What ELIZA showed

ELIZA did not show that machines understand people.

It showed that a simple program could produce some of the outward signs people associate with understanding.

The program was a little like an actor working from cue cards. A fitting line can feel sincere even when it came from a prepared rule.

We normally judge attention through behaviour. Someone answers at the right moment, remembers a detail, or reflects our words back to us. ELIZA could imitate some of those signs with rules for matching and rearranging text.

Reports of people responding to ELIZA belong to its history, but the original paper was a description of the program, not a controlled study of how users reacted.