Project: ELIZA

In the mid-1960s, Joseph Weizenbaum at MIT wrote ELIZA, a ground-breaking program in artificial intelligence that allowed users to converse with it using plain English. ELIZA was meant to sound like a Rogerian therapist: friendly, non-judgmental, and reflective, using questions to get the user to talk further about him or herself. By searching for certain keywords, ELIZA could provide very focused responses, in addition to using a variety of generic statements and questions.

The web has many examples of similar programs now, also known as chatterbots.

Exercises

1. Research the Turing Test. Report your findings, including its connection with ELIZA.
2. Write a program that implements your own idea for an ELIZA. In other words, try to develop a specific personality, in the way that Weizenbaum tried to emulate a therapist.

   Use spiral development (see page 107). Begin with a very simple working program and gradually add features like these:

   • Find out the user’s name and use it occasionally.
   • Use keywords to provide appropriate responses.
   • Have a supply of random responses to fall back on if none of the keywords match.
   • Look for likely types of responses or questions.