Picture, Depiction and Deception: Why Data Visualisations are Cultural Images
Data visualisations are usually created by computers — but they are not technical images. Every visualisation is an interpretation of data and as such a cultural image.
This essay is the sequel to a blog posting that I wrote more than two years ago. The subject is still relevant so it makes sense to re-visit the text and elaborate on the issue. My central point is that a lot of the current debate on data visualisation is not differentiated enough and is based on an antiquated idea of an “image”. So let’s start at the same point as last time:
In the excellent book Design for Information by Isabel Meirelles, I came across a quote from Ben Shneiderman:
“Like Galileo’s telescope (1564–1642), Hooke’s microscope (1635–1703), or Roentgen’s x-rays (1845–1923), new information analysis tools are creating visualizations of never before seen structures. Jupiter’s moon, plant cells, and skeletons of living creatures were all revealed by previous technologies. Today, new network science concepts and analysis tools are making isolated groups, influential participants, and community structures visible in ways never before possible.”
This is a great quote and I very much like the vision behind it. It is obvious that visualisation is an essential strategy to deal with large sets of data. Turning abstract data into a concrete image transforms data into something we can perceive and relate to. Data visualisation translates the unimaginable into a tangible and sometimes interactive image. So the power of visualisation is something I strongly believe in and I totally agree with.
However, I think Ben Shneiderman — and many others in the datavis community — misses an essential point about visualisations. He draws analogies with technical devices. The comparison with telescopes, x-ray and microscopes implies that data visualisation is mainly a technical challenge — a problem of magnification and resolution. Optical devices magnify small or distant objects. Lenses and mirrors direct and focus light waves and are thus revealing existing physical structures like plant cells or galaxies. The assumption is that these devices generate technical images that are objective and show aspects of reality.
So the quote by Ben Shneiderman suggests that data visualisations are technical images. It suggests that the visual representation of complex data structures can be directly derived from the data itself. It suggests that by using visual encodings as data-lenses, you can create an objective image of the data.
I believe that these assumptions are questionable and misleading.
There is a strong argument in media theory for treating all images as symbolic. This includes technical images that were generated by a device. They are also a representation of an entity and need to be decoded and interpreted. So even technical images are not a copy of reality.
Which brings me to my main argument. I believe that data visualisations are not technical but cultural images. As we will see, this distinction is a gradual one — but it is important to note that data visualisations simply cannot be understood as an objective excerpt of reality.
For the further discussion, I would like to refer to Vilém Flusser. He was a highly influential media theorist and philosopher and he dedicated a lot of his publications to the role and significance of photography. In his book “Towards a Philosophy of Photography”, he develops a theoretical framework for understanding the relationship between the photographic image, the apparatus and the photographer. In particular, he questions the objectivity of the photographic image and the apparent control of the photographer in regard to the image-making. But his writings are not limited to photography — he explicitly refers to “technical images” that also include computer-generated visuals.
I believe Flusser’s book holds a number of relevant lessons for the debate on data visualisation that is happening right now. In this context, the parallels between the photographic image and data visualisations are intriguing. Many of the theories, concepts and ideas that Flusser developed for technical images can be applied directly to data visualisation. The following quotes are all taken from his book “Towards a Philosophy of Photography”.
From Technical Images to Cultural Images
Flusser’s definition of the technical image is pretty straightforward: “The technical image is an image produced by apparatuses”. But he directly points to one of the most relevant and culturally important aspects of technical images — the fact that they are often mistaken for the reality they depict. “They [technical images] appear to be on the same level of reality as their significance”. This is a fundamental insight that not only relates to artistic, professional or hobbyist photography, but also to scientific images. Images are always symbolic. Or — as Flusser puts it: “This lack of criticism of technical images is potentially dangerous at a time when technical images are in the process of displacing texts — dangerous for the reason that the ‘objectivity’ of technical images is an illusion. For they are — like all images — not only symbolic but represent even more abstract complexes of symbols than traditional images.”
Important for our debate is also Flusser’s distinction between technical images and traditional images: “With traditional images, by contrast, the symbolic character is clearly evident because, in their case, human beings (for example, painters) place themselves between the images and their significance. Painters work out the symbols of the image ‘in their heads’ so as to transfer them by means of the paintbrush to the surface.”
This distinction is important insofar as it introduces two kinds of image-making. The technical image is produced by an apparatus, the traditional image is crafted by a person. This distinction has nothing to do with qualities. This is not about whether a traditional image is better than a technical image. But the distinction has a lot to do with the way these two different kind of images are perceived. A traditional image is perceived as real in itself — a technical image is perceived as a window on reality. But obviously the technical image is also (and primarily) real in itself.
It is clear that the traditional image is a cultural image. But following the above line of thought, it becomes apparent that technical images are also cultural images. The technical conditions of a optical apparatus do not make an image objective. The use of technology for image generation does not remove the layers of meaning and significance of the image itself. An image is an image. Even if it is produced by a machine.
I think at this point it becomes quite clear where I am heading. Data visualisations and photography share a number of similar attributes. They are both technical images that seem to be an objective representation of reality. But on closer examination it becomes clear that both are essentially cultural images. And they should be treated, interpreted and decoded as such.
The statement that data visualisation creates cultural images is important in two different ways.
First, it is important for the interpretation of data visualisations. As we have seen, they are more than a technical representation of data and facts. Data visualisations are cultural artefacts and need to be interpreted accordingly. And as all other cultural images, there are different ways to read and interpret them. So instead of simply talking about “insights” I would argue for a hermeneutical approach when dealing with visualisations. Visualisations should not only be interpreted on the grounds of their technicality but also in terms of their culturality.
Second, it is important for the creation of visualisations. I am very much aware of a certain suspicion regarding well-designed visualisations. I often encounter the prejudice that aesthetics is just getting in the way of proper data representation and that design just obfuscates objectivity. I strongly believe that this is not true — simply because the objectivity of an image is an illusion. So instead of negating the role of design in the creation of data visualisations, we should embrace and improve it.
I know that many colleagues in our community are uncomfortable with the notion of human interference with data visualisation. If designers “place themselves between the images and their significance” as Flusser puts it, the colleagues fear that the objectivity gets lost. But it is important to point out there was no objectivity in the first place. Actually, the lack of objectivity goes even deeper. As Johanna Drucker has pointed out in her book Graphesis: “Data are capta, taken not given, constructed as an interpretation of the phenomenal world, not inherent in it”. So the collection and generation of data is already culturally biased. And if we stay in the analogy between photography and data visualisation, one could point out the data visualisers do not only use an apparatus — they actively create new ones.
All this does not mean that data visualisation is random, artistic and meaningless. Quite the opposite. Just as photography, visualisations can make strong statements about reality. However, we need to be aware of the fact that both the data and the visualisation are constructions. Furthermore, a visualisation is an artefact in itself and not just a neutral conduit for data. The visualisation helps us to relate to the data — but it obviously shapes the way we perceive it. Just like with a photo, we should not only recognise what is depicted — we need to recognize the picture itself.
Photography and data visualisations can explain, enlighten and entertain. Both represent facts, stories and events. As such, both are meaningful and useful. And both photos and data visualisations are essentially cultural images.