THOUGHT LEADERSHIP Sponsored by dacadoo
Artificial Intelligence (AI) in insurance – Balancing the promise with ethical dilemmas
The insurance sector is on the cusp of entering a new era. At the core of its transformation is the sector’s employment of policyholder data generated from Internet of Things applications
Consumers are widely believed to benefit through more personalised service, faster claims processing and lower premiums. Operators, too, are set to profit from selling more custom designed policies and upselling other relevant products to their inforce base.
In order to pave and sustainably secure the way for AI its proponents, including dacadoo, need to take stakeholder concerns and sensitivities very seriously. Failing to do so could result in regulatory and legislative backlashes with the potential of derailing the triumph of AI and wrecking its many positive effects on Life & Health insurers.
One major stakeholder concern is related to the potential for discrimination. Of course, a powerful argument for ever more personalised underwriting lies in what is known in insurance as adverse selection. An insurer whose premiums are based more on a pooled rate than a personalised one will find high risks being attracted to its portfolio and low risks heading for the exit. Generally, this risk-based underwriting approach as enabled for example by the dacadoo Risk Engine, is seen as fair as long as the personalised rate reflects individual behaviour. Otherwise, regulators may clamp down on price discrimination.
Genome sequencing arguably poses the most complex set of challenges involving the ethics of data use and disclosure. AI and machine learning applications have enabled quantum leaps in terms of interpreting and acting on genomic data as well as in reducing the cost of genome sequencing which has fallen to a small fraction of what is used to be at the beginning of the 21st century. As a result, the number of genetic tests has sky-rocketed, shifting the balance of data power towards the individual.
For insurers, this shift is presenting major challenges. Insurance is primarily based upon the principle of pooling homogenous risks. When a potential policyholder has information about his or her health that is not shared with the insurer, anti-selection could be inevitable, calling into question the viability of the insurance offering.
The challenge of anti-selection versus the potential denial of insurance is most acute in life, disability, critical illness and long-term care insurance.
Existing regulations concerning the use of genetic information generally apply to employment and health insurance. For life, disability, critical illness and long-term care insurance, regulations differ widely, ranging from no regulation whatsoever to prohibitions on using results from existing tests. Therefore, in certain regulatory environments, the insurer is at a severe informational disadvantage, with potentially adverse effects on its financial viability. dacadoo’s Health Score in combination with its imputation engine helps mitigate this problem which could lead to massive disruptions in insurance markets. The Health Score enables a win-win-approach for both the insurer and the policyholder, with fairer pricing for both sides while protecting the consumer’s personal data and complying with rapidly evolving data protection and usage legislation.
“First and foremost, insurance companies have to make sure that customers can trust that the upside of AI outweighs the downside by far.”
In summary, AI holds the promise of revolutionizing the way insurers manage and transfer risk. However, as with any powerful new technology there are challenges which need to be reconciled with the opportunities.
First and foremost, insurance companies have to make sure that customers can trust that the upside of AI outweighs the downside by far. This is primarily about transparency and diligence in terms of capturing, processing and using personal data as well as building and overseeing algorithms – and it should go beyond mere compliance with GDPR and other legislative requirements. At the same time, AI must be shown to make a measurable contribution to an enhanced customer health and overall positive lifestyle experience, less onerous claims procedures and lower premiums that reward individual behaviour.
Once the ethical and moral hurdles described in this article have been cleared, AI solutions, on a sustainable basis and with a “license to operate” from the public, will deliver game-changing breakthroughs (e.g. in diagnostics) and improvements in quality-of-life and lifestyle navigation. Ultimately, AI will be instrumental in making insurance more relevant, appealing and affordable to customers. However, just like in mobile communication where governments defined the GSM standard which enabled a global explosion in mobile communication, we need the right ethical framework for AI to prosper in the insurance industry and the digital societies of the future at large.