It’s logical to conclude that any technology working to make the job of law enforcement officers easier would be embraced with open arms. In most cases, like streamlined CAD platforms or rugged in-car laptops, this is true – yet there are some solutions that receive a bit more skepticism.
While facial recognition technology (FRT) is an investigative tool that can have a significant impact on criminal investigations, it has been met with some concern from those in the general public. To help reduce some of that crime and dispel myths surrounding the technology, below are three of the most common myths surrounding FRT use and the truth behind this innovative technology.
MYTH NO. 1 – FRT IS USED FOR REAL-TIME SURVEILLANCE
You’d be hard-pressed to find any modern action movie that doesn’t show the use of cameras to track a perpetrator as they attempt to evade capture. While showcasing this fictional technology makes for an exciting storyline, it couldn’t be further from reality.
Many believe facial recognition technology is used in conjunction with fixed surveillance cameras and can track a person’s whereabouts or pinpoint a specific person within a crowd in seconds. In reality, FRT is a stand-alone software platform and is not necessarily used in live applications.
“We don’t do live surveillance,” said Terence Liu, vice president of research at Clearview AI. “We do after-the-crime or after-the-fact investigation and there always is a human in the loop to make any final decisions.”
When an officer uses Clearview AI, they upload an image of a person of interest and the software compares that image to a database of publicly available images from the internet.
“Officers should apply FRT with the same standard as if they were given a photograph of a potential suspect and they were comparing it against other photographs,” said Liu. “From a technology perspective, there are clear checks and bounds with FRT that adhere to either common sense or a legal framework.”
Clearview AI is best described as an investigative platform that allows law enforcement to rapidly generate leads to help identify suspects, witnesses and victims to close cases faster and keep communities safe. It is used within compliance of local, state and federal guidelines.
MYTH NO. 2 – FRT SEARCHES PRIVATE PHOTOS ON PERSONAL DEVICES OR PASSWORD-PROTECTED WEBSITES
Facial recognition technology is only as valuable as the pool of images it has to draw from. With Clearview AI, over 40 billion images are stored in a database that continues to grow. The platform’s web crawlers comb the internet for publicly available images only – that means it limits the data it collects to information that has been made available online to the general public. Clearview AI also requires each of its customers to adhere to all applicable data protection laws in their use of the technology and its data.
Using Clearview AI to search for possible matches during an investigation is akin to any manual search where investigators might turn to your typical internet browser. Instead of spending countless hours attempting to pull up all images of “a Caucasian male between the ages of 30 and 40 with short brown hair,” for example, FRT will scour public images it collected and come up with possible matches in mere seconds.
MYTH NO. 3 – FRT IS INACCURATE AND RACIALLY AND/OR GENDER-BIASED
Perhaps the most prevalent myth surrounding facial recognition technology is two-fold, with many believing the tool exhibits bias while simultaneously producing wildly inaccurate results. Activists have made these beliefs clearly known, resulting in some cities backing down from FRT implementation due to citizen disapproval.
Whenever claims of bias or inaccuracy make news headlines, what’s often not included are the test parameters in which these results were created. Theoretically, if a sample size of images is inappropriate or the accuracy threshold is set incorrectly, anyone could end up with results that seem skewed.
Clearview AI’s algorithm has been tested time and time again by the National Institute of Standards and Technology (NIST) and is consistently highly rated for its accuracy. Created in 1901, NIST is considered one of the country’s first physical science laboratories and conducts a range of biometric testing including over 1100 FRT algorithms (as of March 2024).
While there are a range of tests within certain NIST categories, Clearview AI participates in two of these verifications: the Face Recognition Vendor Test (FRVT) 1:1 and the FRVT 1:N. Both tests are designed to evaluate the accuracy of an FRT algorithm and include diverse demographics, ethnicities and genders within the testing environment.
The NIST FRVT 1:1 test submits two images to an algorithm to discern if the images are of the same person. A wide range of image types are used, from straightforward mugshots to what are called “wild photos,” where faces are photographed in any number of angles and lighting conditions. Clearview AI’s algorithm has regularly achieved greater than 99% accuracy in this test.
A substantially more difficult test, the FRVT 1:N functions similarly to Clearview AI’s regular use, where the algorithm is asked to match a test photo to a database of millions of photos. When using a pool of 12 million photos, Clearview AI has determined a correct match at a rate of 99.85% accuracy.
THE MYTHS ARE BUSTED – NOW WHAT?
With an accurate understanding of how FRT works, law enforcement officers can use this tool to significantly benefit their investigative efforts. Without the help of Clearview AI, an investigation could take countless hours, particularly if leads are difficult to come by.
Facial recognition technology can help improve efficiency in several ways, by both helping to uncover new leads and by offering additional information on already identified persons of interest.
“A lead is only a lead, but if you follow up on that lead you can get confirmation of the fidelity of the search,” said Liu. “FRT is a great augmentation tool to further corroborate some of the leads you gather from elsewhere. It could go both ways – it could be a primary lead that helps you find more leads or it could act as a supportive lead.”
FRT has also been proven to be very useful in instances of wrongful conviction. Nationally, 69% of DNA exonerations – 252 out of 367 cases – have involved eyewitness misidentification, making it the leading contributing cause of these wrongful convictions.
In 2017, Clearview AI was instrumental in exonerating a wrongfully convicted man after a fatal car accident in Florida. Andrew Conlyn, a passenger in the vehicle, was charged with the crime and faced 15 years in prison. However, a good Samaritan was also on the scene, helping to rescue those involved, but officers did not capture this person’s name.
When Conlyn’s attorney could not locate the eyewitness in an effort to prove his client’s innocence, he contacted Clearview AI. After just one search, the platform identified the good Samaritan, who testified in court that Conlyn was not driving the vehicle. Conlyn was exonerated a mere 45 days after his attorney reached out to Clearview AI.
“The obvious effect of Clearview AI’s technology is that ‘wow’ factor,” said Liu. “You put any photo in there, as long as it’s not a very low-quality photo, and it will find matches ranked from most likely to ones that are similar in a short second. It’s amazing to see this technology get dramatically more advanced in the past few years, and how well it works at such a large scale. Clearview AI’s leading effort in applying this technology to the field of law enforcement has helped save lives and exonerate the innocent in a very concrete way.”
Visit Clearview AI for more information.
