by Garfield Benjamin
Data visualisation is is becoming increasingly popular, and has produced a massive range of beautiful images that allow us to understand issues that may get lost in the abstraction of pure data. They can even help make inequality visible where it is often hidden. Maps are a great example this, and digital tools have enabled the creation of many stunning visualisations of data across physical space. But visualisation in general, and mapping in particular, is never neutral.
Last year, Google maps changed its projection method from Mercator (one way of creating a flat map from the spherical earth) to a 3D globe model. This corrected the distortion of, for example, Greenland – which is now no longer shown as being larger than Africa. The translation of a map from one perspective to another will always introduce bias, and various cartographers have come up with increasingly elaborate ways around it (some aiming for clarity, others for aesthetic). Below we can compare the stretched Mercator projection (Map 1, first created 1569) with the more compromised Miller projection (Map 2, created in 1942 to iron out some of the Mercator’s issues). More extreme examples include Azimuthal Equidistant projection (Map 3, first created around 1000, now the basis for the UN symbol) and (one of my personal favourites) the Waterman Butterfly (Map 4, first created 1996). These maps show the issues with projecting geography itself, let alone the myriad issues that are introduced when we collect and display data on these maps.
To illustrate some of the issues around visualisation and interpretation, let’s take a look at population and urbanisation data from the ‘Cities’ map of Harvard’s WorldMap project. This map gives us a fairly simple visualisation, showing settlements with 1000 or more inhabitants (the red dots). For some maps below I have also opted to show urban areas (the black-lined burgundy areas) and administrative divisions (the purple lines). First, though, let’s look only at settlements across Europe, without any background map images or other data (Map 5):
Visualising data in this way can show us a great deal. Immediately we can get a sense of the distribution of people across Europe. We can see the greater population of Germany, the ‘hexagon’ of France, the cluster of settlements in Northern Italy, and even the band between Glasgow and Edinburgh. It can also imply data by what is missing – even without borders explicitly drawn in, we can see coastlines clearly defined by the higher proportion of settlements. We can infer the location of the Alps by the band of sparser settlement, and the mountains and forests of Romania are also visible. Others become blurred, such as the Highlands of Scotland which seem to disappear into the sea.
Only by showing the urban areas in Map 7 do we get a more realistic picture of the spread of human activity outwards from the centre, as well as the locus of other major areas.
Another aspect to consider is the format in which data is entered, and the distortions this might create once it is visualised. For example, below (Map 8) is southern Norway. Here we see mostly one large settlement per administrative area, with a few extras along the coastline (as we might expect). So far, so predictable.
But the following map (Map 9) shows two very different styles of visualisation. We haven’t changed the conditions of visualisation, but the differing approaches to entering data (whether it is defining what counts as a single settlement, or how granular administrative divisions are) marks a clear divide at the boundary between Denmark and Germany. It becomes difficult to tell the relative levels of human settlement in this region. Are there many more people in Germany? Are Danish cities fewer but larger? Or do they simply report their data differently?
We require additional information relating to population density (Map 10) to fill in the gaps. But adding additional information also introduces further distortions as the pixelated areas fail to match with administrative or geographical boundaries. Borders can be political but they also display the politics of data!
If we can use a critical reflection on visualisation to view the politics of data, we can also use it to expose the politics of border divisions. Take the two maps below (Map 11; Map 12). In the first, we see various settlements across a few regions. In the second, however, we see the urban area around Saarbrucken spread across the Franco-German border – a contested space throughout history and particularly during and after WWI that still bears this history through mixed-language place names. Contextualised within the region and urban areas, even the settlement distribution of the Saarland has more in common visually with France (and Luxembourg) to the south/west than it does with its neighbouring German region to the north/east.
Maps show political divisions, but they can also visualise the politics of data structures. Advancing our technological capacities to visualise and use data to tackle human issues is important for historical research as well as present and future society. But if Digital Humanities is to retain the critical cultural perspectives that make the Humanities so important, it also needs to take into account the political implications of how data is generated, stored and used. In the Return to Sender project, we are focusing on mapping the movements of postcards to assess the politics of movement in and around WWI. But linking this to their current location in archives, and the digital mediation of the objects as data in online archives, opens the debate up to issues of data politics and how we frame and access collective memory of a time of shared trauma across Europe and for the rest of the world.