Wikipedia Is Edited by Bots. That's a Good Thing.

Bots and anti-bot bots make the edits humans don't want to and keep the site clean

Half of all edits to Wikipedia are made by bots. That stat, from a new study monitoring the site's revisions, might seem like cause for alarm. But according to Wikipedia researchers, it's key to the crowdsourced encyclopedia's success.

For a paper in the physics journal arXiv, Thomas Steiner, a Google engineer and Université Claude Bernard Lyon post-doctoral student, developed an app that monitors the edits on Wikipedia's 287 language sites in real time. Newsweek ran the app for four days and found that, at the time of this post, 46 percents of edits were made by bots.

Bots, which are automated editing scripts running off of Wikipedia users' computers, contribute relatively few edits to the English Wikipedia site. In the period evaluated by Newsweek, they accounted for about 5 percent. Meanwhile, in Spain, that number rose to 14 percent. In Italy it was 53 percent. And in Vietnam, nearly all edits were made by bots.

The reason for the differences across countries, according Wikimedia Foundation researcher Aaron Halfaker, is due in large part to the maturity of the each site. In countries where Wikipedia is newer, and thus the encyclopedia is less fleshed out, there is more work for bots to do.

So what are these bots up to? "They mostly do tedious things that we don't want to do," Steiner told Newsweek via chat. Some bots perform simple tasks like correcting spelling and grammar errors. Others are more specialized, performing "human" tasks like building country pages from census data or using NASA reports to create articles about asteroids.

Indeed, the line between bots and humans often blurs on Wikipedia. Some bots work in tandem with humans to evaluate complicated edits or enforce community rules in Wikipedia--"cyborgs," Halfaker calls them. Over time, says Berkeley computer science graduate student R. Stuart Geiger, programmers have had to make these bots more personal. "Bots that interact with newcomers were very harsh, casting everyone as vandals," he says, and were "turning off a lot of people."

The biggest use of bots, says Steiner, is "detecting vandalism--kids replacing George Washington's name with lower body organs. Stupid things like that."

Vandalism, which has a long history on Wikipedia, may be the bots' greatest success. Halfaker says that a decade ago, a group of Internet trolls began a simple yet obnoxious campaign to deface the site by editing in mentions of a Spongebob Squarepants cartoon character. "They were referencing Squidward all over Wikipedia." In response, Wikipedia users developed a bot that would comb the site's entries, detect the word "Squidward," and undo all the nefarious edits. The fix was a simple one, but would eventually lead to the development of sophisticated anti-vandal bots.

Today, according to Halfaker, the most prolific anti-vandal bot is called ClueBot. It can detect and fix everything from profanity to mashed keys within seconds, he says. It is responsible for almost half of all edits on the English Wikipedia site. "From a computational perspective, its brilliant," says Geiger.

ClueBot also serves to ensure that a relatively small group of vandals can't degrade the site's legitimacy enough that volunteers stop putting in the time and effort to maintain it. "When that bot goes down," Geiger says, "it eerily limits Wikipedians ability to make an encyclopedia."

The fact that humanity needs bots to save the world's knowledge from itself is made only slightly less painful by the fact that these bots, in turn, need humans. As Geiger explains, they are not housed in Wikimedia Foundation servers. Instead, they are developed by everyday users and kept on privately-owned computers. (To run on Wikipedia, they must pass through a Bot Approval Group. If you try to sneak one past the Group it will likely be caught by an "anti-bot bot.")

Geiger remembers one bot he built, called AfDStatBot, that tracked debates between Wikipedia editors. He ran it from his home computer. When he had to move out of his apartment unexpectedly, he found himself without a place to connect the computer to the Internet. "The bot just disappeared because I couldn't take care of it anymore."