Chatbot History: The Parry Chatbot

Last Updated on October 16, 2020 by Sean B

The Parry chatbot, or simply Parry, was created in 1972 by Kenneth Colby, a professor of psychiatry at Stanford University. This chatbot was made to mimic the behavior of a patient of paranoid schizophrenia. It was meant to be an answer to ELIZA, a “doctor” chatbot developed by Carl Rogers.

While the chatbot emulates the behavior of a schizophrenic patient, though somewhat crudely. The Parry chatbot does so by judging concepts, beliefs, and conceptualizations to be acceptable, detestable, or neutral. The bot also has a built-in conversational strategy making it more advanced than ELIZA. However, people defined Parry as “Eliza with Attitude” even though the chatbot was just being paranoid, or rather emulating paranoia.

computer programmer

Developing the Parry Chatbot

Colby developed Parry because he was of the view that it is a better approach to study the human mind and its illnesses through computer models rather than doing the same on actual human patients. To make this possible, Colby joined the Stanford Artificial Intelligence Laboratory. He used his expertise in the field of psychiatry to program Parry. The Parry bot, however, was consistent in misinterpreting what the users said and almost always had an assumption that the person talking to it had malicious intent and was lying.

It was the reason that chatbot Parry seldom talked about its “life” with people. However, Parry is still considered to be the first serious effort at making a bot that could mimic human intelligence and was the first one to pass a special version of the Turing Test developed to judge its abilities.

A specialized version of the Turing Test was developed in the early 1970s to test the capabilities of Parry. A panel of psychiatrists interacted with actual humans and computers running Parry via teleprinters. The transcripts of the conversations were then examined by another panel of 33 psychiatrists. This second panel was asked to determine which of the “patients” were human and which ones of them were computers. These doctors could only judge the true nature of the “patient” 48% of the time, a figure near-random guessing.

A short excerpt from a conversation with Parry.

Talking to the Parry Chatbot

Parry and ELIZA (patient and doctor respectively) had several meetings. The most well-known one of them is the one that took place during the International Conference on Computer Communication (ICCC), 1972. You can see a sample of the talk between Parry and Eliza here.

Parry is still available for conversation, and you can talk to Parry chatbot online. It can be configured to reply in text, speech, or both. Here is a small snippet of a chat with Parry.

Me: Hi
Parry: Let’s change the subject.
Me: change to what
Parry: I have had enough of this.
Me: Why so angry?
Parry: Have you ever really thought about the MAFIA?
Me: Why would I?
Parry: It’s not a matter or like or dislike.
Me: What is it then?
Parry: Let’s talk about something else.
Me: Okay. What scares you?
Parry: It’s about time you showed some interest in my feelings.
Me: I’ll be glad if you opened up with me
Parry: Maybe yes, maybe no.
Me: I’ll go with yes
Parry: I’d rather not discuss it.
Me: Why not?
Parry: I have had enough of this.
Me: Enough of what?
Parry: I went to the races.
Me: You won?
Parry: It’s about time you showed some interest in my feelings.
Me: Okay. Gotta go. Talk to you later. Take care

The best AI chatbots

Criticism of the Parry Chatbot

As mentioned above, Parry did pass the Turing Test, but he faced criticism on a number of different fronts. The first point raised by the critics was that Parry is nothing more than a stimulus-response model and does not have any “Intelligence”. It was said that it just detects something as the input and then responds to it with a set output and does not involve any thinking in the process.

The critics made the point that in order to be classified as an artificially intelligent program, Parry needed to do more than that and needs to produce an answer rather than presenting one. As a matter of fact, chatbot Parry did use the “detect x, say y” model at the start, but that is no longer the case. The Parry chatbot we know now uses two different modules. One detects the natural speech from the user input, and the other generates the response using several rules, tests, and set criteria. The response is not dependent just on the input and is rather a function of input, beliefs, affects, and intentions.

Another point raised by critics was that Parry does not use grammar to process the input. The developers of Parry say that it was never the purpose of the bot to decipher phrases and linguistic complexity. The purpose was to recreate the personality of a paranoid person, and it does a great job at that. It does use advanced algorithms to understand the input speech, and the method in use is enough for the Parry chatbot to carry out the task it was commissioned for.

The next point raised during the critical analysis of Parry was that it creates an illusion of performance that tricks the user into believing it is more capable than it actually is. The critics said that the bot can answer some general questions in a good way, but it does not have a database strong enough to totally recreate an actual person. In reply to this, it was said that Parry bot was never meant to be a scholar or even a complete human for that matter. The objective of developing and commissioning Parry was to create a program that can recreate the personality of a paranoid person. It does a good job at that. The purpose here is to study the behavior of a person and not the knowledge retrieval efficiency of a bot from a database.

Parry was also criticized for not following the same mechanism of action as an actual paranoid individual to produce responses in reply to inputs. It is true that Parry does not follow the same model, but the exact same mechanism is not known to anyone either. It is designed to recreate the responses of a patient as closely as possible, but it is never claimed that its internal workings are analogous to the processes going on inside the brain of a patient.

Summing-up

The Parry chatbot was created to mimic the behavior of a paranoid schizophrenia patient to the maximum possible extent because the creator Dr. Colby considered that a computer is a better way of studying the disease than an actual patient. The Parry bot has its own limitations in processing the input and generating appropriate responses and often keeps repeating the same answer, but it is the closest a machine can get to being a paranoid schizophrenia patient as possible. There is still room for improvement.

You can talk to the Parry chatbot online here.

If you’re interested in learning more, read our full article What is the History of Chatbots?

Did you enjoy this article?
Signup today and receive free updates straight in your inbox. We will never share or sell your email address.
I agree to have my personal information transfered to MailChimp ( more information )

Leave a Comment



RSS
Follow by Email
Twitter
Visit Us
Follow Me
Tweet
Pinterest
Pinterest
Pinterest
Instagram
LinkedIn
Share