The IAIS is an umbrella body which brings insurance regulators from all over the world together in a bid to make supervision of industry more consistent. In a draft paper on the use of big data analytics in insurance, it said the issue was one of a "number of challenges and risks" insurers may encounter when using algorithms to process customer data.
"Algorithms can be complex and are often treated as proprietary and highly confidential in nature," the IAIS said in its paper. "Thus, there can be a lack of transparency and strong asymmetry of understanding between those who design and use algorithms and customers and supervisors trying to understand the outcomes generated by these algorithms."
"Machine learning algorithms are based on historical data and, therefore, generally reproduce the past. This can increase the likelihood that algorithms may perpetuate unforeseen biases, which in turn create risks of errors potentially resulting in inequitable or unsuitable customer outcomes. Additionally, the effectiveness of an algorithm is dependent on the quality, accuracy and completeness of available data, and can be hampered by possible errors in its initial design or programming. There is also a risk that customer segments could be differentiated on the basis of false assumptions or false conclusions drawn by algorithms on the basis of these assumptions, resulting in unlawful discrimination against certain customers," it said.
According to the IAIS, insurers are increasingly using big data analytics software. This includes for delivering automated advice and in other customer interactions, as well as to inform product design, more targeted adverts and for credit assessment and customer identification purposes. They are also using big data analytics to better individualise risks and pricing, it said, highlighting the potential benefits for both insurers and customers that can stem from use of the technology.
"The granularity of data from multiple sources can lead to more personalised and affordable insurance products as well as more efficient servicing for customers," the IAIS said. "Insurers can also benefit from big data analytics by expanding their distribution reach, ensuring more accurate pricing and lowering their cost margins due to claims savings and better fraud detection."
In its paper, the IAIS set out how insurance market regulators might respond to concerns over the use of algorithms. It stressed existing regulatory principles in relation to risk management and the fair treatment of customers are relevant to the use of the technology, in addition to other aspects of regulation focused on transparency and simple, clear customer communications.
Regulators might consider developing codes of good practice and ethical use guidelines for algorithms in insurance, and could also look deeper at how insurers "ensure error and bias-free programming of algorithms to the extent reasonably possible", it said.
"Depending on the complexity and opacity of some algorithm processes, supervisors may also want to consider taking steps either themselves or through reliance on other independent audit or validation parties to conduct sample verification and integrity checks both on the algorithm process itself as well as the outcome of the process in ensuring fair customer outcomes," the IAIS added.
The IAIS' paper is open to consultation until 16 October.
Insurance law expert Charlotte McIntyre of Pinsent Masons, the law firm behind Out-Law, said: "Big data analytics is increasingly used by insurers, both directly in customer interaction through the use of AI, and through the use of algorithms and advanced analytics to design greater individualisation of risks and pricing. Insurance supervisors and regulators around the world are beginning to grapple with this issue, with Charles Randall, chair of the FCA, having delivered a speech on the use of big data analytics in insurance in 2018."
"In his speech, Charles Randall noted the ways in which big data analytics is already impacting the insurance industry, indicating that this is an area to which the FCA will give closer consideration. He stated that regulation is central to the debate around big data analytics in insurance, because such regulation 'will help define whether AI and big data liberate customers or disenfranchise them'," McIntyre said.
"In its consultation paper, the IAIS states that its aim is to better understand the benefits and risks associated with the use of big data analytics and to ensure insurance supervisors and regulators develop proportionate and appropriate guidelines, whilst at all times ensuring the fair treatment of customers is not compromised," she said.