Twitter has become a go-to news source in times of crisis. However, as we learned once again during the Newtown tragedy, inaccurate or even purposefully deceptive tweets can just as easily spread misinformation.
Conventional wisdom holds that eventually Twitter users themselves will sort fact from rumor. Two problems here. You have to pay attention long enough for this to happen. And you have to hope that falsehoods don’t spread beyond social media, where they are more difficult to correct.
A team of current and former Yahoo researchers is using natural language processing and machine learning to address this problem. When their algorithm is presented with a random false tweet and a random true tweet within the context of the overall flow of information, their software assesses the true tweet as more credible 86 percent of the time. The study is in an upcoming issue of the journal Internet Research. [Carlos Castillo, Marcelo Mendoza and Barbara Poblete, Predicting Information Credibility in Time-Sensitive Social Media]
Not exactly a social media lie detector yet, but if this research pans out, it could do a lot to help Twitter’s credibility in crunch time.
[The above text is a transcript of this podcast.]