Emphasizing the Match Between Computer Model and Operating Environment

This essay draws attention to the relationship between the computer models used to power “artificial intelligence”, and the operating environment in which the model is used. Depending on the environment, it may be that additional data, computing power and model sophistication will produce better predictions. Alternatively, the environment may exhibit novel behavior that limits the applicability of large data sets and highly flexible models. In this case, more traditional models or even simple business rules may perform best. Actuaries are encouraged to relate this to the use of models in insurance operations by classifying those operations according to their predictability and controllability.