Pricing is a central activity of actuaries to assure solvency and to manage risk. Life insurance is also a social good, not just a product, and the US population is massively underinsured. Pricing must strike a balance. Unlike a simple 4x4 matrix to produce pricing, today's technology offers the ability to evaluate literally thousands of variables, enabling a kind of precision pricing. It also heightens the possibility of introducing bias and discrimination through the use of unregulated third-party data.
Machine Learning comes with insidious limitations such as Shortcut Learning and Adversarial Perturbations. We will introduce a concept of fairness that can be evaluated with mathematical models such as Conditional Comparative Demographics versus Disparate Impact, or Shapley Scores and Differential Privacy.
At the conclusion of the session, attendees have a better understanding of the ethical risks in digital technologies for pricing and will be given some guidelines.
This session is presented by the Joint Risk Management Section.
| |
|
|
|
|