To improve oversight of facial recognition, expand open-file discovery
Jason Tashea. Photo by Saverio Truglia.
In September 2015, two undercover officers in Jacksonville, Florida, bought $50 worth of crack cocaine from a man who called himself “Midnight.”
During the sale, one of the officers took photos of “Midnight” on an old tracphone, but they did not know the seller’s real identity. As I wrote about in September, during the investigation, the grainy photos—along with the seller’s race, gender and pseudonym—were run through the Face Analysis Comparison Examination & System (FACES), a facial recognition program operated by the Pinellas County, Florida, Sheriff’s Office. Launched in 2001, it’s available to law enforcement across the state and to federal agencies.
Among multiple potential matches, the machine handed back the name and information of Willie Allen Lynch, a local man with a rap sheet.
After being identified by the two officers, Lynch was arrested and charged. With no mention of facial recognition in pretrial documents, it wasn’t until eight days before the trial that Lynch’s attorney learned about the use of facial recognition in the investigation. Even on appeal, Lynch was kept from information about the system and other potential matches the system produced.
Denied the opportunity to challenge the technology used to name him as a suspect, Lynch is serving eight years in prison.
This troubling fact pattern from Florida will continue to rear its head as facial recognition and other novel investigative technologies play a bigger role in the American criminal justice system. To that end, open-file discovery—a hot topic a decade ago—may have renewed purpose that brings needed transparency and accountability to these technologies.
Prosecutors are constitutionally required to hand over all exculpatory evidence to the defense pursuant to Brady v. Maryland. This process relies on the prosecution to determine for themselves what is exculpatory. A clear conflict of interest, this system creates headaches for prosecution and defense attorneys alike. Further, as many scholars, advocates and defense attorneys note, the standard is narrow in what information it covers, especially in practice. Finally, it is rare that a prosecutor is punished for failing to disclose exculpatory evidence—even if the prosecutor knew such evidence existed—as a 2017 New York Times article found.
Realizing the Brady standard is too narrow and hard to apply is nothing new. In 1994, the ABA recommended broader standards when it came to discovery in criminal cases. In the years since, many states have adopted what is called “open-file” discovery for criminal proceedings. Either by court rule or statute, open-file gives the defense expanded access to the prosecution’s casefile, regardless if the information is exculpatory or not. This means fewer back-and-forth discovery motions and appeals are needed. It also satisfies the prosecution’s constitutional requirements under Brady, while requiring less work.
As of 2016, 17 states have open-file discovery in criminal trials, according to a law review article that surveyed state laws. The rest are either closed-file (10 states plus the federal system) or something in between (23 states).
THE REACH OF TECHNOLOGY
While not a silver bullet, expanding open-file to include information about facial recognition systems can improve transparency, while allowing for a zealous defense. Current practice expects that traditional police lineups or photo arrays be shared with the defense, it should be no different when the lineup is done by a computer. Yet, a blind spot in an otherwise expansive discovery regimen allowed it to happen.
To fix this, each state and the federal system should consider open-file statutes or court rules that compel mandatory disclosures pertaining to picture identification broadly—including what system was used, how it was used, the quality of the match, other matches generated in the search, and vitals on the system itself, including any audits—and require police to hand over all investigative materials to prosecutors, as recommended by the Justice Project. (The report, from 2007, did not consider facial recognition technology.) As the organization also recommends, there need to be meaningful penalties if this is not done.
For one thing, the ever-changing pace of technology underscores the importance of expanding open-file discovery. After all, in Florida, which has a open-file discovery, Lynch was unable to access information about the facial recognition system that identified him. Controlled by an expansive statute (Rule 3.220), there are 13 types of material that the prosecution must hand over to the defense within 15 days of request. While covering witnesses, evidence and electronic surveillance, none of the 13 enumerated categories pulled the use of facial recognition into the pool of discoverable open-file material.
Expanding open-file to cover this technology would have helped Lynch learn that he was identified through a facial recognition lineup earlier in his case, as opposed to being surprised a little more than a week before trial. It is unacceptable that he went to trial with his liberty at stake without knowing if there were other possible matches or the relative quality of those matches, also whether his photo had been manipulated by analysts or the vitals of the program itself, including error rates or if the software had been audited.
As investigators adopt facial recognition—often in secret, as recent reporting out of Washington State illustrates—open-file cannot only create a level playing field for defense counsel, but bring a level of transparency and public accountability currently missing in many jurisdictions.
While this would be an improvement over current practice, open-file isn’t a panacea. Even in jurisdictions with open-file, there are missing documents, interviews and other evidence. This may be as much a document management problem as it may be due to an obfuscating prosecutor or difficult vendor.
Further, open-file is also limited by the state’s knowledge. Based on court documents and speaking with people who studied the Lynch case, the police didn’t seem to have much information to share on the FACES system itself—a separate but complimentary concern. To this end, if particular information doesn’t make it into a police report, then it isn’t going to show up in discovery—open-file or not.
Next, the benefits of open-file will not be felt universally, says Donald Jones, a law professor at the University of Miami. Due to structural disadvantages and the limited resources of public defender offices, he’s skeptical that a defender would have the capacity to take advantage of broader discovery rules. This can exacerbate the existing disadvantage in the criminal justice system experienced by poor, black and brown defendants.
Finally, open-file isn’t going to solve the fundamental issues surrounding transparency of a particular device or source code. As has been written in this column before, source code transparency is a necessary component when adopting trustworthy technology in the criminal justice system and protecting due process. However, since the source code is unlikely to be a factor in an investigation, that information will not be revealed through an expanded open-file statute.
Even with these limitations, expanding open-file to include facial recognition will benefit both the state and the defense, while improving the transparency of an opaque process and technology. To do nothing keeps defendants, courts and the public unjustifiably in the dark.
Jason Tashea is the author of the Law Scribbler column and a legal affairs writer for the ABA Journal. Follow him on Twitter @LawScribbler.