A Multi-Task Network Approach for Calculating Discrimination-Free Insurance Prices

In applications of predictive modeling, such as insurance pricing, indirect or proxy discrimination is an issue major concern. Namely, there exists the possibility that protected policyholder characteristics are implicitly inferred from non-protected ones by models, and thus having undesirable (or illegal) impact on prices. A technical solution to this problem relies building a best-estimate model using all (including ones) then averaging out for calculating individual However, approaches require full knowledge policyholders' characteristics, which may in itself be problematic. Here, we address multi-task neural network architecture claim predictions, can trained only partial information it produces prices free discrimination. We demonstrate use proposed find its accuracy comparable conventional feedforward (on information). has clearly superior performance case partially missing information.