I was teaching a student the other day using my "picture file" to assess his English speaking level. After the lesson, I was filing my pictures back in their accordion folder, when it dawned on me I had no "system" for filing these pictures. I just shuffled them together and then slid them into random slots. What a lazy mess. How, I asked myself, could I keep images in order? How do you file images, without simply giving each one a verbal tag, I mean?
The question came inot more focus when I recalled my Google searches for the icons here on my blog. For example, I had to ask for "icon" + "augustine" and see what came up. As long as the image provider had a word or two connected to the images, I could track them down. As you know, I certainly couldn't select a basic *visual* fragment that would then come up in other images.
But why not? Why couldn't we, for example, enter a stereotypical nose icon, or even a certain kind of nose, into, say, Google (or into a police suspect database!) and see what images match it? What categories could we use to put images in order? Geometric? Chromatic? Anatomical? Based on shading? Based on frequency of occurrence? Could we devise a coherent "visual alphabet" for navugating through huge picture databses and photo archives, especially when they lack verbal tags? Or is filing a purely verbal activity?
My best guess is that the apparently innate ability of humans to recognize and classify faces would offer a solution. even if we can't verbalize it, we all have a basic idea of features that make someone look "old" or "dangerous" or "friendly" or "sad," and so forth. So, while the *categories* we use to file faces into are verbally denoted, it seems the facial traits themselves are immediately *and nonverbally* recognizable. If it's ever devised and made workable, I think this technology shouild be called a "Platonic Query," or something like that. Maybe a "Trope Search."
Now that I think of it, this filing system gets into some very deep philosophical waters. Let's talk metaphysics, shall we? We all know what a chair should look like even without a verbal description. "A picture is worth a thousand words." We also pretty well recognize immediately -- unless its at some indefinite boundary between one thing and another -- whether a thing is one thing and not another. Would you have any hesitation to say this is a dog and that is a spoon? (I hope not!) The human mind is such that we know what things are almost without having to think or speak about it.
This feature of the mind provided one of the strongest intutive defenses of Plato's theory of forms (metaphysical realism), according to which, very briefly, there are real, immutable, perfect, abstract "universals" or "forms" of which every concrete thing is but an imperfect manifestation. Every circle is but a dim realization of some perfect circle every geometrician dreams of. Given our innate sense of forms, we know we've erred as soon as we start to draw a triangle when we mean to draw a circle, or when we start to make a mess when we mean to make a ceramic bowl. We know this chair is a chair because, in some mysterious way, we know what "Chair" is. Or so the realists would have it.
Realism's historical nemesis is called nominalism, a theory (defended perhaps most notably by Parmenides, William of Ockham and W.V.O. Quine) that says there are only particular, concrete things that we arbitrarily put into classes based on similar traits. Hence, a chair is a chair only "nominally" because we call it that. We know this chair is a chair because, well, we already know by experience that this chair looks and feels and acts a lot like another thing we've already called a chair. As we gain more experience, the class of chair expands until we can recognize a chair in even the most bizarre art deco, or even in splintered heaps.
A more modern theory of being, standing uneasily somewhere between realism and nominalism, is called trope theory, according to which there are real, abstract things, but each one is totally unique. There is a form of Red -- but only because that one form is based on a particular red object, X. X's redness is indeed a perfect, abstract, repeatedly recognizable entity, but it does not partake of some higher Ideal Redness. Each thing is a form unto itself, and each class of things is a nominal grouping based on similarities between these trope-forms.
At any rate, this (my?) visual filing system would rely on a databse of common tropes. You could, I imagine, look for "light red ball" and then narrow it down even further. Already, though, as you can see, this is collapsing into nothing but a really really specific verbal search engine with lots and lots of, let's be honest, real-world clip art.
Hmmm, this deserves more thought. Email me any helpful ideas, thanks!