This algorithm is quickly clearing old marijuana convictions in San Francisco

A new machine learning algorithm developed by the nonprofit Code for America can read through charging documents, identify codes for various crimes, and then automatically determine which felony convictions can be downgraded to a misdemeanor (those who also committed violent crimes, for example, can't have their records downgraded). Then the tool automatically fills out required forms that the district attorney can file with the court. For those with a criminal record, the changes could have meaningful impacts on their lives. "If you have a felony conviction, or in many cases, if you have a misdemeanor conviction, there are many employers who will not hire you," says George Gascón, San Francisco district attorney. "There are many landlords that will not allow you to rent or lease a place for them. There are certain types of student loans that you would not qualify because of the felony conviction. So we know that having a felony conviction–for offenses that have been legalized–still holds back a lot of people. Though rates of marijuana use are similar in black and white populations, nationally, someone who is black is nearly four times as likely to be arrested for possession. In California, until marijuana use was legalized, black people were more than twice as likely to be arrested. Clearing records can also improve public safety, Gascón says. "There are two major components to reducing the likelihood that people will commit crimes.