Introduction
“Cultural heritage” is a broad term that is worth defining before proceeding with this report. According to UNESCO, the United Nations agency tasked with documenting and preserving cultural heritage, the term can be broken down into tangible and intangible cultural heritage. Examples of intangible cultural heritage include performances, rituals, and languages. This project focused specifically on tangible cultural heritage – both movable, such as paintings and books, and immovable, such as buildings and archeological sites.
Having completed a previous lab on military conflicts in the 20th century with a dataset I built, I was keen to continue working with a similar topic. In parallel, as a result of current events, the destruction of cultural heritage, particularly in wartime, has become a renewed and much-publicized topic of interest. I was intrigued by the contradiction of seeing regular reporting of cultural heritage destruction, but little to no specificity: I found no comprehensive lists or databases on a global scale that reflected the losses in detail. Furthermore, the patterns behind the destruction were often even more obscure. Therefore, I decided to use my final project to try and make a dent in this topic. While I did initially underestimate the complexity, reliability, and subjectivity of the information available, I also didn’t want to back away from the project for those reasons alone, as it would perpetuate the very problems I had identified. If a lack of specificity and recording is the problem, then the first step is to try and understand and compile what is available, what we do know for sure, and what we can confirm to begin defining the way forward.
I chose to organize the visualization in a narrative way – with a clear beginning, middle, and end – to support and tell a story about the patterns and comparisons that were being displayed (Segel and Heer, 2010):
- Provide an introduction/overview of the topic, and show some of the global patterns that emerged from cultural heritage destruction through war and violent conflicts.
- Having established the topic, I wanted to provide viewers with details and thought that a comparison of two countries (one from each century) would illustrate differences and similarities in the ways that cultural heritage is affected during wartime. I chose to compare Kosovo and Iraq for the following reasons:
- They are both very local conflicts with significant amounts of foreign and international involvement.
- The destruction in both countries had significant religious, cultural and ethnic associations.
- Both are located in contentious and historically disputed regions of the world, where borders were redrawn throughout the 20th and 21st century.
- Their major conflicts both occurred in an era of comparable technological development and therefore the capacity for destruction is also comparable.
- Lastly, I wanted to give viewers an understanding of where we can go from here and what future directions for this project may look like. By highlighting unreliable data and gaps in our knowledge, the final section of the visualization can inform actions to begin remediating this problem.
The project goals were to better understand the following:
- What patterns, if any, can be found if we analyze data related to cultural heritage destruction?
- If patterns are identified, can these help us to better prepare, preserve, archive, record, report and/or catalog cultural heritage before, during and after military conflict?
- What data is available, how can it help us and how accessible is it?
- What data is unavailable and why? Are there deliberate attempts to keep this data hidden?
Methodology
As I chose to build my own dataset, the process of putting it together was crucial as it would completely define the outcome of my visualizations. What kind of data I chose, how I recorded it and the dimensions and attributes I chose to keep would dictate the visualizations I would be able to build.
Data Collection
For the networks lab, I used data from Wikipedia to build my own dataset and that was a fairly painless process. Helpfully, Wikipedia already has a lot of their data organized into tables and lists, has links to source materials and further information that helped with fact-checking and also had the benefit of a global network of editors; I thought this would provide a fairly international perspective of cultural heritage destruction from various regions and wars. I began the data collection process by identifying the main Wikipedia pages that would contain the information I needed (these can be found in the references section). I relied most heavily on the List of Destroyed Heritage and used this as my starting point for each country. Informed by my user research, I also began compiling a list of all the attributes that an audience might be interested in seeing in this visualization. This was helpful to keep in mind while going through Wikipedia as it helped me focus on specific things to look for and collect, that I knew would be interesting for viewers. After identifying all of the main Wikipedia pages that would serve as the basis for the project I began organizing the data along the following headings:
This was the most challenging part of the project because the List of Destroyed Heritage was by no means complete. Where specific dates, locations and/or affiliations had not been assigned to a particular destruction, I followed up on the object’s individual Wikipedia page (where available) or tried to find information from other sources that would fill in the gaps. Where the data lacked specificity or was disputed (usually dates), this was noted. While I did not think that I would be using all of the data I collected to create my visualizations, I knew that I only had time to go through the data collection process once and so preferred to over collect and then discard later, which was the right decision. However, this meant I spent more time on data collection than design than I would have liked to. Furthermore, collecting several different dimensions was important to me as I knew that having multivariate data would be best suited for achieving my project goals. As Few states, “when trying to figure out why something is happening, we need to expand the number or type of variables we’re examining” (Few, 2009), which is exactly what I was trying to do.
Data Caveats
By compiling this data myself, I was aware that it would reflect my own biases. This was further complicated by the highly subjective nature of the topic. I have separated my data caveats into two categories: those that explicitly recognize the biases I have imbued in the data, and the ones that served to narrow the scope of what data was included and excluded.
Data Biases
- To understand why cultural destruction was such a prominent feature in military conflicts, I knew that quantifying and recording intent would play a part in my dataset. As I was the one deciding what type of destruction counted as intentional or not, based on the data available, this is a biased measurement. There can also be varying levels of intent, a nuance I was not able to capture.
