The Swiss Re report ‘To BE or not to BE’ (great title!) came out in late 2021, with the aim to ‘…inspire more industry professionals to ethically use BE to inform the design of insurance offerings’ (report here). It’s an important aim and they support it with a good mix of overviews, checklists, research and case studies, both their own and the work of others.
Their work is build around a key principle: ‘We seek to treat customers fairly at all points in the consumer lifecycle.’ And alongside this comes some important assumptions, for example that choice architecture is ubiquitous and that there is no neutral way to frame choices.
There was much in their report that got me thinking. Yes, the report wanted people to ethically use behavioural economics (BE), but what exactly did they mean by that, and to what extent might the choices made by the authors be non-neutral too?
A bit finickity you may think. What’s the point in asking these things? Well, it’s worth asking such questions on two levels. Firstly, getting the choice architecture right across the customer lifecycle is pretty much akin to addressing some of the main ethical risks that consumers encounter with insurance. And secondly, it’s important that the gap that could potentially exist between the ideal of that key principle, and the reality in how firms go about fulfilling it, be recognised, explored and mitigated.
I’m going to respond to the good work in this report with eight challenges that I believe BE people should now address, in order to deliver on that key principle. Some of these eight points could develop into systemic weaknesses if left unaddressed. They shouldn’t be allowed to undermine the significant good that could come out of the ethical use of behavioural economics.
Let’s look at each of those eight challenges and then end with an overview.
Judgement Free
The report provides a number of tools that the authors suggest people use to apply behavioural economics to insurance. One of them is an ethics checklist and it’s described as requiring ‘concrete and judgement-free actions’, and ‘neutral oversight’. Can that really be though?! Any insurer or reinsurer using an ethics checklist, even with the best intentions, will not be able to deliver it ‘judgement-free’, nor provide neutral oversight. BE people are, after all, people, just like the people they’re studying.
This is not being picky. It is important for BE people to recognise that their work is not judgement-free. It never can be. That doesn’t mean that BE people shouldn’t then use an ethical checklist, but that they should do so with recognition that their decisions will be influenced, to some degree, by their professional background, the interests of who they’re working for, and their life experience. That recognition will then help BE people to be open to, and respectful of, those whose interests overlap but judgements differ.
Unethical BE
This question of judgement can be illustrated through two examples used in the report. Both examples point to something positive, but when you explore them in greater depth, there’s more than initially meets the eye.
The first example relates to the oft quoted case of when BE was used to increase the honesty of self-reported data. An influential paper published in 2012 by two well known BE professors found that when people signed an honesty declaration at the beginning of a form, rather than at the end, they were less likely to lie. This finding was taken up by a motor insurer for the self-reporting of the mileage of the insured vehicle.
Then in August 2021, the authors had to retract their paper, when it was found to have been based upon fake data (more here). The irony here is that this was a paper investigating dishonesty.
The second example relates to the Commonwealth Bank of Australia (CBA) and the way in which it increased transparency in the marketing of its credit card products. Yet this was the same bank that in 2021 was at the centre of a major investigation by regulators. The bank had been selling insurance to people taking out credit cards, personal loans and mortgages, often when those people were unable to claim on those policies.
This was no minor case. It cost the CBA over A$23m in compensation to credit card holders and over A$51m to loan protection policyholders. Then there were fines on top of that.
These two examples are recent and powerful examples of the need for ethics in behavioural economics. Yet they are also examples of how behavioural economics is not judgement-free. The challenge here is for BE people to recognise that and then factor it into their approach and the findings that flow from it.
Responsibility
The report contains an interesting caveat. It says that…
"The Swiss Re Behavioural Research Unit makes recommendations we believe are likely to be in the best interests of the consumer, based on the information provided to us by our client/partner. However, the recommendations should be reviewed and assessment by the client’s/partner’s own legal/regulatory/technical/compliance teams."
What this does is hand responsibility for those recommendations over to the client/partner. It’s the client’s/partner’s problem if the recommendations don’t actually end up being in the best interests of the consumer. Is this realistic?
I encountered something similar to this last November, at an EU research workshop on data and ethics in insurance at the University of Bologna. I was with some of the EU’s main researchers into insurance, listening to a detailed presentation by a Swiss Re team on some data analytics work they had been doing. It was all very interesting, until the rather obvious question arose as to the responsibility that the firm took around how this data analytics work was then put to use by clients. The response was immediate - ‘no responsibility’. This was greeted with disbelief and, to be honest, some amusement. It seemed so untenable, except if you looked at it through the lens of Swiss Re’s position in certain markets.
If Swiss Re is saying something along the lines of... ‘we don’t want to be responsible for how our recommendations are used’, then what does this mean for the accountability of BE designers and users? It is common in financial regulations for the manufacturer to remain accountability for what its partners do with the manufacturer’s products. Shouldn’t BE people expect to be treated in the same way, given their involvement at so many points in the consumer lifecycle?
The challenge here is for BE people to explore how responsibilities and accountabilities are shared within the client/partner relationship, so that an equitable balance can be achieved.
Counter Fraud
One of the interesting ways in which the report talks about the use of behavioural economics is in the combatting of fraud. No examples are given about how this is done, which is a pity, for counter fraud is hugely influential in insurance now.
It is important to remember that most counter fraud operations now work from the point of quote and not just the point of claims. If BE is being used from the point of quote, then it would be, in my opinion, the area where most ethical issues would arise.
Behavioural economics studies, to quote the report, ‘how people truly behave, with the insights used to design products and services that go with the grain of human nature.’ Let’s explore this then at the point of quote, with what is called application fraud.
The level of fraud at the point of quote is gauged by a wide range of factors, from how you enter the data, to the data you enter, to the data that other people like you have entered. That indicator of potential fraud is then used to adjust the quote given, both in amount and how it appears on the quote screen.
