We live in a world of cyberbullies and trolls. 41% of U.S. adults have personally experienced harassing or abusive behavior online, while another 28% admit to being abusive to a stranger online themselves. Twitter in particular is a brutally toxic environment. Anti-bullying activist Monica Lewinsky wanted to give people a simple tool to fight this barrage of online abuse.

We live in a world of trolls and cyberbullies, and 41% of U.S. adults have personally experienced abusive behavior online. To help fight online abuse, anti-bullying activist Monica Lewinsky created the @GoodnessBot, a bot designed to replace negativity with positivity in that most toxic of online environments: Twitter. The @GoodnessBot gives users a simple, safe and effective way to stand up to their Twitter bullies — or intervene on someone’s behalf.

Using Brandwatch (a social listening platform) to analyze conversations on Twitter, it was discovered that three cyberbullying incidents occur every second on the platform. Most incidents of cyberbullying go unreported, however, so this kind of toxic behavior has been normalized. People see the harmful messages, but do nothing to stop it because that’s” just how it is on Twitter.”

This is dangerous: not only does bullying, hateful language negatively impact the person it’s meant to harm, it also negatively impacts everyone exposed to it, creating feelings of anxiety, fear and even guilt for not interceding on the victim’s behalf.

By asking people to tag the bot @GoodnessBot in real-time bullying situations, the @GoodnessBot directly reached the people who used harmful language and all of their followers who were exposed to it, spreading positivity where it matters most.

When people see a mean tweet, they just have to reply to it with “@GoodnessBot,” and the bot immediately goes to work. The bot has a vast database of both negative/mean words and positive/complimentary ones. When activated, it searches through the tweet for mean words, then swaps them out for positive antonyms, thus inverting the meaning of the original tweet. A tweet like “Fuck you, stupid bitch!” thus becomes “Bless you, savvy mermaid!”

To design the algorithm for the @GoodnessBot, developers analyzed 62 million bullying tweets, and created a list of 2,375 mean words that occur most often. With this algorithm, they identified all the different ways mean words are spelled on Twitter (bitch, b1tch, bi+ch, biatch, bish, etc.) as well as the nuances of context (“you’re a bitch” is harmful but “yasss bitch” is quite the opposite). They then created a database of affirming /complimentary words, creating a positive antonym for every mean word on the list.

To launch the bot, Monica appeared on The Today Show and encouraged people to call on the @GoodnessBot to intervene in bullying situations on Twitter. People immediately started tagging @GoodnessBot to call out bullying. When others saw the bot in action in their newsfeed, they understood the tool and began to tag the bot in their own social conversations. With no media budget, the bot notched 241+ million impressions and attracted followers in 54 countries. It also changed online behavior; in the week after launch people who interacted with the bot increased their use of positive language by 18%.

With no media budget, the bot notched 241+ million impressions. It attracted followers in 54 countries and received an overwhelmingly positive response — particularly from victims of cyberbullying. “I needed to hear this. I’ve been bullied for far too long,” tweeted one user. Even more remarkable, people who interacted with the @GoodnessBot actually changed their online behavior; in the week after launch they increased their use of positive language by 18%.