On 21st February 2018, Medianama held a discussion in Delhi, on Fake News and Online Content Regulation, with support from Facebook. The following are notes from a talk on ‘Bots and misinformation’ by Mahima Kaul, the Public Policy Head of Twitter India.
Twitter in many ways has been in the eye of the storm when it comes to fake news. The platform has come out an admitted that bots using it may have played a role in influencing the 2016 US presidential elections. The company has promised and implemented various measures it feels will help fight misinformation on its platform.
Mahima Kaul, the Public Policy Head of Twitter speaking at the NAMApolicy event elaborated, “Twitter looks at itself as a platform that’s meant to keep people informed about what is happening in the world and what are other people doing. You have access now to a lot of people you didn’t maybe ten years ago. We also allow pseudonyms and anonymity to protect that space for people who might want to speak about issues but may not want to use their real names.”
Kaul continued, “We do believe that our platform on a certain level does also if used correctly, provide the antidote to a lot of the misinformation and fake news that goes around. This is because we have people like journalists on our platform who are able to now do their job with a lot more speed and accuracy. Fact checkers on the platform are able to counter a lot of this fake news narrative that goes around.”
To highlight the action that Twitter has taken to deal with the issue of misinformation Kaul shared some key stats. “We catch 3.2 million suspicious accounts globally per week, that’s double the amount were able to capture a year ago. We don’t add these numbers to our monthly active users.”
But removing bad players also involves catching bots at the point of creation, Kaul said “The idea is you stop them at the point of creation so they don’t get to even post their first tweet. We catch up to 523,000 suspicious logins in a day. A lot of this is being done through machine learning and automatic spam and bot detection tools, the details of some of which we don’t give out because we don’t want the bad actors to understand how we do it. We are finding that we are improving on this activity at 64% increase year-on-year.”
One of Twitter’s biggest issue is the manipulation of trends and talking point by individuals or small groups running multiple accounts, “We are also trying to detect cluster accounts, a single suspicious entity who may have created multiple accounts to amplify content on the platform. We have used our techniques and have actually managed to catch 5.7 million accounts that way,” Kaul said.
Detecting non-human activity patterns is also important in the process sometimes genuine accounts may get affected, Kaul added, “Sometimes due to false positives, people can get caught up in these filters that we are putting out to catch bot-like behaviour. We generally put these people back in the system after asking them to authenticate themselves by using a captcha or verify a phone number.”
Dealing with abuse
Kaul highlighted what Twitter has been doing to deal with abuse on its platform, “Over the last two years, we have a lot of work done around abuse, something that was becoming and still is a big problem. We looked at that from two points view. One was we have to weigh our policies according to when we made them last and what they are today and see if abuse patterns have evolved and that our policies need to be updated. For example, we had a policy on terrorism, specifically promoting terrorism and incitement. In the last few weeks, we have actually updated the policy to talk about violent groups that may not be terrorist groups.”
“On the other end, we have different tools given to the user. One is you can report a lot more easily now, we ask you for context because with speech context is important. But at the same time from our end enforcement is something you may or may not see. Sometimes we lock a person out of their account, other users may not know and believe that no action has been taken against that person. But if that person has been quite for 12 hour or 24 hours that should give you an indication that the person is probably locked out of their account. It’s like a timeout and we hope that they won’t repeat this behaviour.” Kaul continued, ” Sometimes we tell certain people which tweets to delete.”
Kaul mentioned that a lot of these actions escape the eyes of researchers as they may not show up on the open APIs that Twitter shares with them, “That’s why it’s really important sometimes for researchers to come to us. My colleagues in the UK and US are able to work with academic institutions on this.”
But one doesn’t need bots to amplify bad activities, not when human labour can be cheaper and harder to detect, Kaul elaborated, “Another pattern which is emerging especially in India is human-directed behaviour. Here in India, it is easy to subsidise work by using people instead of machines or bots for bad behaviour. We are trying to find those clusters of bad behaviour, but this a growing problem in some markets, not just India but countries like Mexico also.”
On fake news
Nikhil Pahwa from Medianama asked, “What about verified accounts spreading misinformation on Twitter?” Kaul responded, “We don’t have a fake news policy currently. It is something we need to give some thought over. I don’t know if you want Twitter to be the arbiter of whether this is true and this is not. So how do we deal with this is something that we are still wondering about. What we can do is that we can elevate good content, accounts that are doing good work. Our discover bar now is now more than just trending hashtags.”
Meghnad Sahasrabhojanee, who hosts the ‘Consti-tution’ web series asked, “What about copy-paste tweets and manufactured trends?” Kaul agreed that this was an issue, “We definitely don’t want trends to be manipulated. Firstly it’s not a great user experience if you come and see the top two-three trends and they are not genuine trends as a user you are not going to enjoy the platform. Secondly, it’s gaming our system and we don’t appreciate that at all and we are dealing with it.”