Wednesday, January 16, 2013


A long time ago there was a computer program called Eliza1. You typed a line of text at it, and it replied with a line of text. You could have a conversation with Eliza, in this fashion. The remarkable thing about Eliza is that, at least mostly, Eliza could have a conversation with you without remembering a single thing. You'd think the computer program would remember something about what you'd said, and would be crafting responses based on where we are in the conversation, somehow, wouldn't you? Well, not always.

Eliza: How are you today?
You: I feel sad.

Eliza picks up on "I feel sad" and produces a stock response:

Eliza: Why do you feel sad?

The important thing here is that it doesn't matter how often or when you typed "I feel sad", Eliza will always respond with "Why do you feel sad?" or a randomly chosen similar phrase. Then you might proceed:

You: I'm sad because my mother is sick.
Eliza: How do you feel about your mother?

Again, Eliza will ask you how you feel about your mother any time you say anything about your mother. Had you typed "mother mother mother bat bat cow" Eliza probably would have asked you how you feel about your mother.

This works because, while Eliza is not remembering anything about the conversation, you are. You don't type "I feel sad" at random, and you don't type "mother mother bat bat cow." You type things that make sense. Eliza merely needs to be sufficiently vague to evoke a suitable response, and sufficiently on-track to give the sensation of a coherent conversation. Eliza's responses are loose enough to permit a wide range of responses, but specific enough to evoke responses that are more or less within her scope for reasonable response. The looseness prevents the sensation of a canned and uninteresting conversation. The specificity allows the conversation to remain seemingly on track.

The results are a little loopy, but surprisingly convincing.

Why on earth would we care?

There are certain parallels between how Art communicates and how Eliza does, it turns out. If you squint a bit. Photography has arguably a bigger problem than other Arts: a writer can simply write down words that say what needs to be said; the painter and sculptor construct things, and have more flexibility in how they stimulate you. The photographer is, ultimately, stuck with what is in front of the camera. It's all matters of degree, however. The basic problem is that a static piece of art, a thing that cannot "remember the conversation," is trying to communicate with you.

The lesson to take away from Eliza is that a truly successful communication can be achieved when the Art makes statements that are vague enough to stimulate a response, but specific enough to create a coherent conversation. A photograph can state a question, but the question should not be so specific as to pre-determine the answer, and not so general as to allow anything as an answer.

A photograph that opens a broad question just leaves the viewer puzzled. Macro photographs of stuff open the "what the heck is that?" question, all too often. The viewer walks away. A photograph that answers all the questions it poses isn't very interesting: Here I Am In Front Of The Eiffel Tower isn't very interesting. We might wonder, briefly, who that person is but then we realize that it is the photographer, or the photographer's girlfriend. The conversation ends before it begins.

A photograph that asks us "What is that guy doing?" and allows a modest collection of answers is interesting. If he's doing something specific, but with an unclear motive ("I wonder what he's looking for?") we have a good degree of specificity. We can write a little story about what he's doing, we can check it against other details in the frame, we can back up and write a new story. We are left with a little wonder and mystery, but a mystery that is connected to something straightforward and un-mysterious. Parts are given to us, a whole is left for us to complete. The photograph begins the conversation and nudges it along, we maintain all the context, and a personalized conversation occurs with the image. All that we are and have been informs the context we're keeping as we interact with the piece. What we bring, as much as what the photograph provides, makes the conversation interesting.

[1]  This is not strictly true. Eliza had, I think, a mode of operation that works as described. However, newer implementations of Eliza-like systems exist which work exactly as described.

No comments:

Post a Comment