- By trying to elicit patterns behind this wartime destruction, I thought that recording the types of wars in which these happened would be important (and their corresponding volumes). However, the type of war depends on who you speak to – a terrorist may see his cause as more similar to a civil war than a terrorist campaign – and so this is another are where my judgment is explicitly imbued in the data.
- The fact that I only speak two languages means that there is an added a layer of bias in terms of the information I can access and interpret. For example, there appears to be a section on Wikipedia related to Ukrainian cultural heritage destruction, but it is all in Ukrainian.
Data Scope
- Due to time constraints, I kept the time analysis period between the 20th and 21st century.
- I decided only to compile data for destruction events that occurred during and as a direct result of military/violent conflict. I chose not to pursue cultural heritage destruction under political circumstances (even where these were accompanied by state-sanctioned force) as I believe this is a whole topic in itself and could not be accommodated in the time I had. Examples might include the cultural destruction that supported the narrative and establishment of the Cultural Revolution in China, Franco’s dictatorship in Spain, the rise of the Nazi party in Germany in the 1930s and the anti-religious campaigns of the USSR in the 1950s.
- As mentioned in the introduction there is another category of intangible cultural heritage that this project has not included. I believe there is a serious argument to be made for the effect of military and violent conflicts on the decimation of intangible culture. Through genocide and colonialism alone, native languages, local and religious customs and traditions have disappeared. Unfortunately, the very nature of intangible culture makes its recording harder and its preservation even more so. Therefore, a lack of data and time meant it would not be possible to include this in the project.
Tableau Visualization
Once I compiled my data (an extract below), I imported it into Tableau:
Initially, I wanted my project to be more exploratory, however, I soon found that Tableau constraints would compromise the user experience if I went in this direction. Particularly in the comparison section, between Iraq and Kosovo, I wanted to provide a detailed city-by-city exploratory view, where users could see exactly what types of cultural heritage and which religions were affected. However, if I used a city as a filter in Kosovo, all of the Iraq-related visualizations would disappear, thus rendering a comparison impossible (figures 3 & 4). This was frustrating as I would have to give my participants an experience that was more static than they would have liked, but since the wider project goal was to allow for comparisons, a more narrative view was prioritized.
I also had issues creating maps in Tableau. In my data set, I recorded the city and country of every destruction event. It wouldn’t have hurt to have recorded latitude and longitude as well, as Tableau did not recognize hundreds of cities. Between Iraq and Kosovo, there were around 35 unknown locations, which I then manually input the latitude and longitude for directly into Tableau. A further issue I encountered with maps, was that I could not illustrate the gaps in my data as clearly as I would have liked. I wanted to show the large areas where there was no data and thought that reversing the map, to highlight those areas, would be clearer for viewers. Unfortunately, the image below is the closest I managed to get to a color ‘inversion’ where the grey areas represent those without data:
UX Research and Participants
The UX research I conducted for my project was crucial in informing its design. It was organized in the following manner (all participants were Pratt students):
- A first round, before I began data collection, with 3 participants (1 DAV, 2 MDC);
- A second round, with most of my visualizations in progress and some of my dashboard stories complete, with 2 participants (both IxD);
- A final round, with my visualizations and narratives complete in preparation for the class presentation, with 1 participant (IxD).
Round 1
I used this round to establish the direction of my design. I described the project and topic to participants and then asked them what kind of information they expected to see in a visualization covering this area. By interviewing one DAV student and two from the Museums and Digital Culture program, it soon became clear that these represented the more casual vs. expert user. Both of the students in the MDC program had very clear and specific ideas what they wanted to learn. They wanted to understand not just what type of cultural heritage had been destroyed but the significance of that heritage. They also wanted clear divisions between the heritage itself, such as architecture vs. artifacts, and within these divisions, clear categories and sub-categories. The DAV student had a more general interest and wanted more quantifiable information. They wanted to know quantities of lost heritage per war, and they suggested small multiples to show volume and numbers of different types of destroyed heritage across wars.
Some of the suggestions, such as capturing cultural significance, would be beyond the scope of this project, but the remaining suggestions were largely incorporated into my data headings and my data research process and helped direct the categories I ended up recording.
Round 2
I used the second round of testing to validate the charts and graphs in progress to ensure I was going in the right direction. I asked each participant to explore seven sheets and two dashboards, for a maximum of 10 minutes. I also asked them to ‘think-aloud’ while exploring so I could understand their thought process. I also interrupted them with questions when I wanted to understand more about their navigation choices. I followed up the exploration with two open-ended questions:
- Is there anything you would change in this visualization?
- Is there anything you would like to see more of in this visualization?
This round was significant in helping me shape my final design. The main change I made in my charts was from an observation one of my participants made about my comparative dashboard between Iraq and Kosovo. I initially had each country map side by side, with their relevant graphs underneath (seen in figure 3). As my goal for this dashboard was a comparison, my participant suggested that I actually try to have similar graphs and maps on top of each other vertically, to allow for a more direct connection (figure 6). This way the two bar charts visualizing intent are directly on top of each other, as are the maps, making it easier for our eyes to see the differences. My second participant also asked for more descriptive titles, which I incorporated. Especially in the graphs related to religion, they were not sure whether I was talking about heritage or people. They were also confused as to what the numbers related to – it was not clear that it was a count and I thought that percentages might be clearer instead.
