Insurers' data collection methods 'need to be regulated'

A Sydney academic says regulators must take action to prevent discrimination, exclusion and unaffordability of insurance for some consumers, saying insurers stand to benefit from technological advances in artificial intelligence and Big Data.

University of Sydney law academic Zofia Bednarz says there is limited regulation covering what data is collected by insurers and how it is used.

She says consumers have “very little” control over their own data, and there is a “small window” to correct this before practices are entrenched.

Insurers’ investments in services, software and strategies around big data and AI could “become resistant to subsequent regulation”.

Dr Bednarz says Insurers can lawfully collect data from customer loyalty schemes, social media, website browsing histories, wearable fitness tracking devices, telematics or transaction histories, and consumers may not be aware their data could be used to price insurance.

Protections in current privacy and data protection law are “limited in practice”.

This “datafication” of insurer processes may fuel excessive data collection for insurance contracts, generating a “substantial risk of harm” to consumers from discrimination, exclusion and unaffordability of insurance.

“We often don’t know how it translates into the risk assessment,” Dr Bednarz says. “More transparency is needed. There is a lot of opacity and secrecy surrounding underwriting processes and data practices of insurers.”

In a paper entitled “Is your insurance company watching you online and is it legal?,” Dr Bednarz proposes prohibition of the use of external data, limitations on data use, mandating transparency – including explaining the models used by insurers – and higher privacy law requirements and restrictions on use of personal information to “what can be reasonably expected by consumers”.

“Insurers, using new AI and other models, may be able to collect your online data, and apart from anti-discrimination laws, there are no effective constraints on them using that data to price contracts,” Dr Bednarz says.

“Insurance firms may be using our data…to set prices of insurance products, and we have no real control over how our data is then used, processed, aggregated and combined.

“Virtually every ‘digital trace’ consumers leave can be tracked, and the data extracted may potentially be used for underwriting of contracts,” Dr Bednarz says. “Artificial intelligence and machine-learning tools make it possible to obtain valuable inferences regarding risk prediction from that data.

“Inferences that can be drawn from data are very wide-reaching and many of us would find them uncomfortable.

Dr Bednarz says machine-learning algorithms can correctly guess a person’s sexual orientation from facial images, or depression from social media posts.

“Think about all the things that can be uncovered about us from our grocery shopping history alone – our diet, household size, maybe even health conditions or social background,” she says. “Think about information revealed by our social media posts, pictures, likes, or membership in various groups.”