Lemonade’s Tweet About AI-Driven Fraud Detection Sparks Public Relations Crisis

By Jim Sams | May 28, 2021

Lemonade, a five-year-old insurer that went public last year, generated an online uproar with a Twitter post that boasted about its AI-based fraud-detection skills, forcing the carrier to walk back its comments.

Lemonade on Wednesday deleted a tweet that said its artificial intelligence program “can pick up non-verbal clues” from video recordings of policyholders who use its app to report claims. That was after Vice, an online news site, posted a story quoting AI experts who said “emotion-recognition systems” that claim to detect a person’s mood or mental state are unreliable and inherently biased.

Chief Financial Officer Tim Bixby told the Claims Journal on Thursday that his company did not intend to lead people to believe it was using its customers’ facial expressions to detect fraud.

Tim Bixby

“The language that suggested we use non-verbal clues in hindsight we found inelegant, not a good word choice,” he said during a telephone interview.

Bixby said Lemonade is well aware that the reliability of such technology, known as physiognomy, has been widely debunked.

“Those types of interpretations have been shown to be easily flawed and not to be relied upon,” he said.

The firestorm started Monday when Lemonade posted several tweets to explain how it uses technology to control loss costs and better serve customers. It said that when users file a claim, they record a video on their phone to explain what happened.

“Our AI carefully analyzes these videos for signs of fraud,” Lemonade said. “It can pick up non-verbal clues that traditional insurers can’t, since they don’t use a digital claims process.”

Critics howled.

“As an educator who collects examples of AI snake oil to alert students to all the harmful tech that’s out there, I thank you for your outstanding service,” tweeted Arvind Narayanan, a computer scientist and an associate professor at Princeton University, according to Vice.

After deleting its posts and the ensuing comments from Twitter, Lemonade tweeted that “the awful thread” caused more confusion than anything else. AI is never used to “auto-deny” claims, the carrier said.

“Our systems don’t evaluate claims based on background, gender, appearance, skin tone, disability, or any physical characteristic (nor do we evaluate any of these by proxy),” the carrier tweeted.

Bixby said that Lemonade doesn’t monitor facial expressions to detect fraud and never denies a claim based on a person’s appearance. He said its AI can find information in numerous other ways. For example, words that the claimant doesn’t use can provide clues, as well as the customer’s description of the loss.

He said the Lemonade system can also detect fraud if the same face is videotaped filing numerous claims under several policies issued to different people. He said Lemonade has detected fraudulent claims of that type even when the fraudsters tried to disguise their faces. He said the method of using a video to report claims deters fraud by itself because most people find it difficult to exaggerate while on camera.

Still, the criticism trickled in. Rachel Metz, who described herself as a tech writer for CNN, tweeted that the company’s wasn’t telling the truth when it said it doesn’t use AI to auto-deny claims. She pointed out comments in the company’s initial registration documents with the Securities Exchange Commission filed last June before it began selling stock to the public on the New York Stock Exchange.

“Our claims bot handles the ‘first notice of loss’ for 96% of claims as of Sept. 30, 2020 and in approximately a third of cases can handle the entire claim through resolution without any human involvement,” the filing states.

Bixby said the denial rate doesn’t mean that the AI system is using facial information to make a decision to deny. He said in some instances, the policy might not have taken effect yet. In other instances, the amount of loss may be less than the deductible.

“We would never use a person’s appearance to deny a claim,” he said.

The public relations crisis is ironic, in a sense, given that Lemonade said in its SEC filings that social advocacy is built into its business model. The company is registered as a Certified B Corporation, a private certification that recognizes a for-profit company’s commitment to meet social sustainability and environmental performance standards. The certification fits with Lemonade’s marketing, which aims at tech-savvy young people with few assets.

That marketing appears to be working. On May 12, the company reported that in-force premium grew 89% to $252 million since the first quarter of 2020. The company also announced that it will soon start selling car insurance. The company said that will increase sales in other lines because it has been unable to bundle policies to offer discounts to its customers.

Chief Executive Officer Daniel Schreiber said Lemonade competitors GEICO and Progressive have been in business since the 1930s and spend billions of dollars on advertising each year.

“That all adds up to real heft and we have tremendous respect for these competitors as well we should. But strengths and weaknesses are two sides of the same coin, and all that legacy and bulk comes at the expense of nimbleness,” Schreiber said during a conference call.

Motley Fool reported Thursday that Lemonade looks like a good buy. The company closed at $86 per share Thursday afternoon, less than a peak of $188 but far more than the $29 valuation made at its initial public offering last July.

Was this article valuable?

Here are more articles you may enjoy.