Round 3
In my final round of testing, I wanted to check whether the complete design that I had settled on was violating any heuristics. I decided to use Nielsen’s 10 Usability Heuristics for Interface Design for my evaluation and asked my participant to explore only my three completed dashboards for 5-10 minutes. I then asked them whether they felt any of the heuristics were being violated. They felt that the ‘User Control and Freedom’ and ‘Error Prevention’ heuristics were being violated because whenever they tried to select a specific city or category in a visualization, Tableau made it hard to de-select. This made the visualization feel even less exploratory but ultimately was not a violation that I had control over, unfortunately.
Design
As discussed above, I made significant changes to the layout of my dashboard between Iraq and Kosovo to allow for more direct comparisons. When one of my participants suggested using a different measurement to count and frequency, for better understandability, I decided to use percentages across the project. This is because a large part of the goal for this project was establishing comparisons to find patterns and the easiest way to make sense of these, would be to “normalize the data sets in a way that can make comparisons easier.” (Few, 2009); e.g.: with percentages.
One of the main ways I used color was to support association. Each section of my narrative had a different color that could be seen across all graphs. By using different colors for different sections, I wanted to show users that they were clearly moving into a new part of the story. I also wanted color association to have particular meaning for my first and third dashboards. I used red for the beginning of the story to make an impact, establish the topic as one of destruction and violence. Whereas the for the end, I wanted to leave viewers with a greater sense of stability, of calm, of knowing that despite the data gaps identified, there is a path forward to work towards remedying this, so I felt blue lent itself well to this (MacDonald, 1999).
As I was capturing a lot of different data dimensions, I wanted to use bar graphs to emphasize the independent nature of the data, e.g.: destroyed buildings are a separate event to destroyed artworks. Bar graphs also worked well to emphasize hierarchy, when sorted from highest to lowest value, like the one in my third dashboard below (Few, 2009). As the goal for this particular bar chart was to show the difference between the high volume of building destruction and all other categories, I thought the white space of a bar chart also lent itself well to visually defining that gap:
I used treemaps to represent the religious data I had because treemaps allow us to “readily spot extremes and predominant patterns” (Few, 2009), which happened to be the case in the religious cultural heritage destruction in Kosovo and Iraq. It is a strong visual representation of the amount of Islamic destruction that took place and was well recorded in Iraq, versus the comparably very small amount of Islamic destruction that took place in Kosovo, that likely attests to poor record-keeping.
I used the visualizations below to inform my design in two different ways. The UNESCO network represents intangible cultural heritage and served as a good instruction for how detailed I probably wanted my categories to be (figure 10). The visualization from Afghanistan (figure 11 – Segel and Heer, 2010) was instructive from a color perspective and the effective use of red, in different hues, encouraged me to use it for my opening story.
Findings & Recommendations
The findings contained information that was both expected and surprising. For example, I had assumed that most cultural heritage destruction (particularly those that were not buildings or cities) would be intentional. However, I did not expect it to be 92.54%, which perhaps points to better record-keeping or reporting when the destruction is intentional, as the loss is likely significant to a particular group and therefore recording.
An unsurprising finding was that buildings are the biggest targets of cultural heritage destruction. Especially as technology advanced in the 20th century and bombing increasingly became the tool of choice (Dresden, Hiroshima and Nagasaki, and Guernica to name a few), buildings were often on the front lines and the first to go. The findings show us that we’re good at documenting buildings but not necessarily their contents. This is an area where digital technologies can make a big difference in recording and cataloging beyond the usual suspects of museums and libraries; there needs to be more systematic cataloging of other buildings with cultural significance and their contents.
Lastly, I found the map from figure 5 to be especially interesting, as it showed that there were entire regions that had no data at all; even though several of these have had military conflicts throughout the 20th and 21st centuries. This is helpful as it shows precise areas for improvement but also brings up interesting questions as to why this data may be deliberately buried. This is echoed in the comparison between Kosovo and Iraq, where the data disparity between recorded Serbian Orthodox and Islamic destruction may also point to intentional gaps.
Notable Data Gaps
There are two notable data gaps that I only recognized once the visualization was complete (particularly the map):
- No data on the destruction of cultural heritage through colonialism. The 20th century saw an unprecedented number of decolonization movements and wars and I assumed there would have been an element of cultural heritage destruction in these conflicts (intentional or otherwise);
- Very little data on destruction as a result of genocide or ethnic cleansing, such as the Armenian or Cambodian genocide.
Overall, I was able to make good progress but my findings are not as detailed as I would have liked. They are useful to show areas for improvement, but my recommendation would be to start with an even narrower scope, perhaps at a country or region-level first. A narrower approach could probably accommodate a higher level of detail, which would allow us to get to the heart of the project’s goals: illuminating any patterns and trends behind this topic.
Final Visualization
