The Origins of Eliza: The Chatbot That Fooled the World

Ah, Eliza — the OG of chatbots, the digital therapist who made humans spill their guts to a few lines of code.

It was the 1960s, a time of big hair, moon landings, and… psychotherapist chatbots!
(I was born in 1970- so I missed the 60s by about 10-15 years…)

Psychotherapist chatbots?

Yep. Joseph Weizenbaum, a computer scientist at MIT, cooked up Eliza in 1966.

Named after Eliza Doolittle from Pygmalion, this chatbot wasn’t exactly sentient, but it sure knew how to fake it.

Eliza was designed to simulate a Rogerian psychotherapist — the kind of therapist who sits back and reflects your own thoughts right back at you.

Think of it as therapy-lite: “Oh, you’re sad? Why do you think you’re sad?” Simple, yet oddly effective.

What Exactly Is a Rogerian Psychotherapist?

Carl Rogers, the mind behind Rogerian therapy, believed in letting patients talk freely while the therapist mostly nodded along and said things like, “Tell me more about that.”

Eliza mimicked this style perfectly with its “pattern-matching” trickery.

People found it surprisingly comforting to interact with a program that simply repeated their words in the form of questions.

How Eliza Worked: The Tech Behind the Magic

Eliza wasn’t running on AI wizardry — more like clever scripting. It worked by:

1. Pattern Matching: It recognized keywords in your text.
2. Response Templates: It had a set of canned responses for those keywords.
3. Reflection: It turned statements into questions, like, “I feel sad” becoming “Why do you feel sad?”

So, if you typed, “I’m worried about my job”, Eliza might reply, “Why are you worried about your job?” Brilliantly simple — yet people genuinely felt understood.

The Eliza Effect: When We Believe Machines Have Feelings

Ah, the Eliza Effect — it’s when humans start attributing human-like qualities to machines.

Weizenbaum was startled when people, including his own secretary, poured their hearts out to Eliza. They knew it was a program, but the brain is weird like that.

This effect is still alive today.

Ever felt like Siri or Alexa “understands” you? Yeah, that’s the Eliza Effect in action.

Eliza’s Journey Beyond MIT: From Mainframes to Home PCs

Eliza wasn’t stuck in the MIT labs for long. It was ported to BASIC, a popular language for early personal computers. If you had an Apple ][ (like I did — ah, memories!), you could run Eliza and experience your very own digital therapist at home.

My Nostalgic Eliza Encounter

I remember booting up Eliza on my Apple ][. The green text glowed on the CRT screen:

1
2
3
4
5

**Hello, I am Eliza. How can I help you today?**
> I'm feeling stressed.
**Why do you think you're feeling stressed?**
> My dog won't stop barking.
**How does your dog's barking make you feel?**

It was oddly compelling. For a kid in the ’80s, this was mind-blowing stuff.

The Animals Program: Early Machine Learning

Eliza wasn’t alone. My Apple ][ also had the Animals program, a simplistic learning game. It worked like this:

1. The program guessed an animal by asking yes/no questions.
2. If it guessed wrong, it asked you to supply a new question and answer.
3. It saved this info for the next round.

Here’s a sample session:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

**Think of an animal. I will try to guess it.**
> OK
**Is it a mammal?**
> Yes
**Does it bark?**
> No
**I give up! What was your animal?**
> Kangaroo
**What question could I have asked to tell a kangaroo apart?**
> Does it hop?
**Thank you! I will remember that.**

This was rudimentary machine learning: the program actually improved over time.

Eliza vs. Modern AI: From Rules to Neural Networks

Eliza ran on hand-coded patterns — no deep learning here.

Modern AI uses neural networks and large language models (LLMs).

Instead of simple pattern matching, LLMs learn from vast datasets and generate contextually relevant responses.

In short:

• Eliza: Keyword patterns, no learning.
• Animals: Basic decision-tree learning.
• Modern AI: Neural networks and big data.

Eliza’s Legacy

Eliza laid the groundwork for interactive computing and sparked conversations about AI ethics.

And to any person who tried it when they were a kid (ME!), it was a surreal experience.. Even on a slow old PC from the 1970s or 80s….

Weizenbaum later in life, criticized the unchecked belief in AI, worried about humans relying too much on machines for emotional support.

Key Ideas

• Eliza was the first chatbot, built in 1966 by Joseph Weizenbaum.
• It simulated a Rogerian psychotherapist using pattern matching.
• The Eliza Effect describes how people anthropomorphize machines.
• The Animals program on the Apple ][ demonstrated early machine learning.
• Eliza influenced the development of modern AI and NLP.

References

1. Weizenbaum, J. (1966). ELIZA — A Computer Program For the Study of Natural Language Communication. Communications of the ACM.
2. https://en.wikipedia.org/wiki/ELIZA_effect
3. https://liacademy.co.uk/the-story-of-eliza-the-ai-that-fooled-the-world/
4. https://builtin.com/artificial-intelligence/eliza-effect
5. Ford, M. (2015). Rise of the Robots: Technology and the Threat of a Jobless Future.
6. Russell, S. & Norvig, P. (2010). Artificial Intelligence: A Modern Approach.