Tackling fraud is an ethical thing to do. It penalises those who commit it, deters those who are tempted by it and benefits those who never would have thought of doing it in the first place. Everyone other than the fraudster benefits from there being less fraud around.
The ethics associated with fraud don’t stop there though, for how you tackle fraud also has an ethical dimension. There’s no point teaching fraudsters a lesson by adopting tactics that are themselves ethically questionable, perhaps even illegal.
However with insurance fraud reportedly costing UK insurance in the region of £1.3 billion a year, the pressure on fraud teams to deliver results must be huge. I’ve been told of UK insurers setting financial targets for fraud teams to meet, and in doing so, placing a significant ethical risk at the very heart of such work.
Some in the insurance sector, perhaps those who have worked long and hard at improving insurers’ handling of fraud, may be tempted to think that the sheer size and complexity of insurance fraud overrides the detail of how it is addressed – in other words, don’t the benefits for everyone of comprehensively getting to grips with fraud outweigh the occasional corner that is cut in order to achieve results? That’s a question upon which the public can supply a clear, pre-existing answer: one set of rules cannot be upheld by the breaking of another set of rules. The ends do not justify the means.
Over the years, I have pointed to a range of ethical issues that insurance fraud teams need to consider. In this post, I’m going to draw them together and provide a comprehensive overview of the ethical landscape for how insurance fraud is tackled. This makes it a long’ish post, so grab a cup of coffee and read through the 14 points covered in roughly the same number of minutes.
Lines that Definitions Draw
The line that insurance firms use for establishing whether fraud is present in a claim is not one that has been set by the law, but which has in fact been set by the insurance firms themselves. Their interpretations of ‘fraud’ and ‘proven’ sit below the legal definition, with obvious consequences. The insurer could label a claimant as a fraudster, even if the case against them would not satisfy a court of law. With data sharing on ‘proven fraudsters’ allowed under data protection legislation, such assessments quickly ripple across the market, making the individual virtually uninsurable. With proper governance, this variance in interpretation might be bearable, but as you’ll see from later bullet points, the reality is far from acceptable. More on this here and here.
The Need for Ethical Leadership
With the implications that being labelled as a ‘proven fraudster’ can have on individuals, the investigation of insurance fraud needs ethical leadership. Yet the industry have been told by senior figures on a number of occasions to ‘do whatever it takes’ to tackle fraud. This creates a dangerously slippery slope. Using data about claimants that could only have been obtained illegally and the surveillance of claimants’ young children outside the home are two real life examples of what a poor ‘tone from the top’ can produce. I suspect that the moral case for tackling insurance fraud has on occasions blinded senior figures to the moral case for insurance firms themselves staying on the right side of the law. More on this here.
Guilt and Innocence
I accept (reluctantly but realistically) that insurance firms will from time to time get it wrong when it comes to an individual case of insurance fraud. The more pertinent question however relates to how insurance firms avoid systemically wrong decisions. One factor lies in how you define fraud and proven (see above). Other factors include behavioural bias and some independence in the oversight.
This calls to mind one of the fundamental tenets of English Law, called Blackstone’s Formulation. It says that “All presumptive evidence of felony should be admitted cautiously; for the law holds it better that ten guilty persons escape, than that one innocent party suffer.” Why – because the security offered by the law only exists if the innocent feel confident in its protection. Insurance firms need to remember this when calibrating the definitions and systems they use for investigating fraud. This is not an operation matter, but one of leadership. More on this here.
Mindsets and Ethical Culture
Another aspect where strong leadership is needed is around the ethical culture relating to the investigation of insurance fraud. What struck me most about the two real-life cases cited above was the apparent lack of challenge to how acceptable those actions were. Yet when I’ve spoken with people from the firms involved, they talked about not knowing “…why we didn’t realise earlier that it was a wrong thing to do”.
Two things would have been behind this: firstly, behavioural mindsets such as groupthink and fading (more on this here, here, here and here), and secondly, the acceptance of rationalisations, such as ‘everyone else is doing it’. Those last five words are not called the ‘five most dangerous words in business’ for nothing. The ethical culture within fraud teams needs to be in sync with the ethical values of the business and of the overall profession. Again, this is a matter of leadership.
