Reflections on deep learning and the actuarial profession(al)

28 Pages Posted: 17 Jan 2024

See all articles by Roseanne Harris

Roseanne Harris

University of Witwatersrand, Johannesburg, South Africa

Ronald Richman

Old Mutual Insure; University of the Witwatersrand

Mario V. Wuthrich

RiskLab, ETH Zurich

Date Written: January 10, 2024

Abstract

We discuss some of the professional consequences of rapid advances in deep learning techniques applied to actuarial science. Since actuarial work is highly regulated by standards and professional guidance, we survey relevant aspects of the guidance in the United Kingdom and South Africa, that apply to actuarial deep learning models. A selective survey of recent advances in methodology is then performed, showing how these advances can be used to ensure compliance with guidance on issues such as model understandability, avoidance of bias and discrimination and variability of predictions. We also discuss the current treatment of machine and deep learning in the actuarial education syllabus and make suggestions for a new subject covering these topics in more detail. Finally, we discuss the evolving role of the actuary and briefly consider consequences of large language models on actuarial work.

JEL Classification: Deep learning, actuarial profession, professional guidance, actuarial models

Suggested Citation

Harris, Roseanne and Richman, Ronald and Wuthrich, Mario V., Reflections on deep learning and the actuarial profession(al) (January 10, 2024). Available at SSRN: https://ssrn.com/abstract=4672447 or http://dx.doi.org/10.2139/ssrn.4672447

Roseanne Harris

University of Witwatersrand, Johannesburg, South Africa ( email )

Ronald Richman (Contact Author)

Old Mutual Insure ( email )

Wanooka Place
St Andrews Road
Johannesburg, 2192
South Africa

University of the Witwatersrand ( email )

Mario V. Wuthrich

RiskLab, ETH Zurich ( email )

Department of Mathematics
Ramistrasse 101
Zurich, 8092
Switzerland

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
557
Abstract Views
1,134
Rank
103,389
PlumX Metrics