ELIZAELIZA

Between 1964 and 1966 at MIT, Joseph Weizenbaum built a program that could hold a typed conversation. He named it ELIZA, after the character in Pygmalion who learns to imitate upper-class speech without acquiring the substance behind it. The name was a warning label that almost everyone ignored.

ELIZA's most famous script, DOCTOR, imitated a Rogerian psychotherapist, a persona chosen precisely because it could answer questions with questions. Say 'my mother hates me' and it might reply 'tell me more about your family'. Underneath was simple pattern matching and keyword substitution, with no understanding of any kind.

The trick worked far better than Weizenbaum intended. His own secretary, who had watched him build the program and knew exactly what it was, asked him to leave the room so she could talk to it privately. Some practicing psychiatrists suggested ELIZA-like systems could deliver automated therapy at scale.

Weizenbaum was horrified. He had set out to demonstrate the superficiality of machine conversation and instead discovered how eagerly people project understanding onto anything that responds to them. Researchers now call this the ELIZA effect, and it has only grown more relevant.

The experience turned Weizenbaum into one of the field's earliest and sharpest critics. His 1976 book 'Computer Power and Human Reason' argued that even if machines could make certain decisions, there are decisions they should not be allowed to make, drawing a line between calculation and judgment.

ELIZA remains the ancestor of every chatbot since, and its creator's discomfort remains the ancestor of every debate about them. Sixty years on, both the technique and the unease have scaled up together.