26/04/2023

From bias to black boxes: understanding and managing the risks of AI from an actuarial perspective

From bias to black boxes: understanding and managing the risks of AI from an actuarial perspective Valerie du Preez from Dupro Advisory Actuartech, and Richard Galbraith from Canada Life, preview their presentation to the IFoA Conference 2023

With the increased use of, and opportunities offered by, artificial intelligence (AI) and machine learning (ML), actuaries are having to navigate a progressively more complex risk management environment. Through our transregional and interdisciplinary industry focus group, we have identified gaps and challenges actuaries face when it comes to implementing and using modern modelling techniques from a risk perspective in the context of regulation and actuarial conduct.

The current AI landscape is changing at a rapid pace, as shown by the recent boom in AI releases for public use (for example self-driving, e-hailing services, ChatGPT, and Microsoft Copilot). Considering the potentially alarming pace at which AI is being developed, released and used in various sectors, actuaries see the need to use AI to stay ahead of competitors, with the requisite that it is done in a responsible, ethical and risk-controlled manner.

Our expert focus group has identified key themes that encompass the main challenges actuaries are facing, particularly as it relates to the use of AI for modelling and analytics. These themes include:

  • bias, fairness, and discrimination
  • individualisation of risk assessment
  • interpretability and explainability
  • transparency, reproducibility and replicability
  • validation and governance
  • lack of relevant skills.

While there is some regulation and actuarial guidance on how to manage these themes, it can be difficult to consolidate the disparate guidance, and even more challenging to implement it practically. As such, based on our group’s research to identify the key issues faced within traditional and non-traditional actuarial work, we will evidence the problematic elements of each theme within the context of AI, and present regulation, guidance, best practice and practical recommendations on how to navigate the use of AI as actuaries.

It is not only actuaries working in traditional spheres that need to navigate these challenges, but also those working in non-traditional spaces; along with the rapid development and release of AI, that makes addressing the topic an inevitable requirement for all actuaries.

Research group

  • Matthew Byrne
  • Jean Dessain
  • Valerie du Preez
  • Richard Galbraith
  • Paul King
  • Victor Mutanga
  • Frank Schiller
  • Stefan Zaaiman

Additional contributors

  • Ronald Richman
  • Patrick Moehrke 
  • Lara van Heerden

Learn more at the IFoA Conference 2023

Find out more at Valerie and Richard’s session on Monday 26 June at 15:50. Book by 5 May to take advantage of our early-bird booking offer.

Book your place

  • Share on LinkedIn
  • Share on Facebook
  • Share on Twitter