Weizenbaum's ELIZA imitates a therapist
Two hundred lines of pattern-matching code convince people they are being understood
Quick facts
- Author
- Joseph Weizenbaum
- Institution
- MIT
- Published in
- Communications of the ACM, Vol. 9, No. 1
What happened
Joseph Weizenbaum at MIT published 'ELIZA, A Computer Program for the Study of Natural Language Communication Between Man and Machine' in Communications of the ACM. ELIZA worked by scanning a user's typed sentence for a keyword, applying a scripted transformation rule tied to that keyword, and printing the result back as a question, in the manner of a Rogerian psychotherapist reflecting a patient's own words. If no keyword matched, it fell back on a content-free prompt like 'please go on.' There was no understanding of meaning anywhere in the program, only string matching and substitution. Weizenbaum became alarmed when he found that people, including his own secretary, treated their sessions with ELIZA as genuinely private and confided in it as they would a real therapist, and asked him to leave the room.
Why it matters
ELIZA showed that people readily attribute understanding to a machine that merely reflects their own words back, a phenomenon now called the Eliza effect that remains directly relevant to how people relate to modern chatbots.
How we know
Weizenbaum's original 1966 paper includes the pattern-matching mechanism and a full sample dialogue transcript; IEEE Spectrum's account of the secretary incident is corroborated by Weizenbaum's own later writing on the episode.
Sources
- Joseph Weizenbaum, Communications of the ACM. ELIZA, A Computer Program for the Study of Natural Language Communication Between Man and Machine · Primary source (author-declared)cse.buffalo.edu · Cited as a "primary" source (no stronger domain match).
- IEEE Spectrum. Why People Demanded Privacy to Confide in the World's First Chatbot · Peer-reviewedspectrum.ieee.org · The domain "spectrum.ieee.org" is on our Peer-reviewed registry. · Link is live and its text matches the event's key terms (Jul 2026)
See something wrong? . Corrections with a source get fixed fastest.
Part of a timelineArtificial Intelligence30 events · From a wartime theory of neurons to machines that write, paint, and fold proteinsView all →