So while you tackle cultural issues like behavioural mindsets and rationalisations, what do you try to build up in their place? Critical thinking is certainly a core skill that all fraud professionals should be trained in, for three reasons. Firstly, you’re only as good as your weakest link. Secondly, the impact of many fraud decisions will be enormous – we’re talking of no access to insurance and unable to drive, get a mortgage or some types of job. Thirdly, insurers have taken on the role of police, judge and jury – such a concentration of power needs to be balanced with a matching commitment, in the public’s interest, to the highest quality of process.
The Walk Away
One category of claim to which more critical thinking needs to be applied is the walk-away claimant. These involve the claimant giving up on a claim part way through. This is a characteristic of fraudsters who find themselves too close to detection. However critical thinking will tell you that while all unsuccessful fraudsters walk away, this does not mean that all those who walk away are fraudsters. Some insurers now seem to have grasped this, but it is clear that doubts still linger in places. The great danger is that doubts like this could resurface elsewhere and undermine the soundness of fraud decision making.
As the investigation of insurance fraud has grown, so has the network of business partners and suppliers that help insurance firms fulfil their mission. There’s nothing wrong with partnering with external expertise, so long as their involvement is to the same ethical standards as required of in-house personnel. When dubious work is outsourced to dubious partners, on a ‘no questions’ asked basis, the insurance firm places its reputation into very uncertain hands. With new regulations emphasising individual as well as corporate accountability, individuals careers could be at risk too.
What is important here are insurance firms setting clear and rigorous ethical standards for their fraud investigations, taking steps to ensure that suppliers uphold them too, and monitoring this with robust auditing and ethical due diligence. Industry wide standards can be of some help here, but really only as minimum requirements, for they tend to be set at what every firm will accept, rather than at what the better firms aspire to. They can also suffer from insufficient ownership within individual firms, who often see them as belonging to someone else.
Where a lack of clear ethical standards and boundaries can land an insurance sector is exemplified by what happened to the Korean insurance market in 2012. It centred around KLICS, a fraud database and search system operated by three insurance market associations. The Korean Government not only disciplined those associations, but the sector regulator as well. The problem centred around ‘mission creep’ at KLICS, with data being collected and used for much wider purposes than fraud. I occasionally see signs of such creep happening in UK insurance circles as well, supported by behavioural mindsets that extend fraud perspectives much wider than the public would expect. More on this here.
As fraudsters vary their approach and target, and as insurance firms deepen their experience and expertise, it has become increasingly common for the resulting complexity to be supported by analytical software that falls under the general heading of artificial intelligence (AI). AI allows new patterns to be detected, multiple attacks to be countered: it’s all seems very clever. Yet AI comes with some critical ethical questions. What is the data that the AI tools are being trained on come from? How has it been tested for key attributes such as fairness and non-discrimination? And are those AI tools being deployed in circumstances that reflect their capabilities.
One fundamental attribute is that the algorithms underlying much of AI do not tell you is causations. This means they cannot be used on individual cases, for the correlations at the heart of so much of AI tell you things about the many, not the individual. Have all insurance firms taken this to heart? To be honest, I have my doubts. As a result, there is a risk that these tools will see fraud when it doesn’t exist. And because so much of AI will be ‘black box’ to insurance fraud teams, these cases will just not be picked up by them.
The real value to insurers of AI comes from its ability to not only see what is happening now, but to see where fraud is likely to happen next. In simple terms, a claimant displaying a pattern that fits with previous cases found to be fraudulent, will be subjected to increased checking. And to a large degree, there’s nothing wrong with this, so long as it is proportionate. Yet those simple times now seem long in the past, for such analytics are looking for policyholders who are thought likely to submit a fraudulent claim in the future. They are now being offered renewal or inception terms so onerous that they will take their business elsewhere. This has been described as de-silo’ing fraud investigations, which to a small extent it is, but to a large extent, it is applying fraud-based correlations of unknown significance to large swathes of the public. That’s controversial. Yet there is no apparent evidence for this being accompanied by the serious levels of accountability and oversight that it deserves. There is enough evidence of AI’s weaknesses to justify this.
