Skip Navigation

The Government Is Expanding Its Social Media Surveillance Capabilities

But social media monitoring programs and the algorithms that power them aren’t effective — and may be discriminatory.

May 22, 2019

Federal govern­ment agen­cies such as the Depart­ment of Home­land Secur­ity (DHS) have dramat­ic­ally expan­ded their social media monit­or­ing programs in recent years, collect­ing a vast amount of user inform­a­tion in the process — includ­ing polit­ical and reli­gious views, data about phys­ical and mental health, and the iden­tity of family and friends. DHS increas­ingly uses this inform­a­tion for vetting and analysis, includ­ing for indi­vidu­als seek­ing to enter the United States and for both U.S. and inter­na­tional trav­el­ers.

But while the govern­ment has justi­fied its expan­sion in the name of national secur­ity, there is little indic­a­tion that social media monit­or­ing programs — or the algorithms that some­times power them — are effect­ive in achiev­ing their stated goals. Addi­tion­ally, there is evid­ence that DHS is using personal inform­a­tion extrac­ted from social media posts to target protest­ors and reli­gious and ethnic minor­it­ies for increased vetting and surveil­lance. In a new report, Social Media Monit­or­ing, the Bren­nan Center provides an over­view of DHS social media monit­or­ing programs and the new set of chal­lenges that they are surfa­cing.

The effect­ive­ness of social media monit­or­ing programs is unproven

The recent growth of DHS surveil­lance systems, includ­ing its social media monit­or­ing programs, has been rapid. In Social Media Monit­or­ing, the Bren­nan Center outlines how social media monit­or­ing is used across vari­ous arms of DHS, includ­ing Customs and Border Protec­tion (CBP), the Trans­port­a­tion Secur­ity Admin­is­tra­tion (TSA), U.S. Immig­ra­tion and Customs Enforce­ment (ICE), and U.S. Citizen­ship and Immig­ra­tion Services (USCIS).

DHS social media monit­or­ing programs have expan­ded in paral­lel with the prolif­er­a­tion of social media inform­a­tion and, consequen­tially, the grow­ing number of compan­ies creat­ing products that claim to inter­pret that inform­a­tion. “It’s a coup­ling of the explo­sion of inform­a­tion on social media and the emer­gence of algorithmic tools that purport to be able to analyze it and come up with mean­ing­ful results,” said Faiza Patel, co-director of the Bren­nan Center’s Liberty and National Secur­ity Program.

But despite their expan­sion, the DHS programs have not proven success­ful, even based on the depart­ment’s own meas­ures. For example, after USCIS piloted five social media monit­or­ing programs in 2016, the agency’s own eval­u­ations found the programs largely inef­fect­ive in identi­fy­ing threats to public safety or national secur­ity. Indeed, for three out of the four programs used to vet refugees, “the inform­a­tion in the accounts did not yield clear, artic­ul­able links to national secur­ity concerns, even for those applic­ants who were found to pose a poten­tial national secur­ity threat based on other secur­ity screen­ing result,” accord­ing to a DHS brief.

These DHS pilot programs and their subsequent eval­u­ations high­light several of the cent­ral chal­lenges asso­ci­ated with social media monit­or­ing. One major chal­lenge is the diffi­culty of actu­ally inter­pret­ing what’s in the social media messages and connect­ing them to actual threats. These inter­pret­a­tion prob­lems become even more complex when a non-English language or unfa­mil­iar cultural context is involved. The programs them­selves also carry civil liber­ties risks. “They give the govern­ment a pool of inform­a­tion about people’s personal lives and polit­ical and reli­gious beliefs that can easily be abused. And research shows that people censor them­selves when they know the govern­ment is watch­ing,” said Rachel Levin­son-Wald­man, senior coun­sel in the Bren­nan Center’s Liberty and National Secur­ity Program.

Social media monit­or­ing algorithms are unre­li­able — and could be discrim­in­at­ory

There is yet another key complic­at­ing factor in social media monit­or­ing programs — the increas­ing use of algorithmic tools to review social media posts. These tools and meth­ods, which include natural language processing and algorithmic tone and senti­ment analysis, have high error rates. This makes it ques­tion­able that they are actu­ally capable of achiev­ing DHS object­ives, partic­u­larly because of the open-ended nature of the eval­u­ations they are used for, such as identi­fy­ing national secur­ity threats.

Equally troub­ling, the algorithms that are deployed for social media monit­or­ing are suscept­ible to bias. “Our exper­i­ence with algorithmic tools shows that they tend to oper­ate in a discrim­in­at­ory fash­ion,” said Patel. “They make judg­ments based on prox­ies, and when these prox­ies reflect biases, the results produced by an algorithm simply repro­duce those biases. For example, the biases evid­ent in the early versions of the Trump admin­is­tra­tion’s Muslim ban could be coded into an algorithm, result­ing in the flag­ging of many Muslims as a national secur­ity threat.” Since even before the ban, federal agen­cies such as the FBI and the Depart­ment of Defense have used reli­gious beliefs as mark­ers of danger­ous­ness.

The hard ques­tions that DHS needs to consider

One barrier to address­ing DHS’s expan­sion of its social media monit­or­ing programs is the lack of visib­il­ity into the full scope of the depart­ment’s surveil­lance capab­il­it­ies, a gap this report seeks to address. In addi­tion, there is currently minimal over­sight of these programs, includ­ing from Congress.

“Congress should look closely at these DHS programs and ask the basic ques­tions,” said Patel. “In what contexts is the Depart­ment monit­or­ing social media? How is it veri­fy­ing the accur­acy of accounts being attrib­uted  to indi­vidu­als? What kinds of decisions is it using this data for? How is the inform­a­tion being shared? And how is the effect­ive­ness of these programs being meas­ured?”

Read the full Bren­nan Center report, Social Media Monit­or­ing.