By Jimmy Williams, CEO of Urban Jungle
I think we’ve all got used to the idea that the insurance industry can use its data and models to decide on how much we pay. We accept our profession can impact our car insurance and that renters pay more than homeowners - but that doesn’t mean it's right.
At our business, we hire and interview a lot of young people, and it’s interesting just how many think it’s completely unacceptable to decide on premiums in this way. They don’t think someone should be denied insurance or have to pay due to factors largely beyond their control.
When I set up Urban Jungle, it was because I was upset at how I’d been treated while I was renting in London. If you’re young and living in shared accommodation, a lot of the insurance industry won’t talk to you, or will make you pay through the nose.
Understandably, a lot of young people feel that’s an injustice, and they’re right.
Older people can be impacted, too. Once you’re past retirement age, a whole range of factors apparently make you higher risk, and it’s hard not to feel a lot of this is discrimination.
Honesty is the best policy
One of the best ways to decide on someone’s risk level is the extent to which they are telling the truth. If someone provides information which turns out to be false then that can raise their risk profile. I’ve no problem discriminating against fraudsters, liars or cheats; quite the reverse. But I do have a problem with charging people due to their age, marital status, race, or the country of their birth, and that’s why we take big steps to never do that.
Algorithms
The insurance industry has tonnes of data on us all and has, for many years, created and used algorithms to assess risk. However, if these are used without any consideration of ethics, it can take us down some dark paths. It’s possible to find all sorts of correlations in data and, therefore, deem one person to be more risky than another, because they meet certain criteria.
One demographic may have a higher claim rate than another, but it doesn’t mean people living in that area are more dishonest.
We are now using more and more data - big data - and this is powering the development of artificial intelligence.
However, we control what goes into those algorithms and what they should be measuring. When we create such AI-based software without ethics and diversity in mind, we end up with some awful results. AI can learn to discriminate against certain groups, including racial groups, if the initial data suggests it should do so. For example, AI may find correlation with a certain surname, but that surname may be particularly prevalent in a certain ethnic group. If we allow a machine to believe these two factors are linked then it can end up behaving in a racist way.
We have to start bringing more ethics into insurance and working on what factors are causal and where there is simply a correlation. We use data to look for behaviours with indicate that someone is lying or likely to commit fraud, independent of what demographic they are from. By allowing us to pick out those who are lying, we can offer insurance to loads of people that others simply write off for being in the wrong demographic.
Somewhere along the line, someone saw a correlation between a person getting divorced and being more likely to make a claim.
There’s no reason why one should affect the other.
Key information
In order to provide insurance, you need to establish someone’s identity. The two most important pieces of information are name and date of birth. Once you’ve also got an address then you can be pretty sure you’ve got the right person. You do need a date of birth, because John Smith might have a son called John Smith who lives at the same address. However, just because you take information from people, it doesn’t mean you have to decide your pricing on it.
So, the insurance industry actually doesn’t need to take a vast amount more information to decide on your risk. But, more importantly, it can decide on which factors it uses to decide on your risk - your name and date of birth should not be taken into account. When you’re providing insurance, it shouldn’t matter whether someone was born in this country, lives in a certain postcode, is divorced, married, owns a property, rents or is self-employed.
|