The Cost of Insurance Fraud
Insurance fraud comes in many forms. The Association of European Investigators estimates that approximately 10% of claims are fraudulent in some way. This may include relatively minor types of fraud, such as exaggerating the size of a claim, as well as more egregious acts of fraud. All fraud adds up, and the total cost has a major impact on the insurance industry. According to Insurance Europe, detected and undetected insurance fraud costs European insurers and their customers approximately €13 billion each year. Fraud can occur in many different ways. For example, fraudsters may stage crashes, injuries, break-ins or other insurable incidents and file claims. Other times, the claim is real, but the amount is exaggerated. Insurance fraud can also happen at the application phase when applicants intentionally provide false information. The variations are nearly infinite, and scammers are always creating new ways to defraud insurance companies. This means that insurance companies need to up their game, too, and new technology provides just the way to do that
How Machine Learning Can Thwart Fraud
Before we can talk about how machine learning can thwart fraud, we need to understand exactly what machine learning is. According to MIT Technology Review, most artificial intelligence advancements are the result of machine-learning algorithms, which can find patterns in massive amounts of data. When Netflix gives you a spot-on recommendation, or when Google figures out exactly what you’re searching for, it’s because of machine learning. But the applications of machine learning go way beyond search engines and entertainment companies. Both the insurance industry and machine learning thrive on data, making the two
a perfect match. A human might not be able to detect a pattern of fraud in data from thousands or millions of claims, but machine algorithms can.
GDPR, Data and the Insurance Industry
The General Data Protection Regulation created new requirements for how companies handle data. Many of these new regulations have a direct impact on the insurance industry. For example, according to PrivSec Report, certain types of sensitive data – the type that insurers often need – cannot be processed without explicit permission. Additionally, individuals must be informed about automated decision-making processes, such as those that insurers may use to detect fraud.
This is something that European insurers need to consider, but it does not mean that machine-learning algorithms should not be explored.
According to Nvidia, the GDPR means that insurers need to be able to explain how decisions are made. Toward this aim, the new Intelligent Voice program uses machine learning to identify fraudulent behavior in real time during calls to insurers, and the University of East London is working to show how exactly the decisions are made.
Currently, machine learning is not perfect. For example, legitimate claims may be flagged as fraudulent. In some cases, AI programs may become racially and culturally biased due to the biases of the data being fed into the system. Insurers also have to deal with increased regulation.
However, these drawbacks can – and are – being addressed. Artificial intelligence and machine learning are still in their naissance, but the potential for anti-fraud cannot be ignored.
Want to navigate the future with a tech partner you can trust?