Nieman Foundation at Harvard
Twitter bots are not just vectors of hate and misinformation, an army of trolls ready to pounce on any heated issue.
The Hindustan Times, exploring productive ways to automate parts of the reporting process, has tested Twitter bots to deliver in real-time everything from election results and facts about candidates to sports coverage to air quality alerts, hoping to offer readers more immediate public service information and to save reporters time.
“We started this as an experiment to test the waters,” said Piyush Aggarwal, a news apps developer at the Indian English-language news organization Hindustan Times who worked on these projects with developer Abhinash Jha. “What would engagement look like? Would we be able to pull it off? How automated do we want it to be?”
On the simpler end, the Times has set up triggers, for instance, to tweet from its regional accounts whenever the air quality measures in those areas hit a threshold (the data is pulled from an updating map supported in part by sensors Hindustan Times staff planted across the country):
Are you visiting Sector 6 #Panchkula? #AirQuality is getting bad out there. Use our map to check. https://t.co/VuNzTsB4jb
— HT Punjab (@HTPunjab) June 6, 2017
It’s also tested more complex coverage from a newly launched HT Realtime handle, starting first with reporting the five-state assembly elections in India in March, using public data about the 7,000 or so candidates submitted to the Election Commission of India website and made available before election day, as well as the live results feed from the Commission on the day of. When candidates for a seat were within 50 votes of each other, within 500, or if a candidate was winning by a large margin, the account tweeted. But it also tweeted out factoids about winners, including salient information on education level, personal wealth, and when certain candidates had been accused in criminal cases. “All of these insights, it’s very difficult for a human to find in near real-time,” Aggarwal said. (Each winner was only tweeted about once. The factoid categories were ranked; if, for instance, the winner was not the oldest or youngest, the bot would grab the next most interesting bit of information.)
Twitter, Aggarwal said, temporarily blocked the Times’ API because it was tweeting so regularly it was mistaken for spam.
Gadarpur elects Arvind Pandey of @BJP4India. He has been accused in 12 criminal cases – the highest in the state. #UttarakhandElection2017
— HT Realtime (@htrealtime) March 12, 2017
Lambi goes to Parkash Singh. At 89, he was the oldest candidate to contest from the state #PunjabElection2017 https://t.co/jyVNxpZHeG
— HT Realtime (@htrealtime) March 12, 2017
The Times used the same HTrealtime Twitter account to live tweet coverage of the month-and-a-half-long Indian Premier League cricket games that ended late last month, using match data from structured JSON feeds it got from its partners (eight teams, 60 matches, and around 200 players makes for a lot of possible combinations of insights, and a lot of tweets). A script pulled from a spreadsheet containing links to live blog analysis from Times reporters, including them in the automated tweets about the games.
Mumbai Indians win by 1 run. #IPL2017 #RPSvMI https://t.co/lDAXeES6eP
— HT Realtime (@htrealtime) May 21, 2017
“We thought, why not use that structured feed into Twitter commentary, in real time, so that we can also free up the time of the reporters from these more mundane tasks?” Aggarwal said. “We settled on certain use cases for the bot, like which team has won a toss, who were the opening batsmen, when players score 50, 100, 200 runs, match results, who was the man of the match.”
Lendl Simmons & Parthiv Patel have come out to open for @mipaltan against @RPSupergiants. #IPL2017 #RPSvMI https://t.co/lDAXeES6eP
— HT Realtime (@htrealtime) May 21, 2017
For now, all automated live tweets have been funneled into the HTRealtime handle, but down the line the automated coverage will be separated out by topic, with distinct handle for sports, entertainment, politics, and so forth. (The team is also working on a Facebook Messenger bot, though it’s approaching that platform cautiously as users there often expect a smooth, natural conversation with the bot, which is no easy feat.)
“This has so far just been for the elections and cricket. We’re also planning to do this with other sports, really any event for which we’re getting a structured feed — it could be football, tennis, entertainment, music. This is just the beginning,” Aggarwal said.
