liveBook · Manning

ELIZA

As explained in Understanding AI

ELIZA is an early natural-language conversation program that generated responses primarily through predefined pattern-matching rules.

ELIZA can sound thoughtful without understanding anything. It detects keywords, matches familiar sentence patterns, and places extracted phrases into predefined response templates. section 1.1

Its response process is simple:

  • Scan the input for keywords such as “feel,” “my,” or “think.” section 1.1
  • Match a structure such as “I feel X” or “I think X.” section 1.1
  • Reuse part of the sentence inside a fixed question template. section 1.1

A convincing but empty exchange

User: I feel unhappy about my job ELIZA: Why do you feel unhappy about your job? The reply sounds relevant because it mirrors the user’s words and resembles a therapist’s open-ended question. Yet it does not interpret the situation, recognize emotion, or form an opinion. When language falls outside anticipated patterns, the response may remain grammatical while missing what matters. section 1.1 section 1.1.1

Figure 1.1 The general mechanism of rule-based systems. A fixed rule library written in advance by humans defines the set of triggers, conditions, and actions the system can recognize. When input arrives, the rule-based system scans for a matching condition and fires the associated action. No learning occurs: the system can only handle situations its designers anticipated. Anything outside the ruleset causes failure or a degraded response. ELIZA is a direct instance of this architecture.

Trace ELIZA as a fixed rule-based pipeline: input triggers, matching conditions, and predefined actions, with no learning beyond the rules designers supplied. section 1.1.1

How Understanding AI covers it

How deep it goes
Explains keywords, pattern matching, templates, and edge-case failure, stopping short of implementation code or learning algorithms.
Examples
Uses therapist-style dialogue, keyword-triggered questions, and an emotionally serious statement that receives a mechanical reply.
What you can do after
Evaluate/understand why fluent responses can result from predefined patterns rather than interpretation or emotion.
How it connects
Connects ELIZA to pattern matching, rule-based AI, modern AI systems, and the Turing Test.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Understanding AI. The section links point to the parts of the book it draws on.