[NetBehaviour] Exploring Tate Art Open Data 2

Rob Myers rob at robmyers.org
Mon Feb 10 01:19:29 CET 2014

On 09/02/14 03:15 PM, Michael Szpakowski wrote:
> This seems a measured and fair response. I wasn't having a pop - rather
> rehearsing my perplexities out loud..

Yes that's how I took it.

> I *do* think there are some further issues, especially in the case of
> historical collections, about what is *available* for the data to be
> extracted from -

Certainly this is a visualisation of "the world as seen from the Tate


And the Tate Collection contains...the art collected by the Tate over
the years under changing curatorial fashion and competence (contrast
post-Impressionism in French collections and Action Painting in American
collections with the examples of each in the Tate and it's not even funny).

I'm not claiming more in these posts. But I didn't start by positively
articulated my assumptions about the nature and limits of what I am
doing. I should have done that.

Moving beyond the Tate, one of my next projects is to recreate then
build on some similar visualisations of Wikipedia's articles about
artists and movements. Which I can then contrast with the Tate data.

There are a couple of reasonable questions here. Why contrast existing
privileged institutions in order to (apparently) police historical
consensus, why not try to build something new? And why take the
pre-existing units of those institutions' art histories as the basis of

Firstly this is provisional work with a political motivation. I'm
learning how to do this (everyone is), and I want art open data to be  a
success. So for these reasons I'm visualising a high-profile data
release with a well understood technical and ideological structure.

Secondly I'm working on techniques to discover commonalities and
groupings of artists and artworks that do not rely on pre-existent
imposed categories. The first part of that is going to be in the next post.

> and this suggests the danger of circularity here:
> "I've found no great surprises in the Tate collection data, which I take
> as a confirmation of how well existing models have done".

That conclusion disappointed me, though, because I was looking for clear
surprises as they would be a means of validating this approach. I
earlier described what work differences in qualitative or quantitative
data from prior models would give art historians to do. By "models" here
I mean art historical narratives rather than other data. So this is more
"despite my best efforts different approaches appear to agree" than "A
because B because A".

"Tate categorized this data with these labels, and examining the data I
found them" would be trivial. But when the labels are pre-decided, their
relation and frequency can still be interesting. And if those labels are
systematically mis-applied (as judged according to some external
criteria), that is also interesting.

I'm not trying to play fast and loose with a kind of quantum
insititutional critique / historical insight that is the one if it fails
to be the other. It's possible for this to fail in both modes. ;-)

- Rob.

More information about the NetBehaviour mailing list