In a pair of open letters to Letter to The Honorable Elaine C. Duke, Acting Secretary of Homeland, a coalition of more than 100 tech liberties groups and leading technology experts urged the DHS to abandon its plan to develop a black-box algorithmic system for predicting whether foreigners coming to the USA to visit or live are likely to be positive contributors or risks to the nation.
The letters warn that algorithmic assessment tools will be prone to religious and racial bias, in which programmers get to decide, without evidence, debate or transparency, what kind of person should be an American — which jobs, attitudes, skills and family types are "American" and which ones are "undesirable."
Further, the system for predicting terrorist proclivities will draw from an infinitesimal data-set of known terrorists, whose common characteristics will be impossible to divide between correlative and coincidental.
In the new comments, we urge DHS to stop storing social media surveillance in the A-Files for the following reasons:
Chilled Expression. Activists, artists, and other social media users will feel pressure to censor themselves or even disengage completely from online spaces. Afraid of surveillance, the naturalized and U.S.-born citizens with whom immigrants engage online may also limit their social media presence by sanitizing or deleting their posts.
Privacy of Americans Invaded. DHS's social media surveillance plan, while directed at immigrants, will burden the privacy of naturalized and U.S-born citizens, too. Even after immigrants are naturalized, DHS will preserve their social media data in the A-Files for many years. DHS's sweeping surveillance will also invade the privacy of the many millions of U.S.-born Americans who engage with immigrants on social media.
Creation of Second-Class Citizens. DHS's 100-year retention of naturalized citizens' social media content in A-Files means a life-long invasion of their privacy. Effectively, DHS's policy will relegate over 20 million naturalized U.S. citizens to second-class status.
Unproven Benefits. While DHS claims that collecting social media can help identify security threats, research shows that expressive Internet conduct is an inaccurate predictor of one's propensity for violence. Furthermore, potential bad actors can easily circumvent social media surveillance by deleting their content or altering their online personas. Also, the meaning of social media content is highly idiosyncratic. Posts replete with sarcasm and allusions are especially difficult to decipher. This task is further complicated by the rising use of non-textual information like emojis, GIFs, and "likes."
EFF Urges DHS to Abandon Social Media Surveillance and Automated "Extreme Vetting" of Immigrants
[Aleksander "Sasha" Danielyan/EFF]