conversational character

Depending on their purpose, and what we expect of them, chatterbots are equipped with a more or less complex and elaborate artifi ficial intelligence (AI) (see artificial intelligence). Automated assistants and computer game characters are usually expected to operate within a limited knowledge area. For these to function satisfactorily, it may be suffi fficient that they know how to identify and match key words in the question with a predefined fi answer in their database. However, to fluently converse on a number of nonspecifi fied topics— as is required to pass the Turing test— a more sophisticated AI based in natural language processing may be needed. Some of today’s chatterbots are even designed to learn from their previous conversations—in other words, developing their AI as they speak.

A relative of ELIZA named PARRY demonstrates how the opposite principle may also be employed to create a convincing chatterbot. Written by Stanford psychiatrist Kenneth Colby in 1972, PARRY is a program that simulates a patient suffering ff from paranoia. When Parry does not know how to answer, he will aggressively spit out a line of conspiracy theory, thereby forcing his world onto ours. While totally out of context, his response is nonetheless plausible. A similar trick of the trade is used in artist Ken Feingold’s robotic AI installation “Head” (1999), modeling a poetically inclined, slightly disturbed and confused elderly man. Engaging in conversation with Head d requires a signifi ficant share of interpretative effort, ff but it may also be greatly rewarding to those who are willing to invest in it. Another artistcreated chatterbot is Stelarc’s Prosthetic Head , which simulates the artist’s own personality. As it is designed to learn from its conversations, however, this chatterbot may gradually become “more autonomous in its responses” until the artist will “no longer be able to take full responsibility for what his head says” (Stelarc 2003). While most chatterbots are designed to engage in conversation with human partners, artistic experiments have been made in which chatterbots converse with each other. The result may be hilarious, as in director Annie Dorsen’s theatrical per for mance “Hello Hi There” (2010). Here, Dorsen stages two chatterbots self-reflexively fl discussing the famous 1971 debate between philosophers Michel Foucault and Noam Chomsky, on the concept of human nature.

(Johns Hopkins University Press)

This presentation reconsiders one of the most famous works of electronic literature: Joseph Weizenbaum'sEliza/Doctor. Created in the mid-1960s, this conversational character's success led Janet Murray to name Weizenbaum "perhaps the premier" literary artist in the computer medium. Such evaluations, however, don't take into account what happens during the playful engagement that the system's freeform textual interaction encourages: a breakdown that reveals the shape of the underlying processes. An alternative to this is extremely constrained interaction, which can help maintain the illusion. But a more exciting direction is to design processes that reward readers as they are revealed.

What program could have the effect on today's popular consciousness that Joseph Weizenbaum's Eliza had in the mid-1960s? Eliza ignited numerous productive controversies about language, intelligence, and people's relationships to computers. The system has been hailed as the first and most important work of electronic literature. While other, more complex works have been innovative, challenging, and literary in ways that are perhaps more sophisticated, Eliza was an incisive program of great impact. We consider the provocative program within the contexts of computing from the 1960s to the present. Then, we identify several qualities, some of them not very obvious, that a similarly provocative literary program would need today.

Harger's interview with Mateas and Stern focuses on the development of theiir conception of interactive drama and the Façade project.

Full title: "ELIZA — A Computer Program For the Study of Natural Language Communication Between Man And Machine"

ELIZA is a program operating within the MAC time-sharing system at MIT which makes certain kinds of natural language conversation between man and computer possible. Input sentences are analyzed on the basis of decomposition rules which are triggered by key words appearing in the input text. Responses are generated by reassembly rules associated with selected decomposition rules. The fundamental technical problems with which ELIZA is concerned are: (1) the identification of key words, (2) the discovery of minimal context, (3) the choice of appropriate transformations, (4) generation of responses in the absence of key words, and (5) the provision of an editing capability for ELIZA "scripts". A discussion of some psychological issues relevant to the ELIZA approach as well as of future developments concludes the paper.

(Source: Author's abstract)