Data and Power pt1 – the starting point for insurers’ data ethics journey

The most significant ethical theme that insurers need to address over the next three years is data ethics . Just as most insurers have a digital strategy, there is a growing expectation that those strategies should now have a data ethics component to them. So how should an insurer go about incorporating data ethics into their digital strategy? It’s more easily said than done, for data ethics is more complex and nuanced than most insurers realise. In a series of posts over the next few weeks, I’m going to explore some of the key issues and approaches that insurers should tune into.

Let’s begin with two obvious questions. Why does data ethics deserve such careful attention? And why can’t insurers handle data ethics in the same way as they’ve been handling ethics in general. The obvious example to answer these with is the pricing review, being a clear precursor to the wider debate about the ethics of what can be done with data and analytics.

We know that in 2016, the UK regulator reported that it was fairly relaxed about the data practices being used in general insurance pricing. Their message was that apart from a few wayward practices, all was well. Move on to 2018 and the consumer group Citizens Advice submits its super-complaint. This sets off a regulatory investigation into pricing practices, particularly those associated with lifetime value modelling (LVM).

What this represents are three perspectives at play around the fairness of insurance pricing:

• the insurer’s view of the fairness of the LVM they’re using;
• the regulator’s view with regard to what insurers were doing with LVM;
• the consumer group whose view was very different.

Powerful Voices Shaping the Data Ethics Debate

What this illustrates is that not everyone thinks of the ethics of data and analytics in the same way. Sure, data ethics is a developing field and so differences of opinion are to be expected. Yet there’s more to it than that. What we’re seeing is the emergence of some powerful voices wanting to be involved in the shaping and direction of the data ethics debate in insurance.

Remember that quote from the economic sociologist José Ossandón that I’ve referred to in earlier posts: insurance price is… “not simply a matter of supply and demand, but rather the product of a wider range of actors, including regulators, lawyers, policymakers, members of parliament, consumer associations and representatives of the industry.”

You might question this, but look at the evidence. In 2018, a single

caused the most widely adopted pricing model in retail GI to be put under regulatory scrutiny and deemed unfair. In 2011, a consumer group fundamentally changed underwriting practices by winning its test case on the equal treatment of gender by insurers.

What these cases illustrate is that fundamental change in insurance is just as likely to come from outside the market, as from inside it. Both cases dealt with issues of fairness and equality, and those are issues central to data ethics. What this points to therefore, is that data ethics is going to be shaped not just by insurers and the regulator, but by powerful actors from outside the sector as well.

Don’t ‘Go It Alone’

Should the sector and its regulator seek to ‘go it alone’ and attempt to define and steer the data ethics debate in the way they feel is best, they will not succeed. They may think they are the ones with the technical expertise to deliver ‘data ethics’ for insurance, but if the last ten years of insurance tells us anything, it is that other actors are prepared to, and able to, stamp their views on that debate.

What this points to then, is that the way in which data ethics in insurance takes shape must be seen through the lens of ‘data and power’. In this digital age, power no longer just rests with firms, trade bodies and regulators. The days are now fading when their experts can say ‘this is what is right, and this is what should happen as a result’. Other voices can now make themselves heard, and some of those voices are prepared to challenge how insurers handle their data ethics issues.

So what should an insurer do? Well, those who choose to just ‘do data ethics on their terms’ will over time struggle at best, fail at worse. Ignoring those other voices, ignoring the  way in which they can take over an ethics debate even in a relatively ‘technical’ sector like insurance, is, to all intents and purposes, just burying one’s head in the sand.

Some Fundamental Realities

“Can’t someone just tell us what to do”, some will protest – “after all, we’re good people and just want to get on with doing business.” Yet the pricing complaint showed how dangerous a strategy that could be. The regulator may have signalled as recently as 2016 that all was well in the pricing world, but they got it wrong. It turned out that they weren’t the only actor with views that mattered.

“Yeah, but isn’t everyone doing digital now? Why can’t insurance just get on with it like everyone else?” And I can understand that frustration, but the sector has to face up to two fundamentals. Firstly, that the sector has become a gatekeeper for so many aspects of everyday life: driving a car, owning a home, and making sure dependents are looked after. That brings responsibilities with it, whether the sector likes it or not.

