Mexico has seen an escalation in the usage of bots to suppress twitter trends. Bots in Mexico severely interfere with free speech, online organizing and activism.
Attacks by bots on the hashtag #YaMeCanse were confirmed beginning of December. Mexican tweeps have been altering the hashtag #YaMeCanse to avoid the twitter bots. At the time of publishing this article, they are currently trending #YaMeCanse18. 1 number is added every day or so to keep the tag trending while avoiding the bots.
Alberto Escorcia, a specialist with social networking site Loquesigue.net, described the bot activity as an “attack on freedom of expression,” which users themselves noticed.
This video by Loquesigue.net is an animated visualization of bots attacking the hashtag #YaMeCanse in real time. The bots shut down the hashtag in under 15 seconds, effectively neutralizing the trend. Notice that the human accounts have an organic shape and movement while the bots are clearly automated and rigid.
Journalist Carmen Aristegui interviewed Escorcia regarding his analysis. He said the hashtag protest was overwhelmed by bots, so Twitter removed it from the list of the 10 most talked about topics. He said that a trending topic is created when there is some novelty, “many people write things suddenly very fast,” and that is what Twitter detects: “the speed of tweets”.
The tag #YaMeCansé trended for 27 days at a constant speed; but the bots intervened and lowered the rate of speed, plus they switched the conversation. Escorcia explained:
“A twitter trend is created when there is something new, when a plane crashes, when there is an accident, or when an Attorney says something really bad; suddenly many people write things very fast, the key is the acceleration and speed of the tweets, not the quantity.”
When bots attack, automated accounts enter a trending topic. I’ve counted about 54,000 fake accounts with the monitoring program that I have.”
Escorcia continued: “From what I’ve researched, if an engineer or bot operator can handle around 43 to 45 fake accounts, then these bots lower the speed, or the novelty index. Since Twitter programs its algorithm by measuring the velocity and rate of trends, the bots standardize the speed.
Besides stopping the trend there is another very important issue – disconnecting the conversation. When people interact on a trending topic, networks are created. But with bots, these networks are diluted because the fake accounts create disjointed, meaningless messages, just to rack up numbers. Some have even put pornography into trends, or contradicted things that real humans are saying. Then the bots not only slow the trend, but also decrease the human connections, and people lose interest. It’s really very planned.
For bots to attack a trending topic the size of #YaMeCanse, at least one person is required to operate 50 bots each. So to run 50,000 bots, you would need thousands of people. That doesn’t cost just one hundred thousand pesos, it can cost up to 2 million pesos… up to two million pesos per hour to attack a trending topic like this.”
Video: Demonstration of attacks on hashtags #YaMeCanse and #USTired2:
Hashtag #USTired2 which was created as an Ayotzinapa solidarity protest in the US, was also attacked by bots. You can see the bot attacking #USTired2 visibly in the below screenshot, appearing as a turquoise blue object above the main #YaMeCanse trend. The bots stand out in visualizations as unusual – Escorcia refers to them as “tumors” in the animations. When #USTired2 was under bot attack, #YaMeCanse was also affected peripherally since the tags were tweeted together.
Some researchers call these bot attacks “tecnocensura or tecnofascismo” in Spanish. Bots used in this manner are a weaponized censor. One obvious question we have after witnessing a botnet take down a hashtag in 15 seconds… since this technology exists and works at stopping trending topics, why are ISIS hashtags permitted to trend unimpeded?
We might also ask if any topics have been killed off before they had a chance to trend? If trends are being monitored and a botnet can be activated quickly enough to neutralize a new tag before tweeps notice, it would act as an invisible censor.