Assignment 4: Eliza Analysis

Eliza, also known as Doctor, is a computer program based off of a Rogerian psychotherapist. The program allows for a user to interact with a program similar to the way a patient would interact with their therapist. After playing with Eliza for a few hours, I realized that it had practically mastered the Rogerian psychotherapy method. Instead of giving you a straight forward solution, Eliza employs the Rogerian methods and only gives back questions instead of answers. In some cases, Eliza even places the user back on path when it fells that the conversation is deviating away from users talking about themselves, Eliza quickly attempts to get the user back on track. Eliza is constantly digging for information about the user. For instance, if one was to tell Eliza that they didn’t like school, Eliza would ask questions based off of that input, which causes the user to go deeper into the problem. The program mimics the Rogerian psychotherapist method to semi-perfection, meaning that Eliza has some bugs.

For starters, Eliza suffers from a dated script. The program doesn’t understand common language used today, such as “yep” or other slang terms. In addition, Eliza’s script automatically assumes everything is a problem, unless you refer the Eliza itself. As a result, Eliza will attempt to try and make the user go deeper into statements which have no relative meaning outside of the initial one (i.e. 1+1 = 2). Furthermore, the responses “yes” and “no”, always seems to suggest the type of attitude a person possesses. For instance, during a conversation with Eliza, I answered the question “Does that question interest you?” with the answer “no”, and Eliza assumed I was being negative whereas I was merely answering the question. On the other hand, when I answered that same question “yes”, Eliza assumed that I was a positive person. Again, I was only answering the question I was presented. In my opinion, Eliza’s script should be changed to deal with responses that aren’t of a therapeutic nature, such as dealing with “yes” and “no” answers after the program asks the user a question.

Another problem with Eliza comes from the use of Instant Massaging. Since instant messaging allows for users to communicate with another human across the internet, it has become almost like using the phone. Therefore, like normal talk, users of instant messaging have developed their own lingo. By using acronyms and symbols, such as “lol” (laughing out loud) and “: - )(smiley face)”, users are able to reduce the amount a typing to communicate while getting the message across to the other user. In addition to IM lingo, another problem with interacting with Eliza is that the script doesn’t allow the program to have a conversation. In addition, IM makes humans think that they are talking to a human, rather than a computer. Therefore, they expect a response to be “human-like” response. AT first, Eliza seems to be a genuine human. However, due to the programs limited understanding of the human language and limited response patterns, Eliza’s humanism is quickly stripped.

Due to the way people interact with messaging clients, Eliza needs to be able to converse outside of the normal therapeutic methods. Going back to the positive and negative responses, Eliza should be able to understand that no isn’t always referring to something negative, and yes isn’t always a positive remark. Next, Eliza’s sentence structure needs to be changed. The program has a tendency to think everything is a statement that can be formed into a question. Eliza frequently uses improper sentence structure, mainly because her script seems to assume she is taking one verb and one noun. Eliza should be modified to “read” sentences and make judgment about nouns and verbs, and then make an accurate response. Lastly, Eliza’s understanding of numbers and symbols should be changed as. As of now, a numbers and symbols generate random responses from Eliza, which eventually lead to her bail out statement, “I do not fully understand you”. If I were redesigning Eliza, I would update the script to include symbol and number recognition, as well as responses based on certain inputs of numbers and symbols (such as a happy comment if : - ) is input).

Even though Eliza is unrealistic, it does serve its purpose fairly well. If I were to have a conversation with Eliza the way it was designed, I probably wouldn’t be able to tell the difference between an actual Rogerian psychotherapist and Eliza. However, since Eliza can’t interpret anything outside of the script, it makes the program unreliable for actual conversation.

Download My Conversation with Eliza