about compared to the prospect of the subtle poisoning of the world's information supply that might become possible as hackers became more sophisticated.

This danger is still with us, and the only remedy is to do what journalists used to be paid to do: check your facts. Twice. How do we automate that?