“Are Some Things Unrepresentable?” by Alexander Galloway

There’s a software architecture pattern known as MV* (the asterisk serves as a stand-in for various permutations of MV(something), e.g. MVVM, MVP), the most common of which is MVC, which stands for Model-View-Controller. It sounds a bit jargon-y at first, but the idea behind it is rather simple: any application can be thought of in three disjoint parts. We can take, as an example, a to-do list application.

For this application, there is the Model, which houses the data. In the Model, the tasks might be represented in a data structure like an array. Each task might be as simple as a string containing the description of the task and may be as complex as an object which has additional attributes like a due date, whether the task was completed, etc. The View is how that data is represented to the user, and can be most typically thought of as being the same as the UI. The Controller is the part of the application which mutates the data—for example, on user input.

This framework provides a tidy way of thinking about software because it provides a distinct separation of roles, and highlights one of the key points Alexander Galloway makes in “Are Some Things Unrepresentable:” that “data have no necessary visual form.” After all, we are able to keep data and its visual representation completely separate. This realization is a bit of a Modernist approach to interpreting data, by understanding that a representation of data is not the same as data itself, a la Magritte’s pipe.

Galloway goes on to discuss a corollary of this idea, that data have no necessary visual form, which is that “data have no necessary information.” This is another statement that at first glance seems a bit counterintuitive, but also makes sense. Data (thought described in terms of qualitative or quantitative, scalar or vector, etc), carries with it no meaning. It must be given meaning (literally, to be “given form”) in order for it to have any consequence. People’s entire livelihoods are predicated on this fact. For us to draw conclusions from data, they must be arranged in neat little rows: we have to order them in a time sequence or cluster them based on similar characteristics or compare groups of them to each other.

This is also nothing new. There is, of course, that well-known line by Mark Twain about “lies, damned lies, and statistics.” Data, because it has no form, can be made to take any form. The veracity of the stories these forms tell now are now no longer related at all to the veracity of the data itself.

The above two statements, if nothing else, should be enough to dispel that stubborn idea that the people who spend their lives thinking and dealing with information (software engineers, especially) are somehow absolved from the conclusions drawn from and systems built on top of that data. The data itself may be completely descriptive, but what we decide that description to be lies entirely on our judgment.

P.S. For sake of coherency, I have neglected to talk about Galloway’s second (and equally important) thesis in “Are Some Things Representable,” which is that “only one visualization has ever been made of an information network,” a statement which seems contradictory to the notion that data have no necessary visual representation (after all, doesn’t the fact that a certain type of data is only ever represented one way mean that it is equal to that representation, they are one and the same?). Galloway has some good discussion about why these statements can be simultaneously true, but I’d like to propose that it’s entirely possible that we have only ever come up with a single visualization of information networks because that is the only representation that we can make sense of. To answer Galloway’s question in letter and not in spirit, there are indeed things which are “unrepresentable” to us: anything in the fourth dimension fits that bill. And while there are attempts to translate four-dimensional objects into three (by chopping off one of those dimensions), it is predicated on a loss of information. It is very possible then that the one representation of an information network we are able to make is just one of many possible representations that exist—just not to us.