Data Ethics Risks in Insurance
The insurance sector faces a complex range of data ethics risks. Some they share with other business sectors, but several are unique to insurance. This makes it vital for the data ethics risks in insurance to be mapped and assessed with great care. Those risks will fall into four broad categories:
- data risks - the ethical issues associated with the collection of large datasets and the uses to which they are put
- algorithm risks - the ethical issues arising from how algorithms are configured and the uses to which they are put
- practices - the leadership and management given to how the firm uses data and analytics
- governance - the organisational and individual oversight given to the firm's digital strategy
The last few years has seen a swing in the dialogue about data and analytics in insurance, from a strong focus on the opportunities, to an appreciation that there are data ethics risks too. Whatever stage your digital strategy has reached, some form of data ethics risk assessment is now expected.
So who’s expecting it? The regulator for one, and boards of directors increasingly as well. Why? Because of the growing realisation that data and analytics can throw up some significant issues around fairness and discrimination. Such issues have even been discussed at Parliamentary committee.
One marker will certainly be the FCA’s forthcoming review of data ethics issues in financial services. Their expectation will be that for every digital strategy, there will be a data ethics risk assessment.
What they will look for in such assessments will be structure and challenge. The first will tell them about completeness, in terms of both scope and depth. The second will tell them about thoroughness, in terms of ethical culture and objectiveness. So by all means have a long list of data ethics risks ready, but be sure to evidence the structure and challenge in how it was put together.
Contact me to find out about the type of support I give to insurers undertaking a data ethics risk assessment.
Quote | FCA Business Plan 2020/21
Key outcomes we want to achieve - customers are not unfairly excluded from GI&P products and services, particularly in light of increasing use of data algorithms, which can automatically discriminate against consumers with protected characteristics.
Survey | 5 min read
This survey from 2015 found insurers were paying very little attention to equality issues in relation to consumers.
How you think of data influences how you collect and use it. And that in turn forms the bedrock of your data ethics exposures. These articles explore some of the issues involved.
Fairness and Discrimination
These two issues present insurers with the greatest ethical risk from data and analytics. Insurers need to have a thorough understanding of their origins, as well as the how they are evolving and why. In several articles over the years, I have examined the bigger picture.
Quote | Professor Melissa Terras, Edinburgh University’s Futures Institute, 2019
These words from an influential academic sum up one of the key data ethics risks that insurers face..
All data is historical data: the product of a time, place, political, economic, technical, & social climate. If you are not considering why your data exists, and other data sets don’t, you are doing data science wrong.
A Critical Friend
The role insurers most often ask me to perform around data ethics is being a critical friend. I review their approach to data ethics risk from an independent and informed perspective. The reason they often give for using me in this way is that I am seeing risks in ways that others tend not to, and I am backing this up with analysis founded upon academic research. My approach is pragmatic, but also challenging when needed. it is free of parallel interests, and is firmly focussed on building trust in insurance.
Get in touch if you'd like to have an informal chat.
The Influence of Power
The key lesson that insurers should learn from the recent pricing review, and apply to their work on fairness and discrimination, is that data and ethics are tied with power. This I explore in these four articles...