The public supports the overall idea of insurance firms tackling insurance fraud. The notion that honest policyholders have been paying more premium because of fraudsters has struck a chord with them. And that support is important to the insurance sector, but is it being put at risk through a new practice that strikes at the very heart of how insurance works. Claims optimisation involves the settlement amount being determined not by the cost of the insured claim, but by what the claimant may be willing to accept. The logic is centred around the notion of ‘why pay claimants more than they’re prepared to accept’. It’s neat but it’s flawed.
This matters for insurance fraud people, for what claims optimisation does is break with the notion that claims are paid according to the insured loss that has been incurred. If insurers break from that, then the public may well ask why they should be required to claim only for the insured loss that has been incurred. Here’s the argument that will emerge: surely if one side tries to pay less, it is not out of order for the other side to try to claim more. And in these circumstances, the public approval given to tackling insurance fraud will crumble. It is for this reason that I have referred to claims optimisation as the Achilles Heel of insurance fraud investigation.
Let’s turn now to a sunnier aspect on the insurance fraud horizon. Researchers working with a US motor insurer found that simply by moving a declaration asking the policyholder to confirm that the mileage information being provided was true, from the back of the form to the front of the form, resulted in noticeably increased mileage figures being declared. This was an ethical nudge: a small change designed to encourage different behaviours (in this case more truthful mileages). When taking out a policy or reporting a claim, it is common to be confronted with long and serious warnings about fraud. How much more refreshing it would be if anti-fraud messages appealed to customers’ integrity, rather than presented pre-scripted doubts about their honesty. More on this here.
As the sector’s response to insurance fraud has grown, so have the resources allocated to it. And with such resources come expectations, about what is to be achieved in return. People and processes should of course be aligned so as to uncover and tackle as much fraud as possible. At the same time however, there also needs to be controls in place to disengage how such work is assessed and rewarded from performance measures relating to case or cost counts.
Employees and investigators should be paid for doing a good job, but should not be incentivised on the amount of fraud they uncover, otherwise the system will end up being played and ‘fraud’ found where it doesn’t actually exist. There are evidently still firms in the market willing to performance manage fraud staff on a ‘savings made’ basis. What this does is cast doubts on the ability of such firms to properly manage other aspects of their fraud work.
Tackling insurance fraud is now a significant sector initiative. After its initial success in third party personal injury claims on motor policies, its remit has been expanding, with the many policyholders now being assessed as well as the few claimants. Yet as it has grown, the framework of accountability within which such initiatives are deployed has remained relatively unchanged. Cross sector initiatives such as the UK’s Insurance Fraud Bureau are overseen solely by insurance people, with independent voices being noticeable by their absence. This is fine when such initiatives are getting off the ground, but not fine when that remit keeps on expanding.
Given how wide the various anti-fraud measures are spreading, and how serious the consequences can be for someone labelled as a fraudster by an insurance firm, it seems only reasonable for a ‘voice of the customer’ to figure in that accountability framework. This is not asking a lot, for insurance regulations often ask for just such a thing, but it could achieve a lot, bringing a non-insurance perspective into definitions like proven and fair, and into processes and standards, such as relating to performance management.
A Concluding Looking Ahead
The sector’s initiatives to tackle insurance fraud have come a long way in the last 10 years. And while much has been achieved, questions have been raised. Let’s look 10 years ahead this time. Tackling fraud will draw on data and analytics in ever increasing amounts. And so will regulatory oversight, with tools like machine learning already being talked about for monitoring key regulatory themes like fairness and non-discrimination. Insurance firms need to prepare for the emergence of regulatory oversight capable of forensics levels of monitoring.
Insurance fraud investigation needs to prepare itself for such scrutiny. Questions such as those raised above will hang around, until some controversial decisions emerge and the regulator is given a sharp prod by Parliamentarians to get a grip on matters. The days of insurance fraud being something of a sectoral black box are numbered.
It’s common now to hear of insurance fraud investigation entering a new era. This could be true in more ways than some insurance people think. Questions like those raised above are going to be shaping that new era. Address them and the future will be brighter.