A classic Daniel Dennett article considers a curious chapter in AI history, where researcher Kenneth Colby used the Turing Test to see whether psychiatrists could distinguish between delusional patients and his natural language paranoia simulator ‘PARRY‘.
PARRY was designed by Colby, who was both a psychiatrist and computer scientist, in an attempt to simulate the psychology of paranoia. In particular, the programme was designed to replicate paranoid delusions about being persecuted by the Mafia.
Dennett’s 1990 article, entitled “Can machines think?”, discusses whether the Turing Test is an adequate test of machine intelligence.
Dennett notes that PARRY is the only programme known to have passed the Turing Test – psychiatrists were unable to distinguish between real patients and simulated ones.
Ironically, PARRY was based on the ELIZA programme, which was designed as a text-only parody of a therapist.
Colby’s programme was the first attempt to produce a computer simulation of psychosis, a project which now typically involves artificial neural network simulations of information processing models of the mind, rather than conversational interaction.
Link to article “Can Machines Think?” (seems to have a few scanning errors).
PDF of same article.
Link to article “Simulating Psychosis”.