From supermarkets to hotels to makeup, there are an increasing number of industries using facial recognition as a way of identifying, targeting, and interacting with their consumer base. Its use for security purposes is obvious, especially in retail stores or high-security buildings, like banks or airports, and there are an increasing number of cities where police forces are trialing different technologies, like Amazon’s Rekognition.

More and more companies, especially in the high-tech sector, are developing new facial recognition tools in response to increasing demand. A graph by CB Insights illustrates the tremendous growth in US facial recognition patent applications in the past decade.

(source: CB Insights)

Despite the obvious benefits facial recognition can provide, like automatically checking a guest into a hotel or seamlessly opening your mobile banking app, lawmakers and regulators alike are calling for increased monitoring of the software and its uses. It isn’t difficult to understand why, as numerous studies have found most facial recognition software tends to demonstrate both a racial and gender bias that could have alarming consequences for law enforcement, as well as within the insurance industry. 

Facial recognition in insurance, the pros and cons

Insurance is among the industries eyeing facial recognition with increasing interest, particularly life insurers. The idea is quite tempting: provide a quote to a potential customer simply by using facial recognition software to analyze a selfie. 

Insurtechs like Lapetus Solutions are already providing this technology to insurers, leveraging its facial recognition technology to analyze selfies for common health problems, including visible damage due to smoking. It’s currently testing its technology, called Chronos, with a number of US insurers, including Legal & General America

Together, this partnership developed a product called SelfQuote, which generates a life insurance quote from a user-submitted photo by estimating the person’s age, gender, and body mass index (BMI). Though the product is no longer online, it’s not far fetched to suppose this kind of technology could be leveraged in the future, especially when combined with the  considerable amount of data technologies like wearables can generate.

Dangers of facial recognition

While the idea of a selfie being enough to get a realistic life insurance quote sounds ideal for both the insurer and customer, the flaws inherent in existing facial recognition technologies make its use questionable. 

Proven race and gender biases are inherent in existing facial recognition, like Amazon’s Rekognition. Studies show the technology has a harder time differentiating features on people of color, women, and gender nonconforming individuals. 

While this is especially problematic in policing, it’s easy to see why this could be an issue for insurers as well. If an insurer is relying on this technology to recognize specific characteristics, and it inherently has serious difficulties doing that with people of color, it’s impossible to guarantee it will function as intended for some customers.  

Facial recognition regulation

The primary concern around facial recognition technology isn’t necessarily how it’s used or works, it’s how it’s regulated (or isn’t). Federal legislation and regulation on the technology is so woefully lacking that even companies like Amazon and Microsoft, who create, develop, and market the technology, have publicly called for increased regulations.

The issue of regulation needs to cover a wide variety of existing and potential issues surrounding facial recognition, beginning with how the technology is developed and trained to recognize people. According to Joy Buolamwini, a Massachusetts Institute of Technology researcher and founder of the Algorithmic Justice League, the overwhelming bias it expresses when identifying white men versus women or people of color comes from the photo data set used to train the technology, in which photos of white men are the majority. 

Despite bi-partisan support for increased regulation on the issue, as well as reports of high ‘false positives’ and racial bias issues, federal action has been slow. This is concerning because facial recognition is already being tested and implemented in many states in everything from law enforcement to street cameras, almost entirely without granted permission from its targets. Some states, like Illinois, and cities, like San Francisco, have a jumpstart on federal intervention and have implemented strict laws of their own.

How ethics play in

Outside of regulatory and federal action, businesses looking to use facial recognition technology in their products should consider the ethical implications of its shortcomings. 

If an insurtech develops facial recognition software that’s prone to ‘false positives’ or trained to favor a race or gender, even if unintentionally, should it still market that technology, like Amazon has? If an insurer knows there’s a chance a technology it’s interested in could result in incorrectly analyzing a selfie and issuing the wrong quote on an insurance product, should it still consider using facial recognition?

While it comes down to the individual company to decide what is and isn’t ethical, the cases against facial recognition are not subtle nor debatable. The use of new technologies is a significant issue in the insurance industry now, especially considering AM Best’s new innovation score, which analyzes the long-term health and strength of an insurer based on how innovative it is. In situations like these, however, the desire to innovate should be tempered with the potential fallbacks of a particular technology. Until regulation and technology can catch up and correct these issues, is it ethical for insurers to consider using facial recognition?

If you have a topic you'd like us to cover, or you'd like to become a contributor, we'd love to hear from you! Please email our editorial team with an overview.