3/7/2016

Visualizing data in 3D: Handling complexity through visceral and tactile experiences of data

 

‘Can you do Addition?’ the White Queen asks Alice in Lewis Carroll’s Through the Looking Glass, before rattling off  ‘What’s one and one and one and one and one and one and one and one and one and one?’ Information overload leaves Alice at a loss – unable to answer. Had Alice been given the same amount or even more information in a way that was intuitive and played to her strengths rather than her weaknesses as a human being, things might have been different. Therein lies the power and the potential of data visualisation, because once you see all the data together, holistically, you can identify patterns a lot more easily.

Data visualisations can be incredibly effective at showing where we are in the world today in a much longer term context because data on things like populations, deaths and marriages have been recorded pretty consistently in many different countries in some cases for over a quarter of a millennia. For example, in Norway and Sweden records stretch back hundreds of years, giving a real sense of seismic shifts in how “what’s typical” has changed over several generations.

InformationOverload.jpg

Image: Eilidhmcauley.

So we can see how a number of problems and challenges, particularly those to do with decline in fertility and increasing longevity, have emerged largely from the many hugely positive changes that have happened over the last few generations. For centuries cartographers struggled with how to create meaningful maps of the world. How could they show on a two dimensional surface (page or parchment) a three dimensional surface or landscape? We have the Ordnance Survey to thanks for developing a clear method of showing the topography of the world, the valleys and the mountains as well as the paths and how they vary over the world’s surface.

The same principles apply to mapping social trends such as ageing, mortality, fertility and migration. Using something called a Lexis surface, it’s possible to add three attributes to three dimensions:

  • year (or another measure of time) to the x axis
  • age (or another measure of relative time) to the y axis
  • a third variable, which co-varies with year and age, to the z axis

In 2013, together with Laura Vanderbloemen and Danny Dorling, I published an article in the International Journal of Epidemiology showing a series of maps of all-cause mortality in England and Wales from 1841 onwards, as well as some other European countries, in which the ‘height’ of the Lexis surface was determined by death rates for each single year and age in single years, meaning tens of thousands of values could be represented on a single image.

Lexis_mortality_curve_1903_336.jpg

Image: Graph of Lexis mortality curve (Wikimedia public domain).

Like on an orienteering map, contour lines were used to indicate paths along the surface in which the ‘height’ was constant. During the first half of the twentieth century, and despite the deaths of the two world wars (appearing as shards jutting up through an otherwise flat surface, many of these contour lines, representing a given risk of death, crept upwards: a risk delayed, to be faced a few years later in life. From the 1950s onwards the hurdles have still continued to rise steadily and inexorably, suggesting longevity is still increasing: a cause for quiet celebration for all but pension fund managers and economists who tend to see the world through the lens of the working age dependency ratio.

Rather than simply mapping the Lexis surfaces, however, modern computer technology has allowed the surfaces to be seen and explored directly. After some experimentation with computer generated images, rendering the surfaces as interactive virtual structures, with the support of the University of Glasgow’s Chancellor’s Fund, I have used 3D printing to turn many of these surfaces into real, physical structures. In total more than 40 separate data cubes, each 8cm by 8cm by 8cm in size, showing log mortality rates, population sizes, and fertility rates from dozens of countries who submitted data to the Human Mortality Database and Human Fertility Database.

dataviz-collage.jpg

Image: University of Glasgow’s Photographic Unit.

This has enabled me to compare how many children the women in as many as 50 separate cohorts in around 40 different countries have and at what age. By adding a contour to my ‘map’ I can see the average age at which the women from each birth cohort achieves replacement fertility levels (2.05 children per woman), without which its population will eventually go into long term decline and where a country will have more retired and elderly and not enough younger people in work to support them economically and directly.

Looking at complex data in this way undoubtedly has the potential to provide meaningful evidence for those interested in taking a long term and rational view of a host of pressing social issues. It’s not to be confused, however with the rather more popular term infographic, which has become almost synonymous with data visualisation.

Pretty images that capture attention may have a role to play in communicating research findings but they have begun to set certain expectations around what people want or anticipate from a data visualisation. Some say a good data visualisation is something that you should be able to understand and interpret in a matter of seconds. If this were true for the written word, then prizes for great literature should go to billboard adverts and tweets and road signs rather than novels and novellas.

This misconception is causing real challenges for those of us who want people to slow down enough to start to explore the information contained in these images and not just skim over it. To be able to interpret and understand them fully takes time and effort. I am hopeful that the extra visual, visceral and tactile experience of working with data visualisations printed in 3D might encourage people to make that extra effort.

Imagine how it could have helped Alice!

This article was originally published by LSE Impact Blog, republished under a Creative Commons Attribution 3.0 Unported License. Read the original article here.

Comments