Image © Peli Grietzer,“A Theory of Vibe”
Disclaimer: Most of the insights in this post have already been addressed by semiotics, and won’t strike anyone familiar with that discipline as novel. This is more just an attempt to reframe and re-analogize a process than to advance actual arguments.
Delving into the world of machine learning has me interested in encoding as an analogy for art processes. Consider, for instance, the mental image which so many readers generate in an encounter with a text. This translation from text to image is merely the inverse of what Samuel R. Delaney describes as his writerly process in The Jewel-Hinged Jaw: Notes on the Language of Science Fiction.
When I am writing I am trying to allow/construct an image of what I want to write about in my mind’s sensory theater. Then I describe it as accurately as I can. The most interesting point I’ve noticed is that the writing-down of words about my imagined vision (or at least the choosing/arranging of words to write down) causes the vision itself to change.
Drawing on his existing knowledge of image-word relationships, Delaney takes a mental image which is a composite of objects and class attributes to which he has previously been exposed. (For example, he cannot “invent” a chair from thin air, but instead takes a known object and then mentally modifies its size, shape, or color, perhaps even combining it with features from other known objects.) Delaney then takes those words which he believes best correspond to the objects in his mental image, constrained of course for all sorts of fictive demands like narrative and thematic coherence, as well as linguistic draws like sentence rhythm, alliteration, and novelty. (There is a process of exchange, which Delaney notes toward the end of his paragraph, in which these textual demands shape his original mental image in turn.)
When the reader approaches a text, he builds his mental image in a process symmetrical to Delaney, but utilizing a different set of known images and class attributes so that the lossy text decodes into a new and unique image. Like Delaney, the Delaney reader will find word-world correspondences, where a “stereotype” of an object is modified by qualifying language. This stereotype is highly contextual even before it is explicitly described: the reader of Madame Bovary does not summon a quintessential modern living room to mind; he summons a historical living room, modified by pertinent qualifying and often implied or assumed information.
Good writing requires encoding mental images in a way that the text will generate coherent and well-built mental images for its audience audience, said this audience possessing a wide-range and highly varied set of word-image correspondences, or “stereotypes.” Delaney again:
When the corrections as we move from word to word produce a muddy picture,
when unclear bits of information do not resolve to even greater clarity as we
progress, we call the writer a poor stylist.