Online news articles can lose their appeal in as little as an hour. That is the message from two statistical physicists who analysed the way people access information on the user-driven news site

Fang Wu and Bernardo Huberman of HP Labs in Palo Alto, California, US, studied Digg in an effort to understand the way online news readers consume stories.

Through a statistical analysis of the site, the researchers discovered that just a handful of stories hog most people's attention and most links seem to lose their appeal in just 69 minutes. Wu and Huberman say the finding could perhaps help website designers find new ways to keep people interested when faced with an avalanche of information.

The team say that Digg provides an ideal "natural laboratory" for observing the collective behaviour of online news readers. This is because the choice of links on Digg, and the prominence these are given, is determined by users.

Digging around

Each link is submitted by a single person and others then vote either for or against it, to determine its prominence on the site. If a story receives enough positive votes (or "diggs"), in a sufficiently short a period of time, it will appear on the front page of the site. Likewise, a story that receives too few "diggs" will be demoted to a less prominent page. This popularity information available for anyone to see.

First of all, the researchers analysed the number of diggs received by the 29,864 most popular stories posted to the site during 2006. They constructed a histogram to represent this data and found that it follows a "log-normal" curve.

This means that the majority of stories receive relatively few votes while just a few get the majority of clicks. The pair say this partly reflects the way Digg operates, since stories that are posted to the front page naturally receive more exposure and, in turn, more positive votes.

Wu and Huberman then tracked the numbers of "diggs" received by 1110 stories, minute-by-minute, in January 2006. They found that digging decays in a "stretched exponential" way and the popularity of a story fades after just 69 minutes. In other words, the rate at which an article attracts votes slows down the number of diggs continues to grow but at a slower and slower rate.

Exponential slowdown

Again, this may be partly due to the way Digg gradually shuffles stories off the main page. But Huberman believes the result can also be interpreted as the collective "attention span" of Digg users. But he thinks the same may be true of those who read other online news sites. "The stretched exponential slowdown [or long tail] we observed is, in my view, universal and does not depend on the nature of the channel creating the news," he claims.

The finding could perhaps point to better ways of or varying content on a website, so as to maximise exposure to readers. "It might mean prioritising what goes where on a webpage," Huberman adds.

Clay Shirky, who researchers social software and peer-to-peer technologies at New York University, US, says the results are no great surprise. "It's how everything, from websites to movies to books gets hot," he told New Scientist.

He says this, and similar research, could help those designing websites that rely on social interactions between users. "For social site designers, the choice will be whether to try to heighten or dampen the effects," he says.