AI code of conduct for claims launches

A voluntary code of conduct for using AI in claims processes and the supply chain was launched in the city of London yesterday by the Code of Conduct Claims Working Group.

Eddie Longworth, founder and director of JEL Consulting and brainchild of the code, said that customers and policyholders need to be able to understand how AI is being used to make decisions about their insurance claims.

He said the idea for an AI claims code came while reading about a health insurer in the US that was being sued over policyholder insurance claims that an AI had rejected in mere seconds.  

“A class action was brought against the insurer because they had managed to process and reject 300,000 claims in two months. That’s an average of 1.2 seconds of consideration for each claim.”

The AI had enabled this high speed decision process. 

Longworth said that the policyholders in the case had countered that this process was not fair or equitable. 

“The question they asked the insurer was ‘can we justify each and every one of those decisions even though it only took 1.2 seconds?’ The answer was no, they couldn’t,” he said.

He said his vision for the code was to address this “gulf of misunderstanding and communication between the supposed dark arts of AI and the recipients of the outcomes of AI, I.E. the policyholders and the claimants”.

Under the code any information that is used in AI, for example programming or decision making, must be verifiable and open to audit.

“Anyone should be able to come along and understand what the AI has been used for and the conclusion it has made,” he said.

The second requirement of the code is that there must be a logic behind the use of AI. 

Longworth said: “If I as a consumer or claimant said: ‘Explain this to me, what is the logic and how have you reached that decision?’ I need to be able to understand it without being an expert in AI.”

The code attempts to tackle the issue of bias in AI by requiring signatories to ensure that the AI they use or deploy avoids bias whenever possible.

He explained that AI can be automatically biased if it is trained on information that is biased. “If you ask ChatGPT to define a middle class English person, that person is always white. But that is not reflective of the population.”

The code also requires a governance mechanism and a right to redress for claimants be put in place. This is to ensure that if a policyholder disagrees with a decision made by AI and wants to appeal, the appeal decision should not be made by the AI. 

“The code of conduct is about putting some stakes in the ground. It’s saying to ourselves as an industry and as practitioners that AI is upon us. It’s not new, it has been around for decades, but it is gathering pace at an almighty rate of knots and particularly in claims and supply chain spaces. So it’s time to get hold of this and say ‘how are we going to control this?’.”

Claims and the supply chain will remain the focus for Longworth and the group of experts behind the code. He said: “I’ve already been asked: ‘When are we going to do a code of conduct for the whole of insurance including underwriting?’ and the answer is we are not going to do that.”

To sign up to the code, visit: https://www.aicodeofconduct.co.uk/

Did you get value from this story?  Sign up to our free daily newsletters and get stories like this sent straight to your inbox.