Els Aarts, Media Technology, NMNT, 2016.
ELIZAristotle
In this project, I decided to look into the first chatbot ever made: Eliza. I want to know how she thinks, and how people reacted to her when they first met her. Thereafter, I decided to improve her a little by making giving her more content; as she is essentially a counselor, I wanted her to not just ask questions but also give some useful advice on life. This is done by implementing the thoughts of a very old icon: Artistotle.
Who is ELIZA?
ELIZA was one of the first 'chatterbots', a computerprogramme that conducts a conversation via auditory or textual methods. She was developed by Joseph Weisenbaum in 1966, as a part of the MIT Artificial Intelligence programme. Her most famous ‘script’ is DOCTOR, which is designed to resemble a so-called ‘Rogerian psychotherapist’, whose therapy consists of mostly asking questions to make a patient develop a ‘sense of self’. This was done by using simple pattern-matching techniques: ELIZA does not have a very extended vocabulary, but mostly mirrors the user’s input. When said “My toe hurts because someone stepped on it”, ELIZA could give an answer like ‘Why do you say that your toe hurts because someone stepped on it?’.
Like dr. Frankenstein, Weizenbaum was ultimately not very happy with what he had created[1]. People got quickly and deeply emotionally involved in talking to ELIZA. In his paper “Computer power and Human Reason: From Judgment to Calculation” (1976), he writes ‘I had not realised that extremely short exposures to a relatively simple computer programme could induce powerful delusional thinking in quite normal people.’ Apparently Weizenbaum’s own secretary asked him to leave the room so that she and ELIZA could have private conversations[2]. Weizenbaum found it extremely difficult to convince people that ELIZA did not really understand what they were thinking. To his horror, people begun to telephone Weizenbaum and plead for ELIZA’s help to sort out their problems. Also, some psychiatrists wanted to use ELIZA on their patients.[3]
In Weizenbaums aforementioned paper, he states that we should never allow computers to make important decisions because they will always lack human qualities such as compassion and wisdom.[4]
Why Aristotle?
Aristotle (384-322 BC), the Greek philosopher who was arguably a media icon, but an icon for sure, tried to advice people on how to behave with his ethics of virtue. His ethical theory still inspires people on how to deal with (moral) issues[5],[6], as his wisdom stretched almost all fields of study that existed at that time.
To honor Weizenbaum and to make him less afraid of his own invention, I thought it would be a good idea to enhance ELIZA with the characteristics she lacks of, namely true wisdom, compassion and morality. And who would is almost an embodiment of these virtues? Yes, mr. Aristoteles himself.
How are they combined?
I downloaded source code based on ELIZA’s original code[7]. Since I run another version of Python, and I am not a very experienced programmer, it took me some time to adjust and debug it, and to investigate how ELIZA works. She has two main functions, reflections and psychobabble. In reflections, she looks for pronouns and mirrors them: I:you, am:are, your:my etc. In the function psychobabble, she looks for keywords to mirror the question, or to ask a question in a specific theme.
I tried to implement some witty quotes from Aristotle, but also tried to preserve the ‘empathy’ that people experienced while talking to ELIZA. I gathered quotes from Aristotle’s work on different subjects, sorted them by subject and made them into new keywords. I tried to match as much of his quotes to already existing keywords and paths in ELIZA.
Example of a conversation with ELIZAristotle
Hello, I'm ELIZAristotle. How can I help you?
> Hello! I am Els.
Hello... I'm glad you could drop by today.
>Thanks!
Please tell me more.
>Actually, I can't tell you about my personal problems, because I'm doing this for a project at the university.
Young people are in a condition like permanent intoxication, because youth is sweet and they are growing.
> Oh, I need to think more about your very wise reflection.
To be conscious that we are perceiving or thinking is to be conscious of our own existence. So, elaborate on oh, you need to?
> Well, uh, I just need to be happy I guess.
To enjoy the things we ought and to hate the things we ought has the greatest bearing on excellence of character.
> Okay, I'll do that then!
A likely impossibility is always preferable to an unconvincing possibility.
> Byebye ELIZAristotle!
Very interesting.
> quit
Thank you, that will be $150. Have a good day!
Sources
[1] Copeland, J. (2015). Artificial intelligence: A philosophical introduction. John Wiley & Sons.
[2] Idem.
[3] Idem.
[4] Weizenbaum, J. (1976). Computer power and human reason: From judgment to calculation.
[5] Bickford, S. (1996). Beyond friendship: Aristotle on conflict, deliberation, and attention. Journal of Politics, 58, 398-421.
[6] Hartman, E. M. (2008). Reconciliation in business ethics: Some advice from Aristotle. Business Ethics Quarterly, 18(02), 253-265.
[7] https://www.daniweb.com/programming/software-development/code/380743/eliza-aka-therapist-facelift. , retrieved at 25-03-2016.