Twenty Questions

In the game of Twenty Questions, we exploit both our ignorance and questions as the condition and the medium of our play. A brief comparison with riddles helps explain how. In riddles, we use some blocking mechanism—generally by phrasing the mystery item in words different than we would normally expect—to generate a space of explicit ignorance that must be traversed to solve the riddle. In Twenty Questions, the random selection of some object or class of objects, of an “animal, mineral, vegetable, or other” persuasion, functions as such a blocking mechanism. The more obscure the mystery item, the greater the space of ignorance, and thus the more difficult the particular game. Ever tried picking a pipe cleaner? How about a specific number in the thousands? How about Twenty Questions itself?

The questioner has at her or his disposal a variety of strategies for posing game questions, consisting generally of yes or no interrogatives. Both the questioner and answerer often step outside the context of the game proper to discuss a variety of issues, such as keeping count of the amount of questions posed thus far, requesting or offering help, expressing frustration or surprise, and asking for clarifications of questions or answers. The latter is crucial for monitoring misunderstanding, given the often ambiguous nature of the brief remarks of which the game proper consists. Finding the right questions, or the most effective questions for discovering the answer, requires both the establishment of a shared horizon of background knowledge and the accrual of a stock of methodical questions that always come in handy in the process of dividing and conquering. The game context itself also plays a critical role in the interpretation of particular utterances. While it is difficult to systematically determine the intentions of utterances in everyday conversations, linguists and discourse analysts have employed Twenty Questions as a context within which those intentions are relatively transparent, and thus more susceptible to accurate modeling. While a variety of utterances have relevance to the game, it is usually clear what move in the game an utterance constitutes.

In 1946, a game show version of Twenty Questions aired on the radio in the United States, appearing on television in 1949 and lasting for six seasons while on both radio and television. The cast consisted primarily of members and friends of the VanDeventer family, sometimes supported by celebrity guest panelists. Fred VanDeventer, a newscaster for New York's WOR, and his family had perfected the deductive process of eliminating possibilities and narrowing the field down to the correct answer, usually in seven questions or less. On one occasion, Robert VanDeventer (aka Bobby McGuire) got the mystery item (Brooklyn) without asking a single question. The game show was exported to the United Kingdom and Canada, as well as the Netherlands, where a version of it still airs as of today.

A recent key episode in the history of Twenty Questions is its union with artificial intelligence in the 20Q phenomena. 20Q is the name of a burgeoning business based on a patented form of artificial intelligence, boasting more than 10 million synaptic connections. Robin [Page 736]Burgener is the inventor, programmer, and architect of the key software components of the 20Q AI knowl-edgebase and game engine. A Web site, http://20Q.net, is the host of 20Q and learns by generating new information upon the conclusion of every game—for instance, noting contradictions between what it previously had stored on a particular subject and what it learned from a particular player's or set of player's opinions on the subject. Because of the interpretive nature of Twenty Questions, the knowledge base accrued by 20Q is more of a folk taxonomy generated from common knowledge and everyday speech rather than a scientific taxonomy of precise correspondence between words and objects. The more people and games 20Q plays, the more intelligent it becomes.

...