A data visualization that doesn’t use data


How do you create a data visualization without data? Is it even possible? To find out, Enrico Bertini and Moritz Stefaner, from Data Stories, interviewed Dietmar Offenhuber about the art of indexical visualization.

Data Stories: What is indexical visualization?

Dietmar Offenhuber: It is hard to describe because indexical visualization is a little bit of a paradoxical thing - it is visualization without representation. I used to describe it as a collection of strategies that allow us to more or less directly oversee a phenomenon without translating it to a symbolic visual language such as charts and maps.

So there's no real data involved in terms of numbers or databases?

Exactly. So that's the second paradoxical thing because indexical visualization is visualization without data (since data is also a form of symbolic abstraction or representation). So, this idea of nonrepresentational visualization is radically different from what we understand under the rubric of data visualization. Maybe the best idea would be to start explaining it through an example. Think of a wind tunnel. We cannot really see the airflow, but if we add some smoke very precisely we start seeing lines and the lines indicate the movement of air and if you do it in a very accurate way, then we have a beautiful 3D visualization. But those lines that we see, those lines of smoke are not really representations. They don't really stand for something else. They are part of the phenomenon that we are actually interested in. So, this is what indexical visualization does. It indicates, it points to a phenomenon, and it frames it in a certain way so we can see it. The term is a reference to Charles Sanders Peirce and his semiology where he said that the index shows something about an object because it's physically connected to it.

The index finger is basically what we point with, right? So is that the root of the term?

Exactly, exactly. So it's really about this act of pointing, and that's the second reason why it's hard to define, because indexical visualization is not really about what it is, but what it does. It is a very performative notion. A I said earlier, it's a collection of strategies. So what are some of these strategies that we can use to produce indexical visualizations? One of the most important aspects of it is connected to producing traces. That's what happens in the wind tunnel when we add the smoke as a visible marker. Or, if we go to the hospital and we do an x-ray and we have to swallow some radioactive trace of material so that those things show up on the x-ray. Or, if you like watching CSI and those different crime shows, it is like making the latent fingerprints visible. Those are all different acts of producing traces to make something visible.

A second strategy would be constraining a phenomenon - taking away degrees of freedom. If you think about a thermometer, it has liquid inside, and the liquid expands with temperature. But you wouldn’t be able to see unless we make sure that it can only expand in a single direction in a very thin glass tube. This act of showing something by constraining it is a very important element. A third strategy is related – it is about framing the phenomenon. This means that we have to add some references to comparison. The New York artist Natalie Jeremijenko did a number of projects along those lines. One very simple one is where she used dust masks that have a grayscale printed. And if you go through air pollution, the dust mask will turn gray. By comparing it with this reference, you can see how bad it is.

So the dust mask becomes the visualization.

Exactly, exactly.

Just by adding that legend, or that scale to it, it suddenly becomes a visualization. That's a crazy thought because it sounds almost like, whoa whoa whoa, everything's a visualization. Do you think there's a limit there? At some point everything might become a visualization of something if you see it like this, right?

That's another hard thing - the problem of specificity. But in this case, what the scale does is actually what differentiates just any phenomenon from something that becomes data. Take the cyanometer. It is a ring that has a blue colour scale on the outside to measure the blueness of the sky. And by that you can basically estimate the humidity of the air. So without this frame of reference - “how blue is the sky?” - it's kind of a meaningless question, unless you start adding a scale so to make it systematic. Then there is a moment when a kind of a systematic observation becomes discreet, becomes data, which is a big difference compared to data visualization where data is already something that we have in the beginning before we start. With indexical visualization, we are really going through this process of interpretation and discretization of ourselves.


Image: Great Basin Cyanometer by Virginia Catherall. Source: BLM Nevada.

So this is actually one very important difference between data visualization and indexical visualization. When we look at data visualization, we usually use data as a starting point, but in indexical visualization, data is the end result of our interpretation.

I think one thing that you briefly mention is this idea of minimizing the distance. By being able to visualize and talk about numbers and statistics, we can understand some phenomenon much better, right? But when you use these numbers as a way to communicate information to others, there is sometimes, I would even say often a very big distance between these numbers and the original phenomenon, and visualization is kind of like, a way to kind of try to reduce this gap. Indexical visualization seems to be a way to get closer to the origin, right?

I think there's a lot of concern about narrative strategies in visualization, telling stories with data, data stories. But, I think the second aspect is the experience of data and of information. A story is when someone else basically guides you and explains something to you, but when we think about traces, we have to put it together ourselves. So this is another very interesting aspect of it and something that emotionally really affects us because we are really experiencing the phenomenon ourselves. I think there's a very beautiful example, which is not really indexical visualization in the narrow sense. That is, it id kind of a physical representation, a physical phenomenon, more in the extended sense. It is Kamel Makhloufi’s visualization of Iraqi casualties during the Iraq War, where I think the data set was originally from WikiLeaks. He visualized it in the most simple way you can imagine; by using exactly one pixel for one person who died. And so the result is a visualization that has a fixed resolution, so it's in a way, unique. You can't really scale it. You cannot really transform it a lot. You have a kind of one-to-one relationship between the visualization and the data.


Image: Kamel Makhloufi's visualization.

Do you have any suggestions on how to start or guidelines? My sense is that there are at least two ways to do physical, indexical visualizations. One is going around hunting in the world, trying to find out phenomena that have already been recorded into something. How do you get started with this? Do you have any suggestions that you can give? 

One very easy answer is basically we have now our documentation of the Indexical Design Symposium. There are a lot of videos of different practitioners, researchers; we have criminologies, forensic scientists, biologists, historians, artists, who basically covered indexical practices in the broadest possible way.

You also have a Pinterest board collecting lots of examples, right?

Yes, I do. It might not be self-explanatory because sometimes you might ask yourself, why I think this particular Pinterest or these images are indexical. There's also a paper that I wrote with my colleague and friend on indexical visualization. He looks into the topic of the signature, “what does it mean to see signatures in a particular context?” So I think there are a lot of things to explore.

This piece is an edited version of an interview that was originally broadcast on the Data Stories podcast by Enrico Bertini and Moritz Stefaner. Listen to the full podcast here.