ELIZA’s Code Reveals Early Anxieties About AI

Researchers recovered the original source code for ELIZA, the 1966 chatbot that simulated a psychotherapist, revealing a more complex history than previously understood. The program, created by MIT professor Joseph Weizenbaum, wasn’t intended to demonstrate intelligence but to explore why people readily attribute it to machines.

The team’s analysis, detailed in their new book Inventing ELIZA, shows the program utilized multiple versions and technical innovations, challenging the common narrative of a simple, early chatbot. Weizenbaum himself was surprised by the emotional attachments people formed with ELIZA, warning of the potential for dehumanization as people treat computers as if they possess understanding.

This phenomenon, later termed the “ELIZA effect,” describes the tendency to project human qualities onto even rudimentary computer programs. The investigation into ELIZA’s code highlights its connection to Alan Turing’s work on artificial intelligence and gender performance, and demonstrates how early design choices continue to shape contemporary AI systems, including large language models like ChatGPT. The researchers argue understanding ELIZA’s origins is essential to addressing the ethical and social implications of today’s AI technologies.