‘Machine Decisions’: Governance of AI and Big Data Analytics
In its latest publication, the CRO Forum examines how to adapt existing governance to cover Big Data tools.
June 4, 2019
As in a number of other sectors, insurers are developing and deploying increasingly complex Big Data tools, first for post-sales purposes such as claims handling and customer service, but also potentially for sales, pricing and underwriting. In its latest publication, the CRO Forum examines how to adapt existing governance to cover such tools, including machine learning algorithms, and what reasonable steps can be taken to ‘know what’s happening inside the black box’, ensure reliability and avoid unfair discrimination.
Big Data and AI present particular challenges due to complexity, self-calibration, autonomy and the potential for unexpected results and unforeseen impacts. The report explores these issues with the use of case studies and a group comprising data scientists, model risk experts and governance experts. The paper focuses solely on the (re)insurance industry and is meant to provide a practical overview for risk managers.
The key principles emerging from the case study discussions include the following:
- It is vital to invest sufficient effort upfront to define the problem the complex tool / modelling technique is expected to solve in pursuit of a business objective.
- Human intervention is and will remain critical
- Training needs should apply at all levels in an organization.
The report then explores the complex area of ethics and bias in data and how such bias, if not understood and managed, can cause significant issues for an organization.
Building on the findings, the paper eventually gives an example of an overall model governance framework for Big Data tools, and provides a checklist that the reader can reference when needed. Key here is that model governance techniques and frameworks that exist today do not need to be fundamentally altered, but can be enhanced and adjusted to meet the evolving needs of complex tools and machine learning developments.
SCOR, as an active member of the CRO Forum, participated in this study.
About the CRO Forum
The CRO Forum focuses on developing and promoting industry best practices in risk management. Comprised of Chief Risk Officers from multinational (re)insurance companies, it aims to represent the members' views on key risk management topics, including emerging risks. It regularly publishes practice and position papers on timely risk management subjects. Frieder Knüpling, Group Chief Risk Officer of SCOR, currently serves as Vice Chairman of the CRO Forum.