PARRY
In plain wordsAn early-1970s chatbot built to talk like a psychiatric patient with paranoid schizophrenia. In tests, psychiatrists reading its conversations did little better than chance at telling it from real patients — an early Turing-test-style result.
A program, written in the early 1970s by the Stanford University psychiatrist Kenneth Colby, that conversed in the role of a patient with paranoid schizophrenia. Where ELIZAELIZA, prop. n.: A 1960s computer program that chatted by turning a user's own words back into questions, like a therapist. Many users felt understood, though the program understood nothing. only reflected its input, PARRY tracked levels of fear, anger and mistrust that rose or fell with what it was told, and it kept pulling conversations back to fixed delusions about bookies and the Mafia, drawing on a much larger stock of language patterns. When psychiatrists were given interview transcripts from PARRY and from real patients, without being told which was which, they picked out the program at about chance level — an early, narrow variant of the Turing testTuring test, n.: Alan Turing's proposed test of machine intelligence: a judge holds typed conversations with a hidden person and a hidden computer, and the computer passes if the judge cannot consistently pick it out. — though critics noted that evasive or incoherent answers are expected of the patients being imitated. In 1972 PARRY and ELIZA's DOCTOR script were connected over the ARPANET for a much-reproduced machine-to-machine 'therapy session'.
PARRY passed a narrow Turing test partly because paranoid evasiveness hid its limitations.
Origin
Developed by Colby with collaborators including Sylvia Weber and Franklin Hilf as a working model for testing theories of paranoia, and described in their paper 'Artificial Paranoia' (Artificial Intelligence, 1971) and in later reports on what they called Turing-like indistinguishability tests.
Related
- ELIZAELIZA, prop. n.: A 1960s computer program that chatted by turning a user's own words back into questions, like a therapist. Many users felt understood, though the program understood nothing.
- Turing testTuring test, n.: Alan Turing's proposed test of machine intelligence: a judge holds typed conversations with a hidden person and a hidden computer, and the computer passes if the judge cannot consistently pick it out.
- chatbotchatbot, n.: Software you chat with in ordinary language, by typing or talking, as you would with a person. The newest ones, such as ChatGPT, can answer questions, draft writing and keep long conversations going.
- ELIZA effectELIZA effect, n.: Our tendency to believe that a computer understands us, cares or thinks just because it produces humanlike words, even when it is only following simple rules.
- Loebner PrizeLoebner Prize, prop. n.: A yearly contest, held from 1991 to 2019, in which chatbots tried to convince judges they were human in short text chats. Its grand prize was never won, and critics called it a publicity stunt.
Sources
- Colby, K. M., Weber, S. & Hilf, F. D. (1971). Artificial Paranoia. Artificial Intelligence 2(1), 1–25.