Blog

The volume of insurance fraud in Denmark is increasing. This is the conclusion in Insurance and Pension Denmark’s latest report about this delicate subject matter. The accumulated amount of fraud related to personal injuries (such as work injuries or early retirement), where the size of the claim easily can amount to a few millions (DKK), has increased by 13%. Generally, when it comes to for example loss of work ability or potential disability, we are entering an area where the data is more sensitive compared to a case that is ‘just’ about my stolen bike. This is where data ethics comes into the picture.

The Director of Claims at the Danish financial services group Alm. Brand also recently pointed out that the amount of fraud tends to increase in correlation with increasing digitization of the claims reporting process and as the personal contact decreases, correspondingly. So how can insurance companies fight fraud – both the harder kind where the particular fraudulent activity can add up to for millions and the softer kind where one tweaks the facts a bit during an online claim filing process?

Data Ethics and AI

In the report “Data Usage and Data Ethics”, which was formulated by myself and Nextwork.as on behalf of Insurance and Pension Denmark, we take a closer look at the fight against fraud and how data are – and can be – applied in this process. It is in no way ethically justifiable to commit fraud or submit false data at the expense of others. The data ethical problem arises when honest customers are being wrongfully closely investigated. Some companies might even abstain from going too far when it comes to fighting fraud because bad stories, which are often the easiest to tell in the newspaper columns, can hurt the brand. But is it ethically justifiable to the honest clients who are paying the price? Is it more ethically justifiable to apply more or less data when it comes to fighting back against insurance fraud? It really gets tricky when add to the picture the fact that some customers actually want the insurance company to be somewhat thorough in the claims filing process so that potential fraudsters are caught. Should insurance companies really add useless questions to the filing process just to make customers feel that they’re doing enough?

This is where technology plays a crucial role because AI with many different datapoints can recognize complex patterns in claims and pinpoint those who commit fraud while diminishing the number of false positives fairly early on in the process. So, here we are dealing with a data ethics which is based on the application of MORE data, not less. And Artificial Intelligence can sort through namely unstructured data and use these data intelligently, so the honest clients can continue to file their claims undisturbed.

This is where technology plays a crucial role because AI with many different datapoints can recognize complex patterns in claims and pinpoint those who commit fraud while diminishing the number of false positives fairly early on in the process. So, here we are dealing with a data ethics which is based on the application of MORE data, not less. And Artificial Intelligence can sort through namely unstructured data and use these data intelligently, so the honest clients can continue to file their claims undisturbed.

Honestly – how does one appeal to people’s conscience in digital touchpoints? Let’s look at Behavioral Economics

Especially in regards to ‘soft fraud’, the field of Behavioral Economics can contribute with preventive tactics that motivate customers to stay honest throughout the claim filing process. The awareness of having a counter fraud database, where for example data about a particular claim is stored for several years, will most likely affect the rationale behind fraud. The prospect of being sentenced to a harsh judgment or penalty might have a similar effect. Knowledge about the counter fraud systems, rules and limits that the insurance companies use also has a certain effect and we do in fact see fraudsters move from insurance company to insurance company as the companies’ fraud detecting systems improve. Finally, there is the ethical awareness related to the claim filing process. Here the data ethical maxim is to create incentive for the customer to reflect ethically when he or she files a claim.

In a previous study, the number of kilometers specified by people that were in the process of drawing up car insurance was compared. The higher the number of kilometers, the higher the premium. Half of the people were to report the number of kilometers and sign a sworn declaration saying that the previously stated information was correct. The other half signed the sworn declaration before they filed the information about the number of kilometers. 13.488 insurance policies were filled and the results showed that people generally reported 10% more in the number of kilometers when they signed the declaration at the end of the procedure compared to those who signed the declaration in the beginning [1].

Literature on the subject of Behavioral Economics suggests other recommendations which appeal to ethical reflection when a claim is being filed:

• Awarding good behavior
• Humanizing the report formula for examplevia a personal note and signature from the insurance company’s Director of Claims somewhere in the digital claim reporting process
• Transparency surrounding the direct and indirect costs related to fraud  on the firm or industry level is also a suggestion which can awaken the fraudster’s ethical consciousness regarding the consequences of his or her fraudulent activities
• Information and education about the claim reporting process and for example the difference between subjective justice vs. objective justice
• Stating the guidelines for reporting
• Stating Terms & Conditions as well as codex  for rejecting claims
• Create awareness about the individual’s affiliation to the insurance community
• A stronger business- or industry image will presumably also diminish the incentive and make it harder for some to justify fraud. This is where we enter an area I like to call “data-branding”, where customers are not hesitant to share personal data with the company that they interact with because they have a trusting relationship based on value transactions, transparency and data ethics.

Ready for organizational change? Claims processing and counter-fraud will blend together

The filing and processing of claims will over time run much more automated with Straight through Processing and immediate pay-as-you-go. The fraud investigation on the other hand will blend with the processing of claims in regards to more usage of digitally already available data and well Thought Through Processes that promote honest claim reporting:

1. Transversal integration of more data points with use of technology: Real-time data analysis and automated pattern recognition will be running in the background throughout the claims handling process.
2. Preventative measures: Behavioral economics strategies and tactics are integrated with the claims handling process with a thought through service design, questionnaire techniques etc.

The questions, which the claims handler (human or machine) will ask during the filing of a claim, will be reduced to a total of two or three questions coupled with a long line of already available data as well as automatic image recognition. Any further questions and data collection will only happen in regards to those for whom there is a really good reason to further ask some clarifying questions.

How will the departments that are handling claims look then – in terms of organization? There will be a need for human competences in the design of the filing process, the questionnaire formulation techniques etc. in regards to prevention, and there will be a need for hands at work in the investigative part, where suspicious cases are being investigated. Much of the outlier detection will be handled by AI and transversal data integration. In short, humans capabilities will be needed in the design and setup of claims handling as well as the investigation of potential (many more detected) fraudsters, i.e. investigation of more positives and fewer false-positives.

[1] Shu L, Gino F, Bazerman M et al (2011)