“Please tell me more about that.” – ELIZA
If you want to understand how language simulation worked before the AI boom, this is your starting point:
🔗 Try ELIZA now in your browser
This demo replicates Joseph Weizenbaum’s original 1966 program. It simulates a Rogerian psychotherapist and responds using simple pattern rules – no understanding, no memory, no intelligence.
Why ELIZA still matters
ELIZA’s success surprised even Weizenbaum. Many users felt understood by a program that merely mirrored their statements with generic replies.
This psychological effect became known as the ELIZA effect:
People tend to attribute intelligence and empathy to systems that follow simple, human-like interaction patterns.
As Andre Adrian notes:
Humans anthropomorphize almost everything – from lightning gods to Tamagotchis. We interpret even predictable machine behavior as socially meaningful, because our mirror neurons project our own thinking into the system.
This tendency is not new, not spooky – but essential to understand when evaluating AI.
Therapist logic without understanding
ELIZA’s logic mirrors Carl Rogers’ client-centered therapy:
- unconditional positive regard
- empathy
- congruence
But ELIZA achieves this through fixed rules and keyword-triggered phrases. If you say “My father is angry”, it might respond with “Tell me more about your family.”
This creates the illusion of connection – not through intelligence, but through structured reflection.
Weizenbaum himself warned:
“The impression of being understood is maintained by the user projecting reasoning and insight onto the machine. The machine does not possess these.”
Historical roots and legacy
- Original ELIZA was written in MAD-SLIP on an IBM 7094
- Later ports include BASIC versions (1970s), a Python CLI version, and this JavaScript demo
- ELIZA inspired follow-ups like PARRY, simulating paranoid schizophrenia
These early programs remain key reference points in the critical study of AI – especially when examining the boundaries between simulation and understanding.
Recommendation:
Use this demo for teaching, tech critique, or personal reflection. It’s a mirror of human psychology, not machine intelligence.