Chances are you've dealt with a bot in your daily online activities. Depending on your age, it may have been an IRC bot that welcomed you to a chatroom. Or, today, it may be a Twitter bot that follows you to lure you into following its master; it may be an email autoresponder or so-called live chat on a website; it may be the plethora of spam you delete on a daily basis, maybe you received one of the 20 million messages sent by 70,000 "female" bots to trick users of the Ashley Madison website. Or perhaps your interaction with a bot today may be changes to the most recent Wikipedia article you've read.
It's already bad enough that many human editors are bizarrely committed to Wikipedia's destruction, seeking to speedily delete new content, but researchers have found Wikipedia bots are now fighting over what their respective owner believes to be the "right" change to an article.
The paper, Even good bots fight: The case of Wikipedia, analyses the interaction between bots that edit articles on Wikipedia, tracking the extent to which they undid each other's edits over the period 2001-2010. It also models how pairs of bots interact over time, and identify different types of interaction trajectories.
The researchers found Wikipedia bots, while well-intentioned to support the encyclopaedia, often undid each other's work, with some fights continuing for years.
For the most part, Wikipedia bots benefit the system. In 2014, about 15% of edits were performed by bots, identifying and undoing vandalism, enforcing bans, checking spellers, welcoming newcomers and so forth.
Yet, not all bots operate harmoniously. The researchers focused on edits and specifically reversions, where an edit had been undone.
Their exploration uncovered that, on average, bots on the English-language Wikipedia reverted the work of another bot 105 times.
These bot conflicts could extend for years, with the average response time being a month, likely due to the combination of time needed to re-crawl articles, and to Wikipedia's constraints on bot activity frequency.
Ultimately, the research showed that bots, despite their predictability, ultimately interact as unpredictably and as inefficiently as humans.
The researchers suggest even relatively "dumb" bots, therefore, have complex interactions, and there are potential learnings for artificial intelligence, as well as automated social media management, cyber-security and autonomous vehicles.