The logic may be 'fuzzy,' but the results help find fraudsters
Identifying the outliers in a claim can help insurers identify fraud earlier in the process.
Tackling fraud has never been more critical or difficult. What’s more, the COVID-19 pandemic has undoubtedly shifted the fraud landscape. Fraudsters are becoming increasingly sophisticated with the arrival of new technologies.
According to insurance experts, fraud increases during times of economic hardship. For example, in a survey conducted after the economic crisis of 2008, fraud bureaus reported the number of referrals and cases opened, on average, increased in all 15 categories of fraud included in the survey. The pandemic caused similar disruptions to our economy and ushered in a wave of behavioral and procedural changes in nearly every aspect of business and government operations and daily life. The changes have presented scammers with new avenues for exploitation, as evidenced by the historic DOJ takedown of 351 suspects allegedly responsible for $6B in fraud — with $4.5 billion attributed to telemedicine schemes. It is easy to see how these uncertain times could be a catalyst for increased levels of fraud.
To keep pace with the fraudsters, insurers, law enforcement and government agencies must look to technology to help scale their anti-fraud efforts. Insurers, for example, have turned to AI technologies to mitigate the increasing amount of risk they face from organized fraud rings. One machine learning tool in particular — fuzzy logic — is finding success in the fraud fight
What is fuzzy logic? How does it uncover fraud?
In real life, human beings often encounter uncertainties that affect their modes of reasoning. For example, when asked a question, human answers sometimes deviate from a straight ‘yes’ or ‘no’ and instead include possibilities such as “most likely” or “maybe.” These ambiguous terms can be referred to as “fuzzy.”
Similarly, fuzzy logic is an artificial intelligence method that solves problems using qualitative (fuzzy) data to produce quantitative (measurable) results. Fuzzy logic is designed to closely mimic the ambiguity (by machine standards) of human reasoning. It can identify outliers in a given data set.
From an insurance perspective, fraud and risk are outlying human behaviors that deviate from what is considered normal within a given peer group, which is a group of individuals that possess similar attributes. For example, a person from Southern California claiming that their car was damaged in a snowstorm is a flag for fraud, as their claim attributes would differ significantly from those of their peer group as other car crashes in Southern California would not likely be caused by snow.
Fraud and risk are outliers that are particularly difficult to identify. This is because these outliers’ most readily identifiable characteristic is being “different” than one’s peers. But how does one define “different?” In a fraud context, the word different is ambiguous. A fraudulent claim or person can be very different, moderately different, slightly different and so on.
This is where fuzzy logic comes into play. We’ve learned how to use fuzzy logic to analyze the degrees of difference that exist in fraudulent claims and activity. This AI tool uses math to express these varying degrees of difference, ultimately producing an assignable score to each claim and person indicating the degree to which the claim or person is an outlier. A score that depicts significant deviation from the norm could then be flagged for closer examination for fraud.
Humans cannot categorize every possible degree of difference that may exist in a claim document across an entire insurance business. So we’ve turned to the sophisticated AI systems used in the aviation and aerospace fields to do it for us. The same underlying technology used to independently fly aircraft is now being leveraged to autonomously choose the mathematical representations of the fuzzy terms found in an insurance claim — performing billions of comparisons to assign a score for the degree to which any claim or person is an outlier.
The goal of a successful fuzzy logic machine is a dependable way to flag fraud. There are countless examples of fraudsters using advanced technology to steal millions and 32% of consumers suspect they’ve been targeted by fraudsters. This makes it all too clear that as fraud fighters, we must keep pace with our adversaries. The implementation of fuzzy logic and AI is key to furthering innovation and success in the fight against insurance fraud.
Gary Saarenvirta is the founder and CEO of Daisy Intelligence, a firm specializing in creating AI solutions for insurers. Saarenvirta previously served as former head of IBM Canada’s data mining and data warehousing practices. To learn more about fuzzy logic and mitigating fraud and risk through AI, download our latest whitepaper ‘Identify Risk and Avoid Fraud with the Halo Effect’.
Reprinted with permission from Daisy Intelligence and the Coalition Against Insurance Fraud.
Related: