Colin Crowell, Global Vice President, Public Policy at Twitter talks to The Hindu on dealing with the challenge of bot accounts and hate speech, and balancing it with freedom of expression.
Election in India is a big focus for Twitter. What kind of interaction do you see happening between Twitter and the political parties here?
For starters, we will try to verify the accounts of major candidates. We will educate them on our rules and tools to make sure political parties are aware of them. But we are also in headlong pursuit of making sure that we can be focussed on problems that may arise during the elections, and one of the chief factors that we look at in the electoral context is disinformation campaigns. We are working hard on malicious automated accounts. We are quite focussed on that.
How will you negotiate between freedom of expression and hate speech?
That is always a fine line that we look at. The reality is that if you hold yourself as a platform that celebrates free expression, you are not able to celebrate it and safeguard it, if people are being bullied into silence. We do have to take action on certain accounts to ensure that people can participate and are not bullied into silence or walk away from the platform.
How comfortable are you, as a company, in this role as arbitrator?
I would put it in two categories. The first are the Twitter rules for impersonation, threats of violence, hateful conduct, serial abuse and harassment. We want to articulate our rules, be crystal clear what they are. We do believe that is our role.
On the other hand there is a broader question in the area of disinformation and so-called fake news, where we do not believe that it is our role to determine what is true and what is false. That is the role of journalists and news media organisations, to hold government and public officials accountable on behalf of the broader public, and it isn’t the appropriate role of a corporate entity to decide what’s true and what’s false, particularly in the context of elections. Sometimes what people consider fake news may be hyper partisan rhetoric.
The safer ground for us is malicious automation and how we attack it. We don’t look at the content of the tweets, what we do is look at signals of those accounts. Our investments in machine learning will aid in finding those patterns and signals.
If you look at what we do, in the context of terrorism, we have removed roughly 9,50,000 accounts related to terrorism in the last three years. Three years ago, when we were removing these accounts, about 25% of them were surfaced to us using technology. In the last six months, that number has gone up to 95% of the total accounts. Amazingly, 75% of these 95% accounts were eliminated before they did their first tweet.
But dealing with hate content will be different from terrorism.
It is different, but the use of technology to solve some of the problems that technology itself poses is something we are investing in and it is showing results. In the last year, for example, we have made some 25 changes in rules and in product launches. In the 14 months since we have done that, we are actually actioning 10 times more accounts than before. Today, for example, we are thwarting 5,30,000 suspicious logins a day, where it looks suspicious to us and before they start, we are challenging them to validate certain aspects of their conduct so we know that they are legitimate.
We challenge 6.4 million accounts a week now, double of what it was two years ago, by leveraging technology. Some things of course do remain challenging, like abuse is different from terrorism and state-sponsored interference campaigns and misinformation provides a degree of difficulty since a sovereign state’s intelligence agency will be much more sophisticated in utilisation of VPNs and data centres in different places to mask their origins. But we are trying to work with the challenge.
How does Twitter deal with government requests for blocking?
We see there are rules that are particular to a country at many places. So this is not unique to India. There are things that from the Indian government’s perspective they would like removed pursuant to the Indian law, so we take those requests.
However, one of the things that Twitter tries to do is be as transparent as possible about these requests. When we receive such requests, we upload the government order to the Lumendatabase.org website. Also, we don’t make accounts and tweets disappear because then the conspiracy theories abound. Instead we put a grey box where the tweet or the account used to be, which states that the content has been removed in India pursuant to the Indian law.
Is there any data on bot accounts from India?
We have not broken down bots by country. A lot of the bot networks use VPNs. Actually one of the signals that we look for is that are they coming from VPN.
Often, people point to an account and say that is a bot, but it’s not. If there is a well organised political party that emails out to its supporters that this is the message of the day. It may look like a bot network, but it’s not. They are really human beings. They are political partisans. It is not against the law or Twitter rules. That is part of the milieu that we now occupy.
