Levels of Sophistication of Humor Intelligent Agents

The idea of intelligent (embodied) agents for humor pioneered by Nijholt (2001, 2002) and Stock (1996; see also the fundamental Stock and Strapparava, 2002) has had an impressive start but, as with all really interesting intelligent agent projects, is far from complete implementation. Mindful of all the difficulties and obstacles on the way to full implementation, most but not all related to AI and NLP issues, this paper attempts to explore one particular aspect of such implementation, namely, levels of sophistication. This aspect is of great concern to software developers and manufacturers because of the customer acceptance factor (see Raskin 2002 on the crippling effect of customer nonacceptance of information assurance and security products on the world-wide ability to protect computer systems). The question is whether, given an implemented humor intelligent agent (HIA), the customer will find the agent useful, interesting, and stimulating enough to purchase after a trial use and to continue using for a long time, thus spreading the good word about the product. To keep a human interlocutor interested another human has to be two not entirely unrelated things: • non-repetitive, innovative, creative, capable of renewing the interlocutor’s interest, and • capable of communication at or slightly above the interlocutor’s level of sophistication.