There’s a big obvious question that doesn’t seem to have been raised so far in relation to the Citizens Advice ethnicity penalty report. It’s this - to what degree does the problem in pricing set out in their report reflect a problem with the diversity of the people working in digital insurance?
Researchers have shown that bias in AI systems reflects historical patterns of discrimination in the workplaces that design and implement such systems. They are in effect two manifestations of the same problem. As such, for the sector to address discriminatory pricing, it must also address diversity in the insurance people whose work influences pricing.
Now some of you will be saying at this point that the sector has been doing a lot around diversity, equity and inclusion in recent years. There’s been a lot of attention to it in the trade press and on corporate social media channels. That’s fine, but at the same time, doesn’t this then make the findings of Citizens Advice’s research even more surprising?
My point is that there seems to be some form of mismatch here. A diverse workforce would be expected to have influenced and challenged pricing practices so as to deliver pricing that wasn’t discriminatory. So is this a pipeline issue then? Is the pipeline of diverse peoples there but taking time to spread through insurance firms?
Pipelines and Exhausts
Or is the pipeline delivering these people but deeper issues around workplace culture, power asymmetries and compensation are hindering their impact? There may be a pipeline, but might there also be an exhaust, causing people to leave the sector as fast as they come in? Behind the ‘good news’ stories, where are the research to find out what is really happening?
One factor to consider is the sector’s willingness to entertain forms of human state sensing in the analytics it is testing for use in underwriting, claims and counter fraud. AI systems that analyse voice and image data for proxies of physical and mental health, for evidence of fraudulent, for the likelihood to complain, are replete with ethical questions. Often built around questionable science, they can act as red flashing lights to the very diversity the firm is striving to build.
It's not only the insurer themselves that should be questioning just how diverse their teams associated with pricing are. What of the suppliers and business partners who deliver the data, who design and train the systems? Should an insurer do business with a third party that falls short on diversity? This wouldn’t be radical – other sectors have been doing this for years now.
I know that many insurers have diversity policies that say something along the lines of their employees needing to be as diverse as their customers. If Citizens Advice’s research stands the test of time, then clearly there’s work to be done.