In a previous post, I outlined how creating an image, initially for a competition, also illustrated how that visualisation process often became an analytical (flânalytical?) technique. Having been inspired by the @metropologeny city maps, as I began planning the vis, it always struck me that tweets seemed to naturally fit the mainly rectangular shapes of the buildings on the map. In being drawn towards the tweets however, I wondered about the other data sources which were part of my study, but temporarily parked that aspect until I’d resolved the technical aspects of producing the image. Now, with that task completed and the image submitted for the competition, I now turned back to the other data. How might blog posts or interviews also contribute to the vis?
Before delving into how I moved forward, perhaps it might help to rewind somewhat and look at how the map was built in the first place. This animation shows the different stages
- One of @meteropologeny’s maps was imported into Inkscape and created as a base layer onto which other layers were added.
- Tweets were dropped on top of the district blocks. Fitting them to the size and shape of the buildings was possible, but I felt they began to lose their inherent ‘tweetness,’ so left them as simple rectangles. This meant I needed to mask out the underlying buildings …
- Which is where the idea for using the Twitter bird came from, although …
- It was important as a flâneur not to lose the sense of cityscape, so the next stage brought that back and introduced the different districts or quartiers as ways to categorise the tweets.
- As explained previously, these tweets were arranged into different quartiers …
- … with the whole street plan reintroduced so one might imagine a walk around the city whilst encountering the kinds of activity seen when wandering the Twitter timeline.
- The street names are formed from blog post titles, each street intersecting the quartiers which the contents of the post exemplify.
- In the final stage, for simplicity, the tweets are wiped and replaced by illustrative snippets from the blog posts on adjacent streets.
The ‘layers’ feature in Inkscape is useful for hiding and revealing different features of complex images as you build them up, so producing the above animation wasn’t a big deal. What I didn’t include in the animation was the complete image with all layers showing. The result was so messy, it became difficult to make out any detail. But then it struck me how similar that was to what the researcher faces at the beginning of analysis with all the data in one unholy mess. In the same way a flâneuse can’t view the whole city at once, by walking from street to street, she can begin to build up a sense of the whole from the parts she sees. So too with the data and so too with the image; all the parts are intertwined. What Inkscape allowed was for them to be separated out into different layers.
In constructing the image, at least the parts which showed data, I placed different data sources on different layers. In addition though, and hardly shown in the animation other than stage 5, the different districts were also on different layers, purely for technical reasons to do with formatting. But then I wondered what the implications might be if different strands within the data could be imagined as being on different layers. Analysis then becomes about revealing different layers, rather than filtering data using codes and themes. When complete, there’s a list of checkboxes we can use to hide or show those different layers in different combinations and perhaps reveal different insights.
I began to reflect on how this might sit with a sociomaterial sensibility, only too well aware that Annemarie Mol (2012) described ‘layers’ in a rather different and perspectival sense. Researching human tastes could be done by thinking of the body as existing on different layers and accessible through different disciplines: biology to explain the physical layer and sociology for the overlying social layer. These piled-up layers each produces a different reality due to its different perspective, but of the same body. What Mol offers is different bodies, each contrasting from other versions due to the different ways in which they are done or enacted in different settings or circumstances.
Were the layers in my vis different simply because they offered a different perspective? Possibly, but I’d like to offer an alternative conception of layers, one in which they are separate only when done so. When they are combined, they become singular. Although checkboxes can be ticked and unticked, when each new layer is added, the image becomes something new. We no longer have separate layers, but each being read through and with the other(s). A bit like the ‘tartan’ I described in an earlier post. The above animation is made up pf a series of images. Each of those images does not contain different layers, but has instead is flat. Each enacts a different version of the data. For example tweets for the most part (although not exclusively), and perhaps inevitably given their size limitations, sit within a single district. Blog posts rove across districts in different combinations. When the street names layer is added, those inter-district threads invite a different reading of the tweet data layer; it encourages you back into the tweets and invites you to read them in a different way. Maybe there are some districts which are not interwoven with others and somehow mutually exclusive, whilst others have several connective threads?
There’s another take though. It’s important to remember that as I constructed the image, one layer preceded the other. The tweets went on first to generate the districts, then the streets and blog post quotes came after. The tweet districts constrained where the connective street names could go. If a single blog post for example, contained instances of ‘exchange’ and ‘support’, there’s no street which connects the two. It’s the base map which imposes those limits; it influences the knowledge which can emerge. One could argue this as a bias, or one could see it as another way to decentre the human interpreter.
To return to the beginning, the image which inspired this and the earlier post was not created for analysis, but as a partial representation of the findings for a competition. What if analysis began with a base map, onto which tweets, quotes were placed, then streets or districts were named according to what the flâneuse observed? And what would happen if the same data were introduced to different base maps? Would the same knowledge still be produced?
Mol, A. (2012). Layers or versions? Human bodies and the love of bitterness.
[Postscript: Before leaving this topic, yes, I had considered that a flat 2D image might not be the best medium to convey a sense of walking the streets as flâneur. I’d contemplated Google Earth fly throughs, Google Streetmap-style renderings, VR environments like those where you can explore art galleries, museums and historical sites. Tweets and texts could be incorporated using billboards, notices, posters, graffiti etc, whilst a soundtrack might have snippets from the interviews playing in the background. The time to produce something of this complexity might be time well spent, if indeed that had been part of the thesis from the start and was an integral part of the study … but it wasn’t. I’ve no idea what will happen after the thesis and this might be the side project to which I return, but my current priority has to be on completion and submission]