ELIZA
A 1960s chatbot that mimicked a psychotherapist by reflecting users' words back at them — and people projected so much understanding onto it that they got emotionally attached.
Think of it like
A very simple mirror: you say "I feel sad," it says "Why do you feel sad?" You go deeper because you filled in the blanks.
Example
Users would confess personal problems to ELIZA and get frustrated when the researcher asked if ELIZA had been helpful — they'd resent the implication that it was "just a program."
How it actually works
Joseph Weizenbaum's ELIZA was a pattern-matching system with about 100 lines of script. Its famous "Doctor" persona used trivial rules: if you mention "mother," ask a question about your family. Users consistently read emotional intelligence into it. This revealed that humans are incredibly quick to anthropomorphize: we assume understanding if a response is contextually plausible. The "ELIZA effect" is still relevant — people mistake fluent text from LLMs for genuine reasoning.
For product teams
Demonstrated that users attribute agency and understanding based on minimal behavioral signals.
For engineers
A pattern-matching chatbot with hand-written substitution rules and template responses; early proof-of-concept for text-based interaction.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome