As Kenya prepares for a presidential election next Monday, it’s trying to prevent a recurrence of the last such poll, in December 2007, when more than 1,000 people were killed in postelection violence. Last time, technology helped incite that violence. This time, the hope is that technology will help prevent a similar outburst. Last time around, a text message came on Dec. 31, 2007, four days after a presidential election that many people in the Kalenjin tribe thought was rigged. The text message said that the most powerful Kalenjin figure in the government, William Ruto, was killed. This wasn’t true. But the rumor went viral, from cellphone to cellphone. “That was around in the morning, and by 5, people were moving with their properties, the houses were being torched, and you’re just seeing smoke,” says a man named Alex, who asked that his last name not be used. Alex was in Kenya’s Rift Valley, where gangs of youths with gas canisters and machetes attacked their ethnic rivals. Now Alex is part of a private research project called Umati that scours social media for potentially dangerous speech — speech like that 2007 text message, which he says wasn’t just some falsehood. It was written to incite. “It was hate speech, because whatever was being written there, on the text message, it was for people to react against certain kind of people,” he says.
