While Twitter has been described as a "self-cleaning oven" and a "truth machine," rumors do slip through and cause confusion, if briefly. A new study building on research done in 2010, analyzing tweets surrounding the Chile 8.8 earthquake, is currently researching the possibilities of sussing out the true tweets from the false ones.

The study uses 16 features to identify whether a tweet is credible or not. It seems reliable information tends to be longer and include URLs, as well as come from people with a lot of followers. The true tweets also tend to be negative in nature, and do not contain exclamation points or question marks.

A new paper by those behind the Chile earthquake study, Carlos Castillo, Marcelo Mendoza and Barbara Poblete, will appear in the journal Internet Research next month with what look to be encouraging results. Their algorithm had an AUC of 0.86, meaning that when it's presented with a false tweet and a true tweet, it would label the true tweet more credible 86 percent of the time.

It's not perfect, and actual people would probably fair better in determining which tweet was more credible. Probably. In any case, it's nice to see we're making strides toward making the Internet a more credible place. My only concern is the escalation factor; that is, the people behind the deliberately false information will only adopt better practices to fool such truth-seeking algorithms.