Review Essay: Interpretation versus Cultural Analytics

Lev Manovich, Cultural Analytics. MIT Press 2020

Joanna Drucker, Visualization and Interpretation: Humanistic approaches to display. The MIT Press. 2018.

Lev Manovich’s Cultural Analytics presents in illustrated textbook form the scaling up of media studies to embrace big data analytics and sources. It proposes a refinement of software studies under a more ambitious rubric. Software as culture. What patterns can be found in, for example, the entirety of the photography collection of the Museum of Modern Art in New York? Or, posts on Instagram with a certain tag? This pattern recognition is computer assisted and thus a quantitative method, even if packaged in apps or functions to catalogue images and texts according to colour, the frequency of words or points of origin. The biggest shortcoming of such textbooks is that they tend to lag the cutting edge of developments by five to ten years. At the heart of this approach are pattern recognition algorithms.

The book is addressed directly to the reader in an encouraging tone, “Let’s look at some examples of sources of data…. You can also start with particular cultural genres and check if there are sites listing large numbers.” (p. 89). While the use of the APIs (Application Programming Interface) of various platforms is more sophisticated, the notion of Instagram as a source of data is evident to most 10 year olds, who of course could benefit from the admonishments that use of this and other platforms such as Eventbrite are characteristic of European and Anglo Saxon societies (and in 5-10 years will these brands even be remembered?). The argumentation is marred by US parochialism; even within North America, not far from New York, some of Manovich’s examples (e.g. stoplights must feature three lights of the same size p. 194) are contradicted by the why the objects are used or designed in Québec. A limit is the methodological focus on how to scale up the analysis of such large data sets shifts from research to “processing” at volume. This distracts from not only from theory building but from ethical questions. Perhaps that is why Manovitch’s text gives the impression of teaching cultural pattern recognition for software scientists.

Cultural analytics appears to be the made-for-TED talks social science analogue to systems which purport to recognize the facial patterns of emotion, and thereby detect delinquents and non-normative intentions and mind sets. Systems such as those deployed by Taigusys in countries without privacy laws such as China (Guardian Jan. 22 2021 ) . Such systems rely on machine learning and detecting patterns of emotional facial expressions in the geometry of eyes, brows, and lips. They are deployed to monitor livestock and humans. Amongst people, however, a poker face is always possible as a facade, even in moments of great stress. The driving of dissent underground is the correlate of the built-in obsolescence of such systems.

This is effective for certain purposes, first and foremost those objectives set by those who control and have access to big data. It is worrying to see the evolution of cultural research in the direction of the State and corporations that regard such data and patterns as proprietary knowledge because it can be leveraged for profit. The limits of pattern recognition is the plasticity and diversity of people. The nature of humans is always to adapt beyond stereotypes in pursuit of the better or just the novel. This is no more true than in the domain of culture. Like the old sociology of class, the impetus of large scale pattern recognition is to (1) detect regularity and (2) cleavages between significantly large groupings which are then analysed for their causal power as (3) determinants of the pattern, (4) reproducing it in other fields. However, in the final analysis, such patterns are by definition banal. The addition digital tools make is that it is easier to identify breaks in patterns (2) especially over time. It would thus be useful to have more in the book on temporality and even to expand the theoretical sophistication to think of cultural topologically as a time-space and medium (see work by Celia Lury and others in Theory Culture and Society).

I doubt that Indigenous cultures in North America would agree with the conception and rubric of “cultural analytics”. Culture is neither digital media nor software. Culture is more than a reduction to patterns or even a language of patterns. It is instead the drift of patterns or topologies of patterns. He places his faith in the ability of quantitative analysis to “‘explain’ the seemingly elusive, subjective, and irrational world of culture?” (p.248) However, decades of cultural studies shows this is doomed to reduce the cultural to isolated patterns, like reducing a body to a stick figure.

The Index does not help with metatheoretical concepts such as place or space (nor does the ebook platform I used allow searching in books), nor does Manovitch remark on it, but it is clear that place is a key connector of variables and an anchor of patterns. Manovich seeks to foreground other anchors and indicators such as colour, transparency, texture and shape. Images themselves are treated as spaces within which geometric relations can be discerned.

Whereas “traditional” graphical visualization is said to be a predominantly spatial representation requiring reduction to points, lines and geometric elements as icons for “real-world objects and their relations,” Manovich proposes “media visualization” by which no such reduction is needed. The emphasis of “real-world” betrays the positivism that lurks in Manovich’s approach. It is not surprising then that he mistakes the immanence typical of all diagrams when they are geometric projections from three-dimensional phenomena to two-dimensional surfaces, whether a poster or a screen. The introduction of temporal variables and animation of patterns (change over time) is the more significant distinction of media visualization.

Has Lev Manovich read Joanna Drucker’s 2018 Visualization and Interpretation? It is a much stronger book. It is a critique of the preoccupations of Manovich’s textbook on the representation or visualization of networks (compare the examples on visualcomplexity.com). What is needed is a philosophy of representation, not a technique. The reason for this is that representations frame our understanding of what is significant about a phenomenon or a situation, something that may not capture or may not change to keep up with the evolution of new developments in a situation over time. In her examples, Drucker critiques the page-like, two-dimensional bias that dominates digital displays which are easily capable of presenting three-dimensional spaces of information – and easily animated, I would add. She critiques the lack of analysis of what were once called feedback loops between our representations and action.

Having borrowed techniques such as diagrams and maps, Drucker asks whether these tools may be subverting attention to epistemology in the humanities and thus undermining critical projects. This is because these visualization tools are based on assumptions that knowledge is observer-independent and fixed rather than interpretative. I would add that knowledge is worked up in the cerebellum anew each time information is assembled into patterned knowing. Visualizations strive for a clarity which fits poorly with the perspectival or ambiguous qualities of humanities knowledge – and I would add that this is a quality of all knowledge. Facts are never truly fixed in science but always provisional and dependent on the next experiment.

I’m intrigued by how this might apply to W.E.B. DuBois’ Paris Exhibition diagrams showing the status of “The American Negro” included in this review.

-Rob Shields (Univ. of Alberta)