Turing's test and believable AI in games

Abstract

The Turing test is perhaps the most famous, most quoted, and probably most often misrepresented and misunderstood measure of machine intelligence. In this article we'll briefly review the Turing test and some of its key criticisms. In particular, we will try to answer whether the Turing test -- or something derived from it -- can be of use in developing and assessing game AI. We will also present a brief overview of a methodology for conducting believability testing for games and highlight some of the problems inherent in any attempt to categorically determine whether or not some AI behavior is capable of convincing, life-like behavior.