The Significance of ELIZA

The year is 1966. Interest in Artificial Intelligence (AI) is on the rise, especially in science fiction circles.

The concept of artificial neurons had just been introduced in 1943, and in 1950, Alan Turing proposed the Turing Test, a revolutionary benchmark for AI that would guide research for decades.

At this time, personal computers were non-existent, and interactive computing was still in its infancy, with innovations like Sketchpad appearing only in 1963. Natural language processing (NLP) was largely unfamiliar to the general public.

Then ELIZA appeared, astonishing both experts and everyday users alike. A program capable of engaging users in coherent dialogue? This was unprecedented.

Today, ELIZA's creation is celebrated as a pivotal moment in AI history.

Ironically, her creator did not foresee this impact.

Who Is ELIZA?

ELIZA is an NLP program crafted by Joseph Weizenbaum in 1966 at the MIT AI Laboratory.

NLP is a subset of AI focused on language and linguistics.

Her most recognized script, DOCTOR, emulates a Rogerian psychotherapist, specifically the techniques of Carl Rogers, who reflected patients' statements as questions.

A brief example would be:

PATIENT: I’m feeling down today. ROGERS: Why do you think you’re feeling down today?

Does this sound familiar? It mirrors ELIZA's responses closely.

Weizenbaum originally aimed to create a 'chatbot' (a term not yet invented) to illustrate the superficial nature of human-computer interactions.

Imagine his surprise when users began to perceive ELIZA as genuinely understanding and feeling!

In his own words,

“I had not realized that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

Challenges in Developing ELIZA

Creating an interactive program like ELIZA was a formidable task. While machines excel at rapid calculations, simulating human-like behavior remains a challenge even today.

NLP is intricate because human language is inherently complex, reflecting the nuances of human thought. So how did ELIZA navigate this complexity? Did Weizenbaum train her on massive datasets? No.

This is where the concept of psychotherapy comes into play.

By primarily posing questions, ELIZA directed the conversation without needing to generate substantial content herself.

She didn’t have to grasp the situation; she just needed to appear to comprehend it.

You might wonder, how does she determine what questions to ask?

This is where pattern matching becomes vital. Several factors are at play:

• ELIZA must identify key terms in the user’s input.
• She needs to handle multiple keywords or phrases appropriately.
• ELIZA must manipulate these terms to craft a response.
• She must know how to respond when no keywords are detected.
• Additionally, ELIZA needs to determine when to conclude the conversation.

How ELIZA Operates

To address the above factors, I will outline the two primary components of her operation: decomposition and reassembly. We will also explore her responses when faced with unfamiliar inputs.

Definitions:

• Input refers to the text the user types.
• Output refers to ELIZA's responses.

Decomposition

When a user submits an input, the program analyzes the text for syntactic patterns, allowing it to break the input into manageable segments.

For instance, consider the input: "I am feeling down."

Here’s how it breaks down:

"I" is the pronoun. "feeling down" is the action. "am" connects the action to the pronoun.

The program then scans for keywords and assigns numerical values based on their relevance. This ranking system determines their importance.

For example, "feeling down" may receive a rank of 1 due to its substantive content, while "I" might rank 2 as it specifies the subject, and "am" could rank 3.

ELIZA also identifies certain terms deemed significant for responses, such as "alike" or "same."

The keywords are organized in a 'keystack,' with the highest-ranked term on top.

• "feeling down"
• "I"
• "am"

Reassembly

Next, ELIZA applies specific transformations based on the identified keywords. Higher-ranked keywords receive priority.

For example, "feeling down" will feature prominently in her output, such as in the response: "How long have you been feeling down?"

The pronoun "I" transforms into "You" (first-person pronouns are converted to second-person).

If the input contains words like "alike" or "same," the response will engage with this notion of similarity, possibly with "In what way?"

How are ranks determined? Statements typically follow certain patterns or include specific types of words, regardless of content. Words aligning with these patterns hold higher ranks. ELIZA was programmed to recognize specific words (patterns and types) that also receive elevated ranks.

Once the transformation is complete, ELIZA reconstructs the output, incorporating relevant 'filler' or 'question' words. The final output reflects this assembly.

For instance, input: "People are rude." Output: "Possibly they are rude."

Here, "Possibly" acts as the filler word, "People" is transformed to "they," and "are rude" is carried over as the main content.

To end a conversation, ELIZA looks for cues like "Bye." or "Goodbye."

What If No Keywords Are Found?

If ELIZA encounters unfamiliar input, she might respond vaguely with comments like "I see." or "Please continue."

Alternatively, she can draw from her 'Memory' to reference earlier parts of the conversation.

Quite clever indeed.

Addressing the Key Question

You’ve come this far! You understand ELIZA's significance, her identity, and how she operates. But why was she named ELIZA?

Cue suspenseful music.

ELIZA's name draws inspiration from the character in George Bernard Shaw's 1912 play, Pygmalion.

The character, Eliza Doolittle, begins as a poor flower vendor. With the help of Higgins and Colonel Pickering, she learns to speak 'upper-class English,' aiming to pass as a duchess.

In a similar vein, ELIZA was designed to refine her dialogue to mimic a human psychotherapist.

This may seem far-fetched, but consider this: In ancient Greek mythology, Pygmalion fell in love with a statue he carved, which subsequently came to life. ELIZA, in her own way, achieved success by emulating 'life.'

As a final note, recall my initial conversation with ELIZA about houseplants? Her opening line was, "How do you do? Please tell me your problem."

What phrase did Higgins teach Eliza to pronounce, which she famously repeated at Mrs. Higgins’ gathering and was remembered for at the Embassy?

> "How do you do?"

For a brief overview of the Turing Test, check out this article.

If you found this exploration of ELIZA engaging and want more, consider joining Medium for access to unlimited articles like this for just $5 a month. I’ll earn a small commission if you use my link above.