Dear Office of Management and Budget:
The Center for Democracy & Technology (CDT) appreciates the opportunity to comment on the Department of State’s (DOS) proposals to expand information collected from 14.7 million immigrant and nonimmigrant visa applicants to include: social media identifiers used during the past five years; as well as telephone numbers and email addresses used in the past five years, and international travel over the past five years. CDT has consistently opposed DOS’s past social media collection proposals for many reasons including the fact that such collection would chill free speech and association and inhibit the right to anonymity, in return for little security benefit. Indeed, we raised these concerns with DOS with respect to this proposed collection back in May.3 Our comment in full to DOS is included as an attachment to this letter. Thus far, DOS has not been responsive to these criticisms, including concerns that they substantially underestimate the expense of such a program. Social media identifiers will yield messy and multidimensional data sets, from which meaning will be difficult if not impossible to parse. Given that the average internet user has seven social media profiles, this proposal would introduce significant noise and little if any discernable signal to the visa screening process. We urge the Office of Management and Budget (OMB), in its oversight role, to fully consider these concerns.
We specifically wish to highlight two issues which we discuss in greater detail in the attached comment:
1) Social media screening is unlikely to yield relevant security information.
- Research demonstrates that expressive conduct like social media activity is not a valid predictor of one’s propensity to commit an act of violence.
- DOS has not demonstrated that social media data will provide it with the ability to better enforce the Immigration and Nationality Act, or that its vetting would be hampered without this data.
- There is no evidence to suggest that DOS has developed a strategy for evaluating social media data.
- Social media screening is comically easy for bad actors to circumvent.
2) Social media content is not easily interpreted and will likely lead to the use of problematic algorithmic screening.
- Social media communications are idiosyncratic and not easily decipherable, particularly those that employ slang, sarcasm, or non-textual information like emojis, GIFs and “likes.” Undoubtedly visa applicants will also use foreign languages.
- DOS is requesting more data than human analysts could possibly review.
- In order to analyze the data collected, DOS may turn to automated tools to assess social media data. However, these tools are would be inherently technologically deficient and prone to discrimination. As CDT has explained in a recent white paper, automated tools for analyzing the text of social media posts cannot reliably interpret the meaning of a post or the speaker’s intent.