How Are Police Using Facial Recognition Software? And Is It Accurate?
Law enforcement is always looking for better tools for analyzing criminal evidence in the hopes of achieving more accurate convictions. And, as technology advances, there are more tools at the police's disposal. One of the newest is facial recognition software and facial examination experts.
The new software has already been used to catch, and convict, alleged criminals. But how comprehensive is the technology, and how accurate are its results?
In 2014, the FBI launched the billion-dollar Next Generation Identification system, claiming at the time to house 51 million photographs and would grow exponentially over the coming years. And the NGI system has already been credited with tracking down criminals, including an alleged pedophile who had been on the run for 19 years. (The city of Chicago also installed its own facial recognition technology, which it claims helped them nab a stickup artist.)
But considering the FBI has admitted to faulty forensic evidence testimony when it comes to hair and bite mark analysis, it's fair to question whether the new facial recognition technology is any more accurate.
Forensic, or Fake?
In 2009, the National Academy of Sciences (NAS) issued a blistering critique of forensic science in which it concluded that, outside of DNA analysis, no other forensic method could reliably and consistently "demonstrate a connection between evidence and a specific individual or source." And a Virginia Law Review article from the same year showed how junk forensic science can lead to faulty convictions: "[I]n the bulk of these trials of innocent defendants -- 82 cases or 60 percent -- forensic analysts called by the prosecution provided invalid testimony at trial -- that is, testimony with conclusions misstating empirical data or wholly unsupported by empirical data."
And The Intercept last week related the story of Steve Talley, a Denver-area financial adviser who was twice arrested for bank robbery. Both times his arrest was based on facial recognition analysis. And both times police got the wrong man. Fortunately for Talley, he had other evidence --phone records and eyewitnesses -- to clear him of the charges. While it seems clear that judges, juries, prosecutors, and defense attorneys should approach facial recognition technology with the same skepticism of other failed forensic sciences, other defendants may not find themselves so lucky when faced with a similar analysis.