CAP – Fruin, Noah Waldrip, “The Eliza Effect”
Anthony Pino
June 21, 2017

Summary: Fruin provides a brief history of the Eliza idea, from his point of view as a child, beginning with BBS, the online Bulletin Board System which operated on a phone line by dedicated individuals in their basements and dens, and enabled subscribers to call into BBS, read messages leave messages, download and upload files, play text-based games and hold conversations in text with others---all a “window into the Internet.” A childhood friend showed him how he downloaded a program that enabled him to hold conversations with others. He called it Eliza (23-24). The next phase of Fruin’s discussion involves introduction of Joseph Weitzenbaum’s mid-1960’s in which one seemed able to have a real-time conversation with a computer (24). Its chief attribute was natural language processing, when Eliza played “Doctor,” a script which resembled psychotherapy (a “chatterbot”). Today the name Eliza is associated with The Eliza Effect, the illusion that a computer is more intelligent (complex and capable) than it really is (25), and a programmer’s choice to severely limit a program’s activities based on underlying processes and often lead to breakdown. Weitzenbaum demonstrated its potential in a dialogue between an unhappy girl and a responding computer. Eliza dialogued using typewriter interaction rather than monitor-keyboard-mouse. Responses were based on a currently active script.

A Script is a bundle of data and simple processes. In Eliza’s case the script involves examining the user’s submission, looking for words that are in a currently active script, and making substitutions (I for you, for example), or for keywords, which are stacked in priority. A decomposition rule is then applied to the word at the top of the stack, which is examined for connections to word patterns. Using these patterns the system reassembles them making substitutions and causing a response. Thus Eliza transforms word patterns and uses them as responses. A related program called Doctor has a process which, when a pattern is not found, effectively stalls for time and seeks to expand the volume of data a user proves. “Tell me more about your problems with your mother…” or “You mentioned earlier that…” (my construction). The system is effectively fishing for more data (my perception). Weitzenbaum concluded that the process is “unmasked” as a mere collection of pre-established procedures, not providing users with a clear picture of its internal processes (31), which are not complex but mechanistic or automatic (my words). People began believing computer science was approaching HAL, the feeling computer model in 2001: A Space Odyssey.

Limitations of Eliza and Conclusions: Text-only responses; the user belief that the appearance of intelligence is, in fact, intelligence; Eliza/Doctor is “a series of linguistic tricks”; Weitzenbaum felt his invention caused “delusional thinking,” and became an arch-critic of popular perception of magical machines arguing that we do a “disservice” to ourselves when we perceive them this way (33). Eliza/Doctor consists of “rote simplicity” (35-36). Initial enjoyment is part of the appeal of Eliza/Doctor (35). Eliza/Doctor plays on “the interpretive expectations of the user” (36). The Eliza/Doctor system is prone to breakdowns, prone to disillusionment and game-playing by users (35). The program can be shielded from breakdown by limiting its use (37).

Questions: In many respects, isn’t our own cognition similar to that of Eliza/Doctor?
(Sometimes important political, religious and social issues are settled by pre-established narratives, vocabulary and scenarios).

Are we better off or worse off for having experienced Eliza/Doctor? (ie, to what extent were problems solved or injuries allayed through Eliza/Doctor?)