ABSTRACT
Effective testing tries to ensure AI meets design requirements and reacts and handles unforeseen situations properly. Experience helps to identify common pitfalls and errors. Seasoned AI developers can identify these and work with sets and categories of tests that try to catch problems, making their systems more robust. This chapter shows some of these strategies, analyzes different scenarios, and provides tricks to help AI engineers approach AI testing with some inside knowledge. It presents an example problem and uses it to illustrates the different tests it can perform during the creation of an AI character. The tests described in the chapter—especially the basic functionality tests—should be applied to smaller parts. The chapter focuses on the different parts that make up AI agent and shows some potential tests that include basic functionality tests, perception stress tests, reaction stress tests, targeting stress tests, unreachability stress tests, navigation stress tests, weapon-handling stress tests, exotic elements stress tests, and group and timing stress tests.
