Project typeResearch writing
Writing sample, technology policy

Facial Recognition

I started with something familiar, unlocking a phone with your face, and asked what changes when police use the same technology to identify a stranger. In this 2019 essay, I argue for limits on how law enforcement uses facial recognition.

SkillsResearch, synthesis, argumentation
TimelineSep 2019
Methods & toolsLiterature review, source analysis

What happens when the match is wrong?

If your phone gets it wrong, you can try again or enter a passcode. In a police search, a wrong match can draw attention to someone who had nothing to do with the case.

The Congress test

In 2018, the ACLU used Amazon Rekognition's default settings to compare photos of members of Congress with 25,000 public arrest photos. The software returned false matches for 28 members of Congress, with lawmakers of color disproportionately represented among those matches. Those results describe that specific test. Read the ACLU's test report.

Read the full paper

Loading the writing sample…

Where I drew the line

A face match can help narrow a search, such as when investigators are looking for a missing person. The paper asks how much weight that match should carry and what the public should know about the search.

Check the match

A match can be a lead. Confirming someone's identity calls for other evidence, especially when errors fall unevenly across groups.

Set rules for use

Even an accurate system raises questions about who can search people's faces, how long agencies can keep that data, and what they disclose about its use.