And secondly, data and algorithms are changing some of the fundamental tenets of insurance, particularly around the way in which pricing, covers and claims are being personalised to ever more granular levels. This means that how insurers digitise their products and services has wider repercussions than many insurers think.

Together, these point to it being pretty important that insurers think very carefully about their approach to data ethics and be prepared to adapt their digital strategies to align with the expectations that they will increasingly come under.

The New Power Relationships

Consider this approach, which I believe to be fairly common now. An insurer will form a project team to work on the detail of their handling of data ethics, build what they learn into implementation and operating plans, and track how this turns out for reporting to the board and regulator.

What this does is pretty much fix the firm’s expectations of what data ethics represents for their firm, and cause them to then have to defend it against detractors of various types. This type of defensive approach will make those adopting it feel like they’re often being challenged and put on the back foot. Not a great atmosphere in which to deliver a transformation of their business.

How can data ethics be done differently then? I think the starting point for a typical insurer is to understand how the new ‘power relationships’ around data and ethics are changing, and approach data ethics in ways that acknowledge this. They need to do what I’ve previously referred to as the most ethical of all activities – listening – and to learn how then to absorb what they’re being told.

The insurer needs to build an engagement plan that sustains this listening and encourages it to become deeper and most meaningful. For most firms, this is a much harder job than they realise, for the sector has a habit of shaping debates according to their own terms.

A New Kind of Engagement

From this type of engagement will emerge questions about what data the insurer is using and for what purpose. It will raise questions about the objectiveness of this, and around the extent to which business decisions can confidently be based around it. It will bring certain values into prominence, moving data science from delivering a rational ‘truth’ (as per “the numbers speak for themselves”), to something that is imbued with values like fairness and equity.

Facial recognition analytics is a perfect example of the problems that arise when such engagement does not happen. Developed, trained and tested by a pool of technologists made up almost entirely of light skinned males, it’s hardly surprising then that researchers have found that darker skinned females are up to forty four times more likely to be mis-classified than lighter skinned males. It can hardly  be surprising then that such analytics are now being banned in several US states and are being challenged here in the UK.

So what does this add up to? One thing it definitely does not mean is that data ethics will be defined and scoped by a joint team made up of insurers, regulators and big consultancies. They typically see themselves as powerful influencers of what the ‘norms’ of data and ethics should be, but their role will end up being only partial, like that of a facilitator. Their weakness is that their approach is centred around data ethics being something ‘they will do for you the customer’.

The Shapers of Data Ethics

Data ethics will increasingly become something shaped by the people impacted, not something done on behalf of the people impacted. Power on the playing field of data ethics is changing and those insurers who recognise this reshaping and absorb it into their approach to data ethics will find themselves more trusted, more respected as a result.

Sure, it feels like a lot of work, and some of it will involve insurers doing things that they’re not that familiar with. Yet think of it as a form of long range risk management for the delivery of your digital strategy. If you like, a form of moral compass that increases your firm’s chances of ending up on the digital superhighway rather than off the proverbial cliff.

Questions to Reflect Upon

A couple of recent blog posts (here and here) have in fact covered some of the steps that insurers can take to get started on this, and I’ll weave them in the following four questions for insurers to reflect up:

• What sort of data relationship does your firm want to have with consumers? This has obvious links with the regulatory interest in ‘purpose’.
• How has your firm been engaging with customers in ways that support that type of relationship? I use an assessment matrix for this that can often raise eyebrows in firms thinking about this for the first time.
• To what extent does your existing management information reflect the type of relationship you want with consumers? Your MI needs to represent consumers in ways that reflect the sort of relationship you want to have with them.
• How should your approach to data ethics reflect these things – relationship, engagement and accountability?

As mentioned earlier, over the next few weeks, I’ll be looking at three important themes within data ethics: discrimination and access, due diligence and learning. And in each of them, I hope to illustrate why recognition of the emerging power structure around data and analytics is so important for your firm getting its approach to data ethics right, and for delivering an effective digital strategy.