Oh good, I did comprehend “Classification” and “Cluster.” I never had such a nice precise language differences available to me although obviously the concepts are familiar.

Your word “problematic” was exactly what I thought it meant — I am most familiar with filing systems that make perfect sense — to the person doing the filing. But to someone else, they can seem totally arbitrary, because of different cultures, purposes, etc. Jer is always stunned and dismayed at my computer “desktop” he organizes his materials completely differently.

If the clouds are categories for a blog, they represent essentially pre-defined categories into which posts are placed (true, new ones can be created at need) hence assigning categories is “classification.” Some of those clouds are, however, displaying prominent words contained in posts, and hence represent emergent categories, “clusters” that are similar by virtue of containing the same words.

By “problematic” I meant that the idea of what counts as similar is not obvious, and depends on culture, purpose, etc. Given a wool cap, heavy mittens, gardening gloves, and a sun hat, you might organize into headgear and handgear categories, or winter and summer categories. Neither is necessarily wrong or right.

thanks for the fascinating AI way of distinguishing categories — “classification” and “clustering.” The latter reminds me of maps and file keeping, both of which can be widely (and wildly) varied, depending upon the intent of the mapmaker or filekeeper.

For example, I have seen (visual) journal “maps” which record, in drawings, scenes in the “path” along which the artist walks each day. These bear little resemblance to the map of the terrain or the roads, but are equally valid in their way to the road atlas we all know.

Clustering also makes me think of those “clouds” that image groups of things that change: for example, “categories” in blogs that garner the greatest number of hits show n as “clouds” — the larger the typeface, the more popular the category, sizes which could change at any given moment.

In both these cases “clustering” seems like a more appropriate kind of category than “classification,” although one might derive a classification from such clusterings.

I suspect, given your comment “more problematic than it seems” that you are working on far more difficult to resolve clusters or seeming clusters of things. I’m thinking of those sets of overlapping ven diagrams that I used to love to suss out when I was an undergraduate. But what a wonderful distinction you have presented me with here — something to ponder while I paint.

Thank you for bringing up an absolutely crucial point! Categorization is a key aspect of almost all cognition. The act of assigning a color name to a color patch, a syllable to a sound, or, at a higher level, naming an object or hearing a word, are examples from perception. Much of our reasoning in other areas is equally based on similar generalization/labeling/categorization (think politics, for example). This strength has the obvious weakness you note; it allows us to bring past knowledge of something to bear, but it can also inhibit accurate observation of what is new or different. In the area of visual art, this has always been the meaning for me of the title of Wechsler’s biography of Robert Irwin, Seeing is Forgetting the Name of the Thing One Sees (I can’t remember if that’s a quote from Irwin or elsewhere–anyone know?)

On a technical note, in AI we distinguish two kinds of categorization, classification of things into pre-defined categories (e.g. given by language or definitions) and clustering of things into groups based on similarity (more problematic than it seems), which groups then constitute the “discovered” categories. My work, by the way, involves doing both kinds of categorization with text documents.

Since our local paper prints NY Times articles, generally about 3 days late, and I also get the Sunday times, I’ve read the article you refer to at least twice now. The conundrum that I found most interesting was that language indeed enables people to identify and work more efficiently — quicker and more accurately. So if you knew two different kinds of blue because your language had words for them, you could find the matching colors in sets that contained various blues kinds more readily. You would work more quickly and more accurately.

However, something of the obverse is also true — that once you’ve acquired the language - say bump and beet for green/blue and red/blue — you tend to lump all blues into one or the other category. You tend not to see other possibilties.

This latter problem has repercussions for students of the arts — we are taught to see efficiently — “tree” or even “oak tree” so we come to think of “oak tree” instead of “the tree that has the faintest blue color in itse early green leaves” (I made that up).

It’s what the drawing teachers tell you from the git-go — drawing is first and foremost learning to see –that is, to see with the inefficent, unlanguaged mind. In that sense, it’s very like Melanie’s Taking Dictation from the Muse. You have to take that dictation before you decide what the shape or color you’ve been given is called.

This seems intuitively obvious, but I loved it that they found a way to research the question. That they could go on and discover that when the language was interfered with, the Russians lost their advantage in dealing with color was striking.

By the way, I’m glad the grizzly didn’t get you — and that you got his footprints.

I hope that you will do a sequence on tracks in the snow (before the Grizzlys’ get you). I love your earlier tracks in the snow http://www.artandperception.com/2008/02/art-walk.html

In my limited bayliwick, artificial intelligence meant neural network modeling. I have to think about its wider application to perception

Jay,

I think the bear was walking a bit faster here, so the large hind foot steps almost past the front foot. Elsewhere, they’re nearly in registry, a common walking gait. Size-wise, the hind foot is the same length as mine but much wider, and the stride is 4.5 feet.

melanie,

I think you’re right that we can learn to bring more to consciousness, and that is likely to improve the process (and result), not to mention make it more satisfying.