Briefing: Lemonade’s phrenology hangover

x-ray

Jen Frost reflects on a staggering self-engineered series of events that saw insurtech Lemonade have to deny using outdated and disproven phrenology in its artificial intelligence claims system.

After feeling forced to deny it uses phrenology in its claims handling systems following public backlash, Lemonade bosses will rue the day that one of its team took to Twitter to attempt to explain how its claims AI works. 

Phrenology is a long-disproved pseudoscience that advocates that a person’s mental abilities can be measured by examining physical traits: in particular the head or skull.  

The idea gained traction into the 19th century, thanks in no small part to the work of physician Franz Joseph Gall, and its impacts could be seen in fields such as early criminology. However, its popularity dwindled and it was for the most part discredited into the early 20th century.  

For the avoidance of doubt: phrenology is a ghastly and discriminatory pseudoscience that has no place in modern medicine, society, or indeed business. It is bizarre that in 2021 a publicly-listed insurance company has worked its way into a position where it even needs to touch on it, let alone deny it is using it. 

Lemonade leapt into damage control mode yesterday, deleting the original series of tweets and equally confusing accompanying graphics (archived screenshots in the gallery below). It is a move that may seem sensible to the insurtech given the allegedly (according to it) misleading nature of the content, but not one that screams ‘transparency’ or ‘trust’, buzzwords that the business has taken great pains to link itself with.  

In a bid to clear things up, Lemonade said yesterday: “AI that uses harmful concepts like phrenology and physiognomy has never, and will never, be used at Lemonade.

“We have never, and will never, let AI auto-reject claims.”

Adding: “We do not believe that it is possible, nor is it ethical (or legal), to deduce anything about a person’s character, quality, or fraudulent intentions based on facial features, accents, emotions, skin-tone, or any other personal attribute.”

It added that its use of video was to make it easier to make claiming easier for the policyholder and to deter fraud. 

“Behavioural economics research, much of what inspired Lemonade’s business model and B-corp status, has shown that we humans are less prone to lying when we’re looking at ourselves speaking in a mirror/selfie camera,” it stated.

So just what went wrong here? 

PR blunder 

The self-styled disruptor has sought to differentiate itself from traditional insurers from the get-go. Publishing regular ‘transparency chronicles’ from its launch and a chatty blog, Lemonade’s comms are a blend of ‘hip’ tech firm, whizzy metrics, and chipper jargon. The insurtech claims to lean heavily on behavioural science.

“Lemonade is rebuilding insurance from the ground up on a digital substrate and an innovative business model. By leveraging technology, data, AI, contemporary design, and behavioural economics, we believe we are making insurance more delightful, more affordable, more precise, and more socially impactful,” it boasted in its 193 page IPO prospectus, a document in which words containing ‘delight’ were used 31 times. 

In seeking to appeal to its socially conscious millennial market of policyholders and investors alike, Lemonade feels like it’s trying to be your mate rather than your insurer. Sometimes it does a great job, as when it faced off with Deutsche Telekom over the colour magenta.  

The problem is, sometimes it feels like your mate is several vodka lemonades in and scrabbling to illustrate the brilliance of a simple point with frantic gestures and drawings scribbled in crayon, perhaps on a booze-soaked tablecloth in some dingy hipster bar. 

The since-disowned tweets stated that Lemonade’s claims handling AI, known as AI Jim, uses “non-verbal cues” to detect fraudulent behaviour. Possibly in a bid to mystify its technology it committed the cardinal sin of failing to mention that red flags are escalated to humans. 

Those of us who enjoy a tipple will be familiar with hangovers. And Lemonade will be feeling one today. Fortunately for the insurtech, its share price has not suffered from the fiasco. However, this is not ideal timing for a phrenology mishap given it is already facing targeted action from activist short sellers.  

Clarity, as many in UKGI will have found themselves hollering as the business interruption test case ticked on into this year, is often the best medicine. Particularly when selling a financial product. Less of the boggling graphics and boastful commentary could go a long way to help people reconnect with the brand and understand what it is really doing. 

AI mistrust 

Besides the obvious PR blunder, the incident has highlighted another key issue and one that runs deep: public mistrust of AI

Lemonade clearly felt it had to reassure concerned parties that it its AI systems are not, in fact, rejecting claims by themselves. 

“The term non-verbal cues was a bad choice of words to describe the facial recognition technology we’re using to flag claims submitted by the same person under different identities. These flagged claims then get reviewed by our human investigators,” Lemonade shared on its blog. 

Adding: “This confusion led to a spread of falsehoods and incorrect assumptions, so we’re writing this to clarify and unequivocally confirm that our users aren’t treated differently based on their appearance, behavior, or any personal/physical characteristic.” 

In a 2019 blog post Lemonade highlighted one case in which AI Jim flagged up a fraudster, described as a “guy” by the insurtech, who donned multiple disguises including a blond wig and a pink dress to make claims. Detecting regular Sherlock Holmes (or perhaps more Moriarty) wannabes looking to bypass its AI claims system is a far cry from the connotations “non-verbal cues” could conjure up and did for many – perhaps hinting that the tech is rather less sophisticated than some of its less savvy investors might hope. 

Recent research by Accenture, which surveyed 47,810 people across 28 global markets in 2020, found that just 7% of consumers would trust a chatbot when making a claim. 

AI offers opportunities to innovate and drive efficiency, but must be properly managed and introduced. Its current abilities and applications are widely misunderstood and mistrusted. People do not want to run the risk of being discriminated against because they do not look or act the same way as someone else’s perceived norm.

Claims 

Putting unlikely phrenology to one side, Lemonade has in part fallen victim to its own ambitions. For years it has pitched itself as David. But as it becomes less David, more Goliath, it looks increasingly like a traditional insurer (whether it wants to or not). It is now publicly listed and targeting rapid growth, and big insurers handle lots of claims. 

There is nothing wrong with that. But there is a fundamental point at issue here that goes beyond rhetoric, beyond technology and beyond social media mishaps. It is a lesson that sensible incumbents have already learnt. Insurance purchasers do not want to hear about rejected claims. 

Gallery: Lemonade’s since deleted thread
  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.postonline.co.uk/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

Best Insurance Employer 2021

Between 15 March and 15 May Post invited insurance employees, through emails, social media and via the brand’s print and online channels, to nominate and rate their employers on a variety of aspects including their management, diversity, work/life balance and whether they would recommend the employer to a friend. More than 2300 responses were received and more than 50 firms were nominated.

You need to sign in to use this feature. If you don’t have an Insurance Post account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: