‘Suppressed’, says BotSentinel creator by volume of robots in pocket tags



[ad_1]

Also read

“I was amazed at what I was witnessing. When the hashtags and phrases related to Brazil started appearing on our BotSentinel lists, we were all surprised, ”said Bouzy. The tool was developed by him in 2018 and monitors posts and posts created and generated by fake accounts, called robots or bots.

Brazilian users have started to follow the Twitter account, which publishes which labels, Brazilian or not, are growing due to the activity of the robots. BotSentinel also monitors and indicates who are “trollbots”, accounts managed by real people who behave like bots, replicate fake news and generate hashtags.

The volume of labels handled by non-authentic accounts in Brazil surprised Bouzy, who explains that the recent importance that the tool has given to Brazilian policy is not directed, but rather autonomous.

Until now, BotSentinel has focused primarily on hashtags and inauthentic activities related to the United States. We do not control what the algorithm discovers; It is completely autonomous. BotSentinel has taken on a life of its own and goes where the data takes. This is surprising, we do not focus on Brazilian politics. We did not tell Bot Sentinel to look for non-authentic accounts, expanding hashtags and Brazilian phrases; he did it alone, “he says.

“OUT OF MAIA” AND 5 OTHER BOLSONARIST LABELS

pocket labels which have been fed by inauthentic accounts: “data-reactid =” 47 “> Specifically about the hashtag #MaiaTemQueCair, Bouzy stated that” there are at least 300 to 400 inauthentic accounts that amplify “that hashtag. five pocket labels that have been fed by non-authentic accounts:

  • #ClosedWithBolsobaro

  • #ClosedComBolsonaro

  • #FechadoComBolsonaro

  • #ForaMaia

  • #MaiaVaiCair

The increase in these hashtags occurs in a period of extreme political tension around Bolsonaro.

HOW DOES BOTSENTINEL WORK?

Bouzy explains that BotSentinel uses “learning machine” and artificial intelligence to classify accounts as bots or trollbots. The filter used to define what an inauthentic account would configure, according to him, are Twitter’s terms of service.

In 2018, the initial BotSentinel model took 13 months from inception to completion. During the process, 2,500 normal Twitter accounts and 2,500 non-authentic accounts were used that were personally tracked by him.

“Approximately 5 million tweets were used to create the first model. The current model uses 25 million tweets from normal accounts and non-authentic accounts that the platform tracks, “he details. Bouzy guarantees that the current model is approximately 95% accurate. “There can be false positives and false negatives. No model or algorithm is perfect, but we are constantly improving. ”

The trigger for the creation of BotSentinel was the 2016 US election, which resulted in the victory of current President Donald Trump. “I realized that everyone called themselves ‘bots’ and I felt there was a lot of suspicion on social media platforms, especially on Twitter. I felt there had to be a tool to help people distinguish between “real” and “false” “, Bouzy justifies.

Another reason, explains the software engineer, is the migration of the public from news networks to social networks. “Every year, more and more people receive their news from social media platforms than news media, and these inauthentic accounts can distort reality and spread inaccurate information on social media platforms.”

THREAT TO DEMOCRACY

Inauthentic stories are a great threat to democratic political systems for him, masking popularity and influencing opinions based on unproven or wholly false facts.

“They (non-authentic accounts) can help someone unpopular look popular and help someone popular look unpopular. They can influence public opinion and elections, so tools like Bot Sentinel are essential to help people distinguish between reality and fiction, “concludes Bouzy.



[ad_2]