I paint data

Whenever I lecture about our cultural analytics work (computational analysis and visualization of large cultural data), somebody in audience always ask: is this art? Many people who were fed the strict diet of bar chart and pie charts do not know what to think about our visualizations which show large image collections sorted by various visual attributes. Like the individual images they include, these visualizations are colorful, sensual, and aesthetic (i.e. they appeal to senses as opposed to only cognition.) So are they indeed art?

Early in my life i was trained in realist painting and drawing. I had an art teacher in Moscow since I was 12, then went to Moscow Architecture Institute and continued painting for a while after I moved to NYC in 1981. Is it possible that my recent work with visualization of large image collection is a return to painting? Yes - but not just any kind of painting.

Realist art does not capture the world mechanically. Instead, it focuses our attention on patterns in visible world. It highlights some patterns and disregards others. The relations and structures which are highlighted may be color combinations, relations between figures, face expressions, gestures, and so on. These patterns often support the "meaning" of a painting, but not always. (Especially when "modern art" starts to develop in the middle of 1850s, the visual patterns become gradually "liberated" from supporting semantics and meanings.)

Similarly, data visualization reveals some patterns in the data. To do this, data is translated into visual representations, organized in particular ways, and coded using color, shading, size or other graphic attributes.

Both realist art and data visualization are constrained by the data. That is, they don't simply sample some bits of data and arrange in arbitrary combinations. Realist art preserves important aspects of visible reality such as relative sizes of objects and convergence of lines under one point linear perspective, the effects of different types of light sources (e.g., sun casts shadows), etc. Similarly, data visualization preserves (and makes visible) relations in the data set - in contrast to "data art" which may use the same data sets and same representational techniques (e.g., computer graphics) to generate abstract compositions which do not preserve any of the relations in the data.

I can now come back to the question: are our visualizations art? if they are, they belong to the tradition of realist art.

In other words: I "paint" data. But my brushstrokes, colors, and composition are constrained by what is this data. The pleasure which I derive is is following these constraints.

Art always thrived on constraints. Visualization is no difference. Thus, while some of us choose to become data "abstractionists", other prefer to become data "realists." And this is where I am.

Lev Manovich.