Abstract: Over the last decade, many police departments have adopted (and some have abandoned) a suite of algorithmic tools designed to help law enforcement proactively prevent crime. This talk evaluates whether location-based tools (e.g as PredPol/SoundThinking) can be used in a way that is adequately fair. I argue that because they concentrate the risks posed by police activity on specific geographic areas, whether we should use these tools depends not only on whether they can effectively reduce the observed crime rate, but importantly also on whether they can promise greater security for the residents of the most at-risk neighborhoods.