The litigation over Facebook’s Tag Suggestions feature in the United State District Court for the Northern District of California continues, with the court this week denying both sides’ bids for summary judgment in a ten-page order. The case, formerly captioned Patel v. Facebook and now going by the name of In re Facebook Biometric Information Privacy Litigation, is on course to proceed to trial in July.
The order contains a few items of interest, both legal and technological. The court re-emphasized its earlier conclusion that a violation of the notice and consent provisions of the Illinois Biometric Information Privacy Act (BIPA), without any additional harm, could constitute an actual injury sufficient to support a lawsuit. We noted last month that the court’s conclusion may be in tension with recent Ninth Circuit precedent. The court, however, did not appreciate Facebook’s attempts to re-litigate the issue, describing a “troubling theme in Facebook’s briefs” of “reverting to the faulty proposition” that a violation of BIPA’s privacy provisions was not by itself an actual injury.
The court then turned to the “voluminous” evidence that each side submitted in seeking summary judgment. Instead of discussing each issue the parties had raised, the court cited just one example of the factual disputes precluding the entry of summary judgment. That example raises interesting questions—bordering on the epistemological at times—of how biometric technology works.
The crucial factual question is what the Tag Suggestions feature actually does to identify and distinguishes users in uploaded photos. If Tag Suggestions engages in a “scan” of “facial geometry,” then under BIPA it collects biometric identifiers and BIPA’s notice and consent requirements would apply.
The parties agree on some things, namely that Tag Suggestions scans photographs of faces, detects the location of certain points—the center of eyes, the tip of the nose, and the edges of the mouth—and calculates a “face signature” from those points. The plaintiffs argue that by locating those points, and progressively scanning the facial features surrounding them, the algorithm can reconstruct an entire facial image. Facebook disagrees, arguing that Tag Suggestions merely analyzes pixels with certain characteristics to create a unique template—something that the algorithm could do just as easily with an image that was not a face.
Where is the line between analyzing and distinguishing between arrangements of features on a face on one hand, and recognizing and remembering a face on the other? A difficult question, especially since, as Facebook puts it, the algorithm “learns for itself what distinguishes different faces and then improves itself based on its successes and failures, using unknown criteria that have yielded successful outputs in the past.” Figuring out exactly how this opaque process works will make for an interesting trial.
That trial is scheduled for July 9. Two pending docket items bear watching between now and then. On May 7, Facebook moved to stay the district court litigation while it pursues an interlocutory appeal on class certification. Briefing on that motion is set to be complete on May 29, and a hearing is scheduled for June 14. The ruling on that motion, and the results of the class certification appeal, could delay or significantly change the nature of the trial.
If the trial does go forward as scheduled, though, the public may have a unique opportunity to learn about the technology at issue here. Under the “Cameras in the Courtroom Pilot Project,” the trial may be recorded and the recordings published online. The court has asked the parties to make any objections to recording the trial by May 22.
We’ll be monitoring—or perhaps we should say “scanning” the docket for any updates.