Cultural Analytics lectures by Manovich in UK (London and Swansea), March 8-9, 2010

lecture 1:
date: March 8, 2010
time: 2pm

Castle Lecture Theatre, London Road Building.
Keyworth Street.
London South Bank University, London

map of the campus

Note: Kate Hayles will lecture first at 1pm

lecture 2:
@ The Computation Turn international workshop
date: March 9, 2010
time: 3:45pm

James Callaghan Building
Swansea University, Swansea, UK

Note: Kate Hayles will delver keynote lecture at 9:15am

Lectures outline:

1 rize of visualization of culture
information visualization as a key new techique for representation appropriate for information society - and also (if it's interactive) a new technique for thinking
simultaneos development of www and infovis in early 90s
2004-: Processing + availability of large data sets + APIs

2 what is visualization?
reduction – use of graphical primitives
layout which reveals patterns

humanities visualization – creating layouts from actual media objects, as opposed to representing them via graphical primitives (no additional metadata is added)


gather (Time montage)

highlight (Anna Karenina, Hamlet) / continuity between a "full" media object and visualization/diagram

sample (gameplay montages, Time slice, folder: Vertov sample)

calculate (folder: Vertov averages)

is this visualization?

Culture visualization today - usually visualization of metadata about artifacts and process ( see visualization of social networks at In contrast we want to show relations between actual artifacts .
Shall we call image graphs and other techniques which involve arranging images (or their sample) in particular ways "visualization"? If we assume that the core principle of visualization is not reduction and (therefore) use of vector primitives such as points and lines (which also form the language of diagrams and sketched in art) but rather the arrangments of elements in a layout which shows patterns/relationships betwen elements, then the answer is yes.
Note also that if normal visualizations consist from symbolic signs (vector elements which stand for the objects and which signify through an agreed code), "direct visualization" uses objects themselves. Therefore it does not involve signification or reduction. (in that direct visualization parallels figurative art which also does not signify real-life objects through signs but represents them in detail, and which also communicates meaning through layout - traditionally called composition. Buy while the goal of figurative art is communication of meaning, visualization only shows patterns - it's up to the researcher to interpret them as meaningful.
20th century cultural theory and media art often focused on close reading of media - zomming in and slowing it down (think of "24 hour Psycho") But after media explosion (social media and archives digitization) we need to learn how to zoom out, fast forward, compress (visually), summarize - so we can make sense of vast cultural landscapes.
Visualization dumensions:
functional - aesthetic;
using established familiar methods - inventing new methods.
Lots of artistic visualization is purely aesthetic - I.e. It's not trying to reveal patterns but only uses visualization principle of deriving images from data to make abstract art.
Where does culturevis fits in? Ideally we want to be both functional and aesthetic ( find forms which best express the particular artifacts and cultures). (examples of my diff visualizations of vertov shot lengths). However if we want to be able to compare many artifacts, we need to use the same technique.

We can use Pierce's signs classification - icon, index, symbol, diagram - in relations to visualization. This scheme describes the types of relations between a sign and a referent. But we can also add a new dimension which describes "how much" a sign represents, so to speak - does it show more or less information about the referent, and what kind of information?
Normally, we think of visualization (i.e. a representation which uses vector elements) and a realistic image (a photograph or a painting) as opposites, one representing the bare minimum structure of an object and the other representing the object's sensorial appearance in detail. But if we look at them using our new "how much and what is represented" dimension, we see that there are all kinds of intermediate cases. Therefore realistic representation and vector visualizations are just the extremes of a continuos dimension.
At the same time, there is a qualitative diff between Marey's chronophotographs which diagram motion and visualizations which may represent relations which are not directly visible (such as economic data) - so on this dimension indexical diagrams (such as Marey) and visualizations are diff categories.

3 cultural analytics: data analysis + visualization
add metadata (via manual annotation and/or automatic analysis), then visualize

Some of the key advantages of this method:

automatic analysis + “image graphs”:
1) understanding meaning and/or cause behind the pattern (for instance, a repeating movement pattern in Vertov is sometimes due to parallel montage, but in other cases its not)
2) revealing additional patterns (for instance, changes in communication techniques across Time covers)

visualizing (analog / continuous) cultural dimensions which can’t be adequately described with language (which uses discrete categories)

visualizing continuously changing qualities over time
particularly useful for 21st century motion graphics and films, but also opens a new direction for understanding 20th century cinema / rhythm / time series analysis
example: movement pattern in "11th Year"
example: Time covers (change over a longer period)
example: US, French and Soviet 20th century films: comparing shot lengths

4 beoynd categories: aggregation without structure
Latour’s arguments: tracking and representing aggregates of objects, without the need to go to another level of model, structure, etc. (“Tarde’s idea of quantification” in Mattei Candea, ed. The Social After Gabriel Tarde: Debates and Assessments.)
Extending these arguments to culture visualization:
from categories (i.e. genres, historical periods, etc.) to a multi-dimensional space of features where we can see objects forming distributions and clusters

Latour: "we should find ways to gather individual "he" and "she" without losing out on the specific ways in which they are able to mingle.... But never in some overarching society. The challenge is to try to obtain their aggregation without either shifting our attention at any point to a whole, or changing modes of inquiry."
Similarly, we need to get away from the standard distinction betwen an individual cultural artifact and larger categories be they "cultures," "genres", etc. We need new ways of studying aggregates - bottom up ( which is what data analysis + visualization make possible.) Here the ability of computers to keep tracks of large volumes of data and navigate through the data at arbitrary zoom levels without the need for aggregation, simplification or averaging becomes crucial. (Rather than seeing Manga space as a map consisting from a few ditinct regions, we can show every single page and observe patterns of contonuos change across this space.) Modern computing allows us to analyze, record and represent "individual variations" (Latour) of billions of entities - thus making possible for social and cultural sciences to become truly scientific in a way still inaccessible for natural sciences (because their entities - such as gas molecules - are still too numerous for computers to represent individually and therefore they have to use general models to represent their structure and behavior.
"Structure is what is imagined to fill the gaps when there is a deficit of information as to the ways any entity inherits from it's predecessors and successors." ( we still have the problem of mapping exactly this information. However we can at least start by refusing structures such as "genre", "period" etc.)
"Individual variations are the only phenomenon worth looking at in societies for which there are comparatively few elements."

visualizations_vs_categories folder
example: Manga analysis
example: modernism folder

Latour: "through the ease with which we can navigate a datascape, we manage to interrupt the transubstantiation of the aggrgate into a law, a structure, a model and complicate the way through which one monad may come to summarize the "whole."
"But the "whole" is now nothing more than a provisional visualization which can be modified and reversed at will, by moving back to the individual cimponents, and then looking for yer other tools to regroup the sane elements into alternative assemlages."
"The whole lost its provilleged status: we can produce out if the same data points, as many aggregates as we see fit, while reverting back at any time, to the individual components."

Do we still need discrete categories?