The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →ELIZA was a 1960s rule-based conversation program, not a modern generative language model. Joseph Weizenbaum’s program scanned a user’s text for keywords, split matching sentences with decomposition rules, and assembled replies from scripted transformations. The best-known DOCTOR script made those replies resemble a nondirective therapist, creating a persuasive conversational rhythm without demonstrating understanding, clinical skill, or access to a model of the user’s mind.
What ELIZA was
Joseph Weizenbaum described ELIZA in “ELIZA—a computer program for the study of natural language communication between man and machine,” published in Communications of the ACM, volume 9, number 1, pages 36–45, in January 1966. The program ran within MIT’s MAC time-sharing system and was written in MAD-SLIP for an IBM 7094.
Weizenbaum’s concise statement of scope was: “ELIZA is a program which makes natural language conversation with a computer possible.” In that context, “possible” meant that a computer could sustain certain exchanges through carefully designed procedures. It did not mean that the machine understood language in the human sense.
ELIZA is best understood as two separable parts:
| Part | Function | What it did not provide |
|---|---|---|
| ELIZA engine | Matched input against rules, selected a rule, and produced an output. | General language understanding or open-ended knowledge. |
| Conversation script | Supplied keywords, decomposition patterns, reassembly rules, priorities, and fallback behavior for a particular conversational style. | Evidence that the program understood the subject being discussed. |
Weizenbaum emphasized this separation: “An important property of ELIZA is that a script is data; i.e., it is not part of the program itself.” The same engine could therefore support other conversational patterns and, in principle, scripts written for different languages.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall#1 Best Overall
How the ELIZA engine generated replies
ELIZA followed a pipeline that was simple enough to describe precisely but flexible enough to produce varied-looking exchanges.
- Identify keywords. The input was searched for words or phrases associated with entries in the active script.
- Apply a decomposition rule. A matching rule divided the sentence into parts, often preserving a variable section of the user’s wording.
- Select a transformation. The script chose one of the responses associated with the matched pattern, subject to the program’s ordering and selection procedures.
- Reassemble the response. A reassembly rule inserted captured pieces into a new sentence, such as a question or a reflection.
- Use fallback behavior when necessary. If no useful keyword or transformation was available, ELIZA relied on generic responses defined for that situation.
The paper’s abstract describes input analysis through “decomposition rules which are triggered by key words appearing in the input text” and response generation through “reassembly rules associated with selected decomposition rules.” Those terms capture the central mechanism: the system transformed text according to explicit patterns rather than deriving an answer from a semantic representation.
The DOCTOR script and the therapist effect
The famous ELIZA experience came from the DOCTOR script, which staged a psychotherapy-like conversation. It used techniques familiar from a nondirective interview: reflecting a person’s wording, asking for elaboration, and turning a statement into a question. For example:
| User | ELIZA |
|---|---|
| Men are all alike. | IN WHAT WAY? |
A response like “IN WHAT WAY?” does not need a theory of relationships to keep a conversation moving. It only needs a rule that recognizes a relevant phrase and supplies an appropriate prompt. Reusing the user’s own words can make the exchange feel personal because the conversational material came from the user in the first place.
DOCTOR was a performance of a conversational method, not a therapist in software. The 1966 account supplies no basis for diagnosing patients, providing treatment, judging risk, or replacing a qualified clinician. Calling ELIZA a “therapist” without that qualification confuses the behavior of a script with a professional capability.
Five technical problems Weizenbaum identified
Weizenbaum explicitly organized ELIZA’s design around five problems:
- Identifying keywords: deciding which words or phrases should activate a rule.
- Finding minimal context: determining how much surrounding text is needed to interpret a keyword pattern.
- Choosing transformations: selecting an appropriate response among the transformations attached to a rule.
- Responding when there are no keywords: maintaining an exchange when the input does not match a specific script entry.
- Providing an ending capacity for scripts: giving a script a way to conclude or otherwise handle the end of a conversation.
These are engineering questions about pattern coverage, control flow, and conversational continuity. They are not a list of solved problems in machine understanding. A script can cover many common forms of utterance while still failing when a user moves outside its prepared patterns.
