Technical Consultant and Lead Data Scientist Definity Financial Corporation
In this session, we will describe one insurer's experience with integrating bias and fairness considerations into its predictive modelling plans. Our objective is to educate the audience on the practical considerations of operationalizing the bias and fairness concepts that have been the subject of presentations and research papers by the CAS and other organizations. We are members of a Working Group that has been spearheading efforts to increase adoption of bias and fairness checks by our company's predictive modelling teams. We will share some practices to achieve this goal and demonstrate their use through a case study. Examples of these practices include developing qualitative guidance on when and how to incorporate fairness considerations into a modelling project plan, ensuring that analysts have access to guidance to assist with selecting technical methodologies for detecting bias, and developing supplementary in-house content to facilitate adoption of these tests throughout the model lifecycle.
Learning Objectives:
Modify a predictive modeling project plan to ensure that bias and fairness are considered at all stages of the model lifecycle.
Recommend approaches to selecting relevant bias detection tests for a predictive model.
Develop a plan to address barriers to adoption of bias and fairness checks.
