Eliza – one of the first chatterbots (chatbot) and one of the first case of Turing Test.
The Eliza Effect:
The tendency to project human traits – such as experience, semantic comprehension, or empathy – into computer programs that have a textual interface. The effect is a category mistake that arises when the program’s symbolic computations are described through terms such as “think”, “know” or “understand”.
Here is an article explaining the origin of Eliza effect and how works before and nowadays:
https://builtin.com/artificial-intelligence/eliza-effect
Here is a version of the chatbot:
https://psych.fullerton.edu/mbirnbaum/psych101/eliza.htm
Although it tricked a lot of people into believing that they were talking to someone who truly understood the input, my experience with this chatbot is not at all pleasant. Despite one possible reason being the chatbot’s over-sixty-years old age, it is extremely annoying that Eliza fails to give any valuable output when you try to discuss serious issues with her, and she constantly answers your question with another question or simply repeats your words.
This is due to the limited scripts (the most famous variation being DOCTOR) encoded in the program written in MAD (programming language and compiler for IBM, developed in 1059 by Bernard Galler, Bruce Arden, and Rovert M. Graham) – SLIP (a list of processing computer programming language, invented by Joseph Weizenbaum in 1960s).
In the end, if a person in need of therapy chat with Eliza and ends up feeling that they are chatting with a human, it might as well be a result of just hearing themselves talk because they get a sense of validation by the loopy responses (a large part is questions) they get from Eliza. In a simpler saying, Eliza simply mirrors users’ language.
However, in another article 1960s chatbot Eliza beat OPENAI’s GPT-3.5 in a recent Turing test study that is contradictory to my observations of Eliza, authors claim:
“First, ELIZA’s responses tend to be conservative. While this generally leads to the impression of an uncooperative interlocutor, it prevents the system from providing explicit cues such as incorrect information or obscure knowledge. Second, ELIZA does not exhibit the kind of cues that interrogators have come to associate with assistant LLMs, such as being helpful, friendly, and verbose. Finally, some interrogators reported thinking that ELIZA was “too bad” to be a current AI model, and therefore was more likely to be a human intentionally being uncooperative.”
——————————————————————————————————————————————————
The whole Eliza phenomenon and this week’s screening Her (Spike Jonze, 2013) lead me to think about AI’s role in practicing psychology, psychoanalysis, and therapy treatment. For example, there is Wysa chatbot (https://www.wysa.com/), a clinically validated AI.
One of the most important advertising points for these chatbots is anonymity, the relief of stress for people who want to avoid direct interaction with people (old-fashioned face-to-face therapy). The AI therapy also solves the time schedule limit, because with a chatbot, patients can do the talk anytime. While scrolling through their website, I noticed that this system is not completely operated by AI and still needs human backup, which reminds me of the discussion in class about how huge a human crowd is behind these AI systems. What’s more, AI therapy is not without its obvious limits and imperfections. They give biased responses, and sometimes even encourage suicidal ideas.
(Minor spoiler) An interesting parallel pops to my mind when I recall Theodore in the film sincerely doing love letters, which I later realized is his job. All these people receive sweet words not even from their own partners. Theodore writes so many of those just like Samantha dates thousands of men. Now Spike Jonze cleverly draws a questionable line between human and AI as well.
Leave a Reply