Overview
In December 1966, MIT professor Joseph Weizenbaum published a computer program called ELIZA in the journal Communications of the ACM. Running on a time-shared IBM 7094, ELIZA carried on conversations by matching user input against a set of pattern-matching rules — most famously mimicking a Rogerian psychotherapist by reflecting the patient’s statements back as questions.
It was the first program to enable a natural language conversation between a human and a machine. In some sessions, users spent hours talking to it. Some reportedly believed they were having a meaningful therapeutic dialogue. A few refused to believe it was “just a computer program.”
How It Worked
ELIZA’s core technique was pattern matching and substitution: when a user typed something, ELIZA would identify a keyword, apply a transformation rule, and respond with a scripted phrase. The most famous “DOCTOR” script turned statements into questions:
- User: “I am sad.”
- ELIZA: “Why do you say you are sad?”
No understanding was involved — only pattern-matching against a database of rules and responses. Yet the illusion of comprehension was powerful enough to fool many users, a phenomenon later called the “ELIZA effect” — the tendency to over-interpret machine outputs as meaningful.
Weizenbaum’s Surprise and Regret
Weizenbaum built ELIZA in a few weeks, partly as a demonstration of the superficiality of computer communication. He was genuinely surprised when his secretary — after observing the program — asked him to leave the room so she could speak with it privately.
In the late 1970s, Weizenbaum wrote Computer Power and Human Reason, partly to retract from the unintended consequences of his creation. He came to believe that assigning computer conversations the same status as human ones was a form of intellectual dishonesty.
Why It Matters
ELIZA was not taken seriously as science by the AI community — it was too simple. But its impact on the broader culture was profound:
- It introduced the “conversation as interface” concept decades before it became ubiquitous
- It generated the first serious public debate about human-computer deception and emotional attachment to machines
- It inspired every subsequent chatbot, including ALICE (1995), Jabberwacky (1997), and ultimately ChatGPT
- The ELIZA effect remains a foundational concept in AI psychology and human-computer interaction research