Twitter to look at your conversation to identify the trolling instincts

The microblogging site is looking at the behavioural perspective, to improve the health of the conversation

Upping the ante against internet trolls, Twitter is looking at behavioural signals to narrow down on cyber bullies. In its newest safety initiative, the social media website is using machine learning to identify behaviours that distort and detract public conversations on Twitter.

“We’ve designed this to work in any market where we have sufficient use, and therefore signal, to do an effective analysis. Our initial tests were conducted in markets around the world,” a Twitter spokesperson told Business Standard.

India is one of the fastest growing markets for Twitter with the country a major contributor to the double-digit user base growth of the social media website.

Actors, sportsmen, and brands in India find themselves in the line of fire on the social media almost daily with customers clashing over political affiliations, religious beliefs and personal biases.

The most recent victims were e-commerce major Amazon and homegrown cab aggregator Ola.

Amazon was targeted by a set of abusive trolls over the use of actor Swara Bhaskar for its campaign around the Indian Premier League (IPL). After the Kathua rape incident, Bhaskar had been subjected to online anger for highlighting the fact that a child had been raped inside a ‘devi-sthan’ (temple). 

Ola had found itself in the middle of a social media storm when a rider, Abhishek Mishra, who claims to be a Hindutva thinker tweeted that he cancelled an Ola ride as the driver was a Muslim. “I don’t want to give my money to jihadi people,” the tweet read.

Keeping such cyber hounding in mind, Twitter is looking at the behavioural perspective, to improve the health of the conversation.

“Some troll-like behaviour is fun, good and humorous. What we’re talking about today are troll-like behaviours that distort and detract from the public conversation on Twitter, particularly in communal areas like conversations and search,” said a Twitter blog post.

For example, if an account has not confirmed its email address, if the same person signs up for multiple accounts simultaneously, the account can be muted by Twitter. Also accounts that repeatedly tweet and mention accounts that don’t follow them, or behaviour that might indicate a coordinated attack could see the same fate, said the Twitter spokesperson.

The new approach has already seen a positive impact, resulting in a 4 per cent drop in abuse reports from search and 8 per cent fewer abuse reports from conversations.

Twitter has made over 30 changes to its products, policies, and processes in the past 16 months to tackle safety, and as of last month, its has been taking action on 4x the number of abusive accounts globally every day compared to the same time last year.