The First Chatbot’s Multiple Personalities

2026-07-15 · Source: IEEE Spectrum · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, long

Summary

The recently published book "Inventing ELIZA: How the First Chatbot Shaped the Future of AI" by MIT Press, based on unearthed source code from MIT's archives, reveals that ELIZA was a far more complex program than previously understood. While widely known for its "Doctor" psychotherapist persona, analysis of the original 1960s code demonstrates ELIZA's sophisticated platform capabilities, including script editing and contextual memory, allowing it to assume multiple personas. Beyond the famous Doctor script, researchers found ELIZA was programmed for roles discussing math, poetry, and even acting as a Socratic tutor (e.g., Intrvw, Canvec, FVP1 scripts) or a calculator (Arithm script). This discovery challenges assumptions about early AI development, highlighting Joseph Weizenbaum's advanced technical innovations and the system's architectural distinction between the ELIZA system and its various scripts, which presaged modern software patterns like plug-in architectures.

Key takeaway

For AI Scientists and Research Scientists developing conversational agents, understanding ELIZA's original architecture offers valuable lessons. You should consider separating core system logic from persona-defining scripts, as ELIZA's system-script dichotomy enabled diverse roles from therapist to tutor. This approach allows for flexible, context-aware bot design, demonstrating how linguistic choices and conversational flow construct convincing, adaptable personas even within technical constraints.

Key insights

ELIZA's unearthed code reveals it was a sophisticated, multi-persona platform, not just a simple pattern-matching chatbot.

Principles

• Chatbot personas emerge from script language and conversational approach.
• System-script dichotomy enables flexible, multi-role AI architectures.
• Constraints shape design, but elegant solutions can still emerge.

Method

ELIZA created personas by using structured data scripts that directed the system to "play" a specific task or role, with later versions incorporating conditional keyword matching for dynamic responses.

In practice

• Design chatbot scripts to evoke specific rhetorical tones and characterizations.
• Separate core system logic from persona-defining scripts for flexibility.
• Use Socratic methods in tutoring bots via conditional keyword matching.

Topics

• ELIZA Chatbot
• AI History
• Human-Computer Interaction
• Chatbot Personas
• Scripting Languages
• Conversational AI

Best for: AI Scientist, Research Scientist, AI Student

Related on AIssential

• The Chatbot That Foretold Why People Share Secrets With ChatGPT — The "ELIZA effect" highlights humans' persistent tendency to project intelligence onto responsive computer programs, a critical lesson for modern AI.
• PersonaKit (PK): A Plug-and-Play Platform for User Testing Diverse Roles in Simulated Full-Duplex Dialogue — Full-duplex dialogue systems need flexible, persona-specific turn-taking beyond rigid "always-yield" policies for immersive interactions.
• Beyond one-on-one: Authoring, simulating, and testing dynamic human-AI group conversations — DialogLab bridges scripted and generative AI dialogue for complex multi-party human-AI conversations.
• AutoPersonas: A Multi-Timescale Loop Engine for Open-Ended Persona Evolution — Separating divergence from evidence-governed absorption prevents persona self-locking while preserving identity continuity.
• Penny: Transition Network Analysis of Learner-Chatbot Interactions in Scaffolded EFL Writing — Learner-chatbot interaction dynamics in EFL writing are non-linear and proficiency-dependent, requiring differentiated AI design.
• Apple’s new Siri AI knows when to shut up — Apple's Siri AI adopts a concise, utility-focused personality, contrasting with other chatbots' verbose and engaging styles to avoid user over-attachment.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by IEEE Spectrum.