On Wednesday, the American Civil Liberties Union of Massachusetts released a damning report detailing extensive social media surveillance in Boston between 2014 and 2016. Thousands of documents obtained through public records requests reveal that police used advanced social media data mining techniques to monitor what they termed “Islamist Extremist Terminology.” The authorities worked with Geofeedia, a social media monitoring startup that both Facebook and Twitter banned from using their data in December 2016, prompting the ACLU’s original records request.
Using Geofeedia, police are able to filter social media posts based on images, keywords, hashtags, and, crucially, location data—all in real time. The Boston Police Department and its intelligence gathering arm, the Boston Regional Intelligence Center, sourced posts from Facebook, Instagram, YouTube, Flickr and Twitter. In Chicago, police tracked people using the #BlackLivesMatter hashtag during protests, Similarly, Boston authorities tracked people tweeting #MuslimLivesMatter online. This surveillance dragnet caught up everyone using the targeted tags, including emoji, in the area, even if they opted not to disclose their location to Twitter itself.
As documents show, the Boston Regional Intelligence Center created an alert category called “Islamist Extremist Terminology” and set up alerts around terror-related keywords, like “ISIS” and “Islamic state,” and words that may be connected to an attack, including “smoke,”“boom,” and “backpack.” BRIC members received an email with a link to any tweet, Facebook post, YouTube clip or other posting containing the keywords coming from a selected time and location. Other alert categories included “Chapel Hill Shooting,” a reference to the 2015 murder of three Muslim Americans. The BPD searched for people using the #MuslimLivesMatter hashtag after their deaths.
But emails reveal BRIC used Geofeedia for much more expansive surveillance than simply location-based alerts. Not only could BRIC see tweets in real time in a specific location, they could highlight an area and view archived tweets going back “weeks or even months.”
Tellingly, a 2014 email references a Target Tracker capable of “a link analysis to show who the target is affiliated with on all Social Media.” This seemingly refers to associational data. When using social media to keep tabs on someone, police often make note of who they’re friends with online and who’s tagged in their photos. Privacy advocates have long balked at the idea of targeting people for surveillance because of who they’re friends with. Of course, people’s friend circles tend to include others of the same race, income, and education level, making associational data a type of profiling.
Finally, Boston police wanted the ability to search tweets in case of a crisis. To that end, BRIC set up alerts to trigger automatically in areas following attacks. Most interestingly, Boston Police wanted integration with ShotSpotters. ShotSpotters are acoustic devices with GPS capabilities. Whenever they detect gunshots, they indicate to police where shots were heard. The BPD was looking into integrating GeoFeedia and ShotSpotters, so that if gunshots were heard in a neighborhood, police would immediately receive social media updates from people in the surrounding area. BRIC’s director, David Carabin, also wanted 911 integration, so a call for emergency services would trigger a social media sweep as well.
The three-year trial for Geofeedia cost Boston $26,698, but the ACLU found no evidence that the social media mining led to any arrests or prompted further investigations. Facebook, which owns Instagram, and Twitter have promised to cut off Geofeedia’s access, but it’s important to remember that all public data is still fair game for police.