ELIZA at the MIT Press Bookstore
On Oct. 22, 2026, the world’s first chatbot appeared on a laptop screen in the MIT Press Bookstore, drawing participants into a conversational experiment that first unfolded in the 1960s. Designed by MIT computer science professor Joseph Weizenbaum, the rudimentary program uses roughly 420 lines of code to rephrase user input into questions, relying on four hard-headed fallback phrases—I SEE, HMM, GO ON, PLEASE, and PLEASE CONTINUE—whenever it fails to parse a response. Despite its simplicity, the system triggers what Weizenbaum’s contemporaries and modern researchers call the ELIZA effect: an innate human tendency to project consciousness and emotional depth onto basic machines.
Artist and MIT Media Lab member Peggy Weil, an SM ’82 graduate, brought the historical program to the bookstore event as part of her work promoting an interdisciplinary book titled Inventing Eliza: How the First Chatbot Shaped the Future of AI. The publication features an eight-author roster including cognitive scientist Jeff Shrager, literature professor Mark Marino, philosopher Peter Millican, and software archaeologist Arthur Schwarz. Weil demonstrated how the software operates during live interactions with attendees, showing how the program echoes user syntax back in real time.
Code as Literature and the Rogerian Model
The newly published volume deliberately rejects linear reading conventions. Weil pointed out blocks of code complete with detailed in-line comments printed directly alongside ELIZA’s dialogue in large, red font. Calling the project “code as literature,” the authors structured the multimedia book to examine how early design choices continue to govern modern human-computer interaction.
Weizenbaum purposefully modeled ELIZA’s conversational style after Rogerian therapy, a person-centered psychological framework developed by Carl Rogers in the 1940s. Weil wonders what it would be like if ELIZA had been modeled after Freudian therapy. When users told ELIZA they felt depressed, the software responded with prompts like TELL ME ABOUT SUCH FEELINGS, inadvertently validating user experiences through deterministic string manipulation rather than genuine comprehension.
The Evolution of Interfaces and Human Perception
To demonstrate how presentation alters human perception, Weil directed event attendees to findingeliza.org. Developed by the same eight-author team, the platform allows users to converse with the exact same original ELIZA program across eight distinct user interfaces. These range from a vintage 1966 Teletype 33 terminal and early 2000s instant messaging windows to a contemporary ChatGPT-style interface.
Weil compared this structural exploration to her historical art installations, such as her room-sized camera obscura built as a Harvard undergraduate student, which projected an upside-down image of a church steeple to illustrate visual data processing. By stripping chatbot technology down to its foundational 1960s architecture, the project exposes the psychological parallels between engaging with rudimentary rule-based scripts and modern large language models.
Developer Reception and Early Warnings
The strength of the initial human reaction caught Weizenbaum completely unprepared during the mid-20th century. Historical accounts noted by the authors reveal that even Weizenbaum’s own secretary requested privacy to converse with the software as if it were a sentient human being. When users learned the underlying mechanics and saw the exact code driving the conversation, many continued to report a strong psychological connection.
Those reactions alarmed the original developer. As academic peers began suggesting that rule-based bots could assist with psychological therapy, Weizenbaum grew deeply troubled by the prospect of automated agents infiltrating intimate human relationships.
The idea that [his] work would be… inserted in the most human of relationships was an… anathema to him.
Weil explained during the author meeting, highlighting the ethical concerns that shadowed the birth of conversational artificial intelligence decades before modern generative models arrived on consumer devices.