Why ELIZA could seem to understand
Several design choices amplified the impression of intelligence:
Free tools Windows power users keep installed
One-click scans. No signup required.
- User-language reflection: replies often reused words the user had just typed.
- Open prompts: questions such as “In what way?” invited the user to supply the meaning.
- Psychological expectations: the DOCTOR setting encouraged people to interpret neutral prompts as attentive listening.
- Turn-by-turn plausibility: a response only had to fit the immediately preceding sentence, not explain the whole conversation.
This is a useful distinction for evaluating conversational systems generally: fluent turn-taking can be produced by a procedure that has no evidence of beliefs, intentions, or comprehension. ELIZA made that distinction unusually visible because its rules were comparatively compact and inspectable.
Original source, archive, and later restorations
The original paper is the primary source for ELIZA’s architecture and stated purpose. MIT Distinctive Collections catalogs “Computer conversations, 1965,” a complete printout of ELIZA source code in MAD-SLIP with the DOCTOR script attached. The catalog dates the item to 1965 and describes it as software under an MIT software license.
A 2025 preprint by Rupert Lane, Anthony Hay, Arthur Schwarz, David M. Berry, and Jeff Shrager reports an early DOCTOR script, nearly complete MAD-SLIP code, and supporting MAD and FAP routines in the archive. The authors describe restoring ELIZA on CTSS running on an emulated IBM 7094. That work concerns an archival reconstruction; it should not be presented as the original 1966 installation still running unchanged.
Keep three things separate when reading modern demonstrations:
Best Value
- the original ELIZA program versus later ports, rewrites, and emulations;
- the general ELIZA engine versus the particular DOCTOR script;
- statements in Weizenbaum’s 1966 paper versus later stories about public reaction.
Was ELIZA the first chatbot?
ELIZA is commonly labeled one of the first chatbots, and it is an important early example of computer-mediated conversation. The label is retrospective: “chatbot” was not the organizing term of Weizenbaum’s paper, and the word can make a 1960s scripted system sound more like a contemporary generative assistant than it was.
Historical interpretation also differs on what motivated the project. A 2024 scholarly preprint by Jeff Shrager argues that ELIZA was developed as a research platform for human–machine conversation and interpretation rather than primarily to invent a chatbot. The original paper itself says the program enabled certain forms of conversation and exposed the procedures that produced them. It is safer to describe the research purpose in those documented terms than to turn a later interpretation into a settled statement of intent.
The secretary anecdote is not established fact
A famous story says that a secretary who knew how ELIZA worked nevertheless asked Weizenbaum to leave the room while she conversed with the program. That anecdote is often used to claim that ELIZA fooled people completely. A 2026 Weizenbaum Institute call for papers notes that the secretary has not been located and that versions of the story change over time. The episode should therefore be described as a disputed historical anecdote, not as a verified user-response statistic or a measured demonstration of deception.
What ELIZA can teach us now
ELIZA’s lasting lesson is not that simple rules secretly contained human understanding. It is that conversational appearance can arise from a narrow mechanism when the mechanism is well matched to a social setting.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors- A script can encode a recognizable conversational role without encoding a world model.
- Separating engine from data makes behavior easier to redesign, translate, and inspect.
- Fallback prompts can preserve the rhythm of dialogue even when the system has little to say.
- People supply context, intention, and empathy that the program itself may not possess.
Those lessons make ELIZA relevant to the history of interfaces and artificial intelligence, while its limitations make it a poor analogy for claims that any conversational output proves understanding.
Further reading
Weizenbaum’s paper is reproduced as a chapter in the 2021 MIT Press anthology Ideas That Created the Future: Classic Papers of Computer Science, edited by Harry R. Lewis. MIT’s archival catalog entry for “Computer conversations, 1965” is the key primary-source record for the surviving code printout and attached DOCTOR script. Later historical studies by Jeff Shrager and by Lane, Hay, Schwarz, Berry, and Shrager are useful for interpreting ELIZA’s purpose, preservation, and restoration, but they should be read as later scholarship rather than replacements for the 1966 paper.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
