DOCTOR mode runs Weizenbaum’s original 1966 script, word for word, on a faithful port of the ELIZA algorithm. Try the famous first line, “Men are all alike.” ELIZA is not a real therapist and should not be used for mental health advice.
ELIZA is typing...
Try these prompts
ELIZA uses simple rules: it scans for words and sentence patterns, then reflects parts of your sentence back using a scripted template.
Show details
When a rule matches (for example, "I feel..."), ELIZA applies a fixed response template and may swap pronouns like "my" to "your." If no rule matches, it uses a fallback response. This is scripted pattern matching, not modern AI reasoning.
What ELIZA Is — and Is Not
Historical Context
ELIZA was created between 1964 and 1966 at MIT.
It was developed by Joseph Weizenbaum.
The most famous script, DOCTOR, simulated Rogerian psychotherapy prompts.
The DOCTOR script used psychotherapy patterns because reflective conversation is easy to model with simple keyword rules.
What ELIZA Is
ELIZA matches keywords and sentence patterns.
It substitutes and reflects phrases using predefined rules.
Its only “memory” is a short queue: when you say “My …”, it may save a line like “EARLIER YOU SAID YOUR …” and replay it later when nothing else matches.
What ELIZA Is Not
ELIZA does not understand language in a human sense.
ELIZA is not trained on a dataset.
ELIZA does not think.
ELIZA is not a licensed therapist and should not be used for mental health advice.
Contrast With Modern AI
Modern large language models use statistical pattern learning.
They are trained on very large datasets.
They generate probabilistic responses rather than fixed scripts.
How to Talk to ELIZA (1966 style)
Use simple declarative sentences.
Avoid slang.
Avoid abstract philosophical questions.
Speak as if interacting with a psychotherapy terminal.
"I feel anxious." "My father upsets me." "I am worried about school."
History Panel: ELIZA in AI Timeline
ELIZA occupies a foundational moment in natural language processing history and public perception of machine conversation.
Timeline
1950: Alan Turing proposes the imitation game.
1957: Perceptron research begins.
1964-1966: ELIZA is built at MIT; DOCTOR script is demonstrated in 1966.
1970s: Expert systems expand symbolic AI.
2010s-present: Large-scale statistical and neural language models become dominant.
Worked Example: Tracing ELIZA's Decomposition and Reassembly
Walk through Joseph Weizenbaum's 1966 pattern-matching algorithm on a real human input sentence to understand why the illusion of comprehension occurs.
Input Sentence: “I feel that my computer dislikes me.”
Step 1: Keyword Scanning and Priority Ranking. The script tokenizes the input and checks its keyword table. Three keywords match: FEEL (precedence rank 4), MY (precedence rank 2), and COMPUTER (precedence rank 50). Because COMPUTER possesses the highest numerical rank (50), the algorithm selects the COMPUTER rule set and ignores the others.
Step 2: Decomposition Rule Matching. The rule template for COMPUTER is (0 COMPUTER 0), where 0 matches any sequence of arbitrary words. The parser segments the sentence into components: prefix \((s_1 = \text{"I feel that"})\), keyword \((k = \text{"computer"})\), and suffix \((s_2 = \text{"dislikes me"})\).
Step 3: Pronoun Transposition Table. The suffix words are inverted using the reflection dictionary: “me” transposes to “you”, “my” transposes to “your”, and “I” transposes to “you”. The suffix becomes “dislikes you”.
Step 4: Reassembly Rule Selection. The script cycles through the reassembly options for COMPUTER: “Do computers worry you?”, “What do you think machines have to do with your problem?”. Choosing the second reassembly template inserts the transformed suffix, producing: “Do computers worry you, or what do you think machines have to do with your problem?” Because the reply accurately echoes the user's specific noun (“computer”) without comprehending its semantic meaning, the user mistakenly infers genuine empathy.
Self-Check Question
If a user enters “I am very unhappy today”, which keyword pattern will ELIZA trigger, and how will “I am” be transformed?
Show Answer
Answer: The rule triggers on keyword I AM (or I). The transposition table inverts “I am” into “YOU ARE” (or “YOU”), resulting in the classic DOCTOR response: “Did you come to me because you are very unhappy today?”
Sources & Further Reading
Primary literature on natural language processing, the ELIZA program, and the philosophy of artificial intelligence.
Weizenbaum, Joseph. “ELIZA—a computer program for the study of natural language communication between man and machine.” Communications of the ACM, 9(1), 36–45 (1966). doi:10.1145/365153.365168.
Weizenbaum, Joseph.Computer Power and Human Reason: From Judgment to Calculation. W. H. Freeman & Co., San Francisco (1976). Weizenbaum's critique of over-attributing human intelligence to computational systems.
Turing, Alan M. “Computing Machinery and Intelligence.” Mind, 59(236), 433–460 (1950). The seminal philosophical paper proposing the imitation game (Turing test).