杏吧原创

ON 18 July, someone in the Russian government edited Wikipedia. They opened up an article titled 鈥淟ist of aircraft accidents in civil aviation鈥� and scrolled down to the entry about MH17, the Malaysia Airlines flight that had been shot down just the day before. At the time, the article blamed terrorists from Russia. The government official deleted that passage, instead pointing the finger at the Ukrainian military for the accident. How do we know this? A Twitter bot caught him in the act.

Twitter bots are programs designed to spew out tweets according to a particular algorithm. Twitter itself estimates about 1 in 20 of its accounts is not human (see 鈥�Bot or Not?鈥�). Bots are notorious for stirring up trouble 鈥搕hey are often used to try to sway political discussions or stock prices, say. Publications, companies or wannabe celebrities can buy fake followers to make themselves look more popular.

But the talk around Twitter bots has started to shift to a more positive note. While many bots are still mindless puppets for hire, others are being devoted to more serious subjects: spreading information, generating poetry and encouraging transparency. As at the University of Puerto Rico says, we are in the midst of a 鈥淭witter bot boom鈥�.

聯Twitter bots can spread information, even generating poetry and encouraging transparency聰

Bots that track Wikipedia edits coming from government IP addresses, like the one that caught the MH17 change, are relatively new. The first was the UK-focused @parliamentedits, which popped up at the beginning of July. Versions for other countries 鈥� including the , and 鈥� quickly followed. Another, called @valleyedits, tracks Wikipedia activity from major tech companies.

鈥淭here is an incredible yearning in this country and around the world for using technology to provide more transparency about our democracies,鈥� wrote , the software developer behind the US version, in his personal blog. For him, these kinds of bots provide a relatively simple way to encourage the spread of information and debate about issues that he holds dear.

The users of other bots take self-expression one step further. , a digital studies professor at Davidson College in North Carolina, sent out a public call last year for people to use bots more as instruments of protest. Sample himself has built multiple bots to satirise groups he disapproves of. One, , imagines the kind of arbitrary data that the National Security Agency is collecting on Americans: 鈥淛udd Kutch of Port Vernon, California shared a photo named PROTESTANTISM on Instagram鈥�. Another named @NRA_Tally combines reports of fake mass shootings with pro-gun arguments commonly made by the National Rifle Association.

Why do this on Twitter? One reason is that it鈥檚 a social media platform that鈥檚 relatively friendly to programmers. Facebook, by contrast, is more difficult to use. Twitter also lends itself to easy sharing. People are more likely to add accounts that they don鈥檛 know than they might be on other networks. Retweets and favourites can help grab the attention of many human users quickly.

鈥淭he information is delivered to you. It鈥檚 much more proactive rather than reactive,鈥� says Rod Plummer, the managing director of Shoothill, which helps run GaugeMap, a website that tracks rising water levels across the UK. Each individual river and lock has sensors that measure water levels and its own Twitter account that updates periodically throughout the day, so locals can track spots of interest to them.

The idea behind the flood accounts is simple: take information people might want to know and put it out in the open. There are quite a few in this style, from earthquake trackers to one that analyses an American football team鈥檚 strategy. But not all bots are so friendly about it. @needadebitcard finds and retweets pictures that people have posted of their debit cards to its 16,600 followers, shaming those foolish enough to publicise such personal information. Social media limelight can be a bit harsh.

As bots gain in popularity, their owners need to wrestle with new ethical questions. Darius Kazemi, a prolific builder of comedic and literary bots, says he tries to set a good example. He鈥檚 developed some basic bot etiquette: for instance, his bots never mention or tweet at Twitter users who haven鈥檛 already interacted with the account. He also keeps a master list of words his bots are never allowed to use, such as racial slurs.

Your bot should not say anything that you would not say in public, Kazemi says. 鈥淚 want to make sure that as bot-makers we are taking these ethical concerns into consideration. I think it would be very easy not to.鈥�

Bot creators must also decide how to handle any errors their algorithms make. Twitter is notorious for spreading misinformation 鈥� corrections gain far less attention. Say there鈥檚 a bug in the code of one of the government-tracking bots, and it claims that someone made a Wikipedia edit they actually didn鈥檛 make. Who is responsible for the error, and what should happen next? Issues like these will be discussed at Kazemi鈥檚 second annual Bot Summit later this year.

Flores says he hopes that bots will evolve over the coming years. Many creators have their bots focus on only one goal, endlessly performing many permutations of the same trick, but he鈥檇 like to see new levels of complexity. Not that single-serving projects no longer have their appeal.

鈥淭witter folds data back on to itself and it makes people appreciate it with different eyes,鈥� says Flores.

This article appeared in print under the headline 鈥淭witter bots grow up鈥