In Plain Language
ELIZA was a program that chatted like a therapist, but it did not understand a word. It just watched for patterns. Say "I am sad" and it flips it round into "How long have you been sad?". That is the whole trick. People still felt it understood them, and some did not want to stop talking to it.
ELIZA matched what you typed against a set of templates and reflected it back as a question. "I am X" became "How long have you been X?"; anything it could not match got a stock prompt like "Tell me more". There was no memory, no model of the conversation, and no representation of meaning — the Rogerian-therapist persona was chosen precisely because a therapist who mostly echoes you back is easy to imitate. What Weizenbaum did not anticipate was how readily people confided in it, even after he explained the mechanism. That gap between how a system works and how convincing it feels is now called the ELIZA effect, and it resurfaces every time a language model sounds thoughtful.
Why It Was Important
Created by Joseph Weizenbaum, ELIZA simulated conversation by using pattern matching and substitution methodology, most famously imitating a Rogerian psychotherapist. Despite having no built-in understanding, users often ascribed deep human emotion to it—a phenomenon now known as the 'ELIZA effect'.
Who Invented It
Joseph Weizenbaum
Computer scientist at MIT; later became a prominent critic of over-reliance on AI.
Applications
- Chatbots
- Human-Computer Interaction
- Pattern Matching
Key Papers
- ELIZA — A Computer Program for the Study of Natural Language Communication Between Man and Machine
Joseph Weizenbaum · Communications of the ACM · 1966