[Paper Review] ELIZA Reinterpreted: The world's first chatbot was not intended as a chatbot at all

Jeff Shrager|arXiv (Cornell University)|Jun 25, 2024

AI in Service Interactions4 citations

TL;DR

This paper reinterprets ELIZA not as the world’s first chatbot, but as Joseph Weizenbaum’s original research platform for studying human-machine communication, interpretation, and the illusion of understanding. It argues that ELIZA was intended to investigate how people attribute credibility and intelligence to machines, not to simulate conversation, and that its accidental fame as a chatbot obscured its true purpose and led to the loss of its original design and intent for over 50 years.

ABSTRACT

ELIZA, often considered the world's first chatbot, was written by Joseph Weizenbaum in the early 1960s. Weizenbaum did not intend to invent the chatbot, but rather to build a platform for research into human-machine conversation and the important cognitive processes of interpretation and misinterpretation. His purpose was obscured by ELIZA's fame, resulting in large part from the fortuitous timing of it's creation, and it's escape into the wild. In this paper I provide a rich historical context for ELIZA's creation, demonstrating that ELIZA arose from the intersection of some of the central threads in the technical history of AI. I also briefly discuss how ELIZA escaped into the world, and how its accidental escape, along with several coincidental turns of the programming language screws, led both to the misapprehension that ELIZA was intended as a chatbot, and to the loss of the original ELIZA to history for over 50 years.

Motivation & Objective

• To reframe ELIZA’s original purpose, challenging the widespread belief that it was designed as a chatbot.
• To demonstrate that Weizenbaum intended ELIZA as a platform for studying human interpretive processes in response to machine-generated language.
• To trace how ELIZA’s accidental public emergence and technical evolution obscured its original research goals.
• To recover and reconstruct the historical context of ELIZA’s creation, including its roots in early AI, natural language processing, and symbolic computation.
• To highlight the long-term consequences of misinterpreting ELIZA as a conversational agent rather than a cognitive research instrument.

Proposed method

• Analyzing primary sources, including Weizenbaum’s 1966 CACM paper, unpublished manuscripts, and correspondence, to reconstruct the original intent behind ELIZA’s design.
• Tracing the technical lineage of ELIZA to early AI research, particularly the influence of Victor Yngve’s COMIT language and the development of SLIP for symbolic pattern matching.
• Examining the historical context of 1960s AI, including the Turing Test and the rise of symbolic AI, to position ELIZA within broader intellectual currents.
• Investigating how the DOCTOR script—ELIZA’s most famous variant—emerged independently and became the dominant public face of the program, despite not being Weizenbaum’s primary focus.
• Using archival materials and oral histories from key figures (e.g., Terry Winograd, John Markoff) to clarify the technical and philosophical motivations behind ELIZA’s architecture.
• Contrasting the original research goals of ELIZA with its later reception, especially the shift from a cognitive science tool to a cultural icon of AI.

Experimental results

Research questions

• RQ1 Why did Joseph Weizenbaum create ELIZA, and how does this differ from the commonly accepted narrative that it was intended as a chatbot?
• RQ2 How did the technical design of ELIZA—particularly its reliance on pattern matching in SLIP and COMIT—support its role as a platform for studying human interpretation of machine language?
• RQ3 What historical and technical factors led to the DOCTOR script becoming the dominant public manifestation of ELIZA, despite not being Weizenbaum’s primary focus?
• RQ4 How did the timing of ELIZA’s release and its unintended public adoption contribute to the misperception of its purpose?
• RQ5 What were the long-term consequences of ELIZA being misinterpreted as a conversational agent rather than a research instrument for studying human-machine credibility and interpretation?

Key findings

• ELIZA was not intended as a chatbot, but as a research platform to study how humans interpret and attribute understanding to machine-generated language, particularly in contexts of ambiguity and incomplete comprehension.
• Weizenbaum’s primary goal was to investigate the cognitive mechanisms behind human credibility judgments in response to machine output, not to create a convincing conversational agent.
• The DOCTOR script, which became the most famous version of ELIZA, emerged independently and was not part of Weizenbaum’s original research agenda, yet it became the de facto public face of the program.
• The original ELIZA source code was lost for over 50 years due to the accidental proliferation of the DOCTOR script and the lack of documentation, despite its central role in early AI research.
• Weizenbaum’s fear that the example would 'run away with what it is supposed to illustrate' was realized: ELIZA became a cultural symbol of AI intelligence, not a tool for studying human interpretive processes.
• The shift from ELIZA as a research instrument to a chatbot icon illustrates the broader danger of how easily people impute agency and understanding to simple pattern-matching systems, a concern that remains critically relevant with modern LLMs.