This is about judgement, not truth. Counter fraud at the application stage is about degrees, not insights. And these are degrees of judgement that could determine one’s ability to buy a car, get a job or buy a house. In other words, life events. So any type of nudging in such circumstances needs to be handled with extreme care.
In the report’s ‘ethics checklist’, the final item is “would you be able to support this trial if the results became publicly available’. It’s good but doesn’t really go far enough. Intent and method are just as important as results. In using BE to tackle application fraud, all three of those things should be available for public discussion.
So the challenge here is for BE people to understand (and be open about) the way in which their nudge work judges our potential criminality and uses that understanding to influence decisions. The potential seriousness and sensitivity of this needs to be acknowledged.
Whose Interests?
The first item of the report’s ‘ethics checklist’ is… “Is the desired behaviour change in the best interest of the consumer?” It’s an important question, not so much in the desired change being in the best interest, but about how that best interest is determined.
I’ve been in several situations where an insurer was adamant that something being done was in the best interests of the consumer. What became self-evident however was that those best interests were the ones the firm wanted the consumer to have.
I once arranged for that difference to be checked and the results were pretty stunning. What was thought to be the number 1 interest of this particular customer base, was found to be number 22. What was actually the number 1 and 2 interests of these customers had never been measured by the insurer. To their great credit however, the insurer's immediate response was ‘OK, let's work out how to measure them.’ This exercise influenced their thinking for a long time afterwards.
I would therefore recommend to BE people that the following goes at the top of their ethics checklist: ‘find out what the customer sees as most in their best interests’. And next to top would then be to evidence how the behavioural change would support that. Insurers need to understand that they cannot just assume what the insurance buying public want.
The challenge here then is for BE people to build closer and more active relationships with the civil society groups who represent consumers. Their insights can be revealing, sometimes startling.
Fairness
Closely aligned with views on what is in the best interests of the consumer are views on what is fair in respect of the consumer. Much the same concerns around the former hold true for the latter. Behavioural fairness is not something that can be determined by the insurer alone, for their views on it are subject to many interests and judgements. I wrote a detailed piece of analysis on this in 2020 – more here.
As personal lines underwriters have found to their cost here in the UK, views on what is fair in insurance are not simple. They are evolving as we speak and in the direction of what consumers view to be fair, not insurers.
So when the report asks whether a trial was fair, it is important that that means fair from the consumer’s perspective. And so the question then evolves into ‘what does the consumer consider fair in such circumstances, and what do we know about this?’
BE people should build such questions into their work, to avoid their findings being treated by consumers, to be brutally honest, as just another expression of relative power. Yes, it is important that BE people “…approach the challenges and opportunities around ethics …through an empirical approach”. To do so however, the observations, the experiences, the testing must all happen in a sphere that respects the importance of the actual consumer perspective.
The challenge here is for BE to evolve as something insurers do with consumers, not do to consumers.
Trust
Research has found that people generally approve of nudges and that this approval increases as public trust in the institution concerned increases. Great in general, but not so great for insurers. Public trust in how insurers use data is low.
Take the findings of independent research commissioned by the ABI and published in 2020, into consumer attitudes to data and insurance. Consumers were said to be viewing insurers’ use of their data “…through a double-layered lens of mistrust.”. The research went on:
“…consumers are primed to feel particularly cautious and sceptical when it comes to their data in the context of insurance. It also means that they are more likely to interpret new developments in relation to their data as designed to work in the industry’s best interests (for example as a means of increasing prices and profits) rather than their own.”
It’s important therefore that BE people working in (re)insurance recognise that their work is taking place in a context of relative mistrust rather than trust. And in such circumstances, those ethical checklists must be tighter, tailored to the right issues and vetted afterwards. If not, the hopes and aspirations of BE people could well hinder rather than help the situation.
Does this point to the need for two key principles? To the existing “we seek to treat customers fairly at all points in the consumer lifecycle” could be added “we seek to rebuild trust in how insurers use consumer data”. The challenge therefore is to build a methodology off the second principle in much the same way as the report puts forward a methodology for the first principle.
Context and Consent
My final challenge to BE people is premised off a key point that the Swiss Re authors make in their report.
"…there can exist no world without choice architecture. For example, as soon as a company enters the insurance business, it will inevitably implement some form of choice architecture. For instance, the insurer will set prices for policies, and it will phrase the descriptions of the policies."
This is hugely important not only for how everyday insurance people understand behavioural economics, but also how BE people understand insurance and the work they do in it.
If there is no world without choice architecture, then every BE project must itself sit within a context influenced by previous choices. That project must then take that context into account, for if it were not to, its results would be off from the beginning.
Now, influencing that context, in terms of BE people’s use of consumer data, are two ethical issues: consent and secondary use. All consumer data used in BE projects must pass through consent and secondary use hoops. And the shape of those hoops will inevitably shape the BE project. They can’t fail to.
So the challenge for BE people is to recognise the influence of hoops like consent and secondary use on the design and implementation of their projects. In other words, to let the context given to their work by data, trust and consumers influence what they do.
Now some of you may be thinking something along the lines of “hey, is that all!” And I agree, it does sound a bit of a complex undertaking to take on. However, there are ways that can help with this. One that I found useful in the past was the work of data feminists and the tools they suggest for understanding the influence of power (more here).
Summing Up
The Swiss Re report on the ethics of behavioural economics raises some important points. At the same time, those points often in turn raise some important questions, some of which are pretty challenging. I’ve tried to cover eight of them here.
To ensure that the good that can flow from good behavioural economics is secured, we need to address those challenges. The opportunity here is that good behavioural economics could act as a powerful lever for undoing that ‘double lens of mistrust’ that the ABI research identified amongst UK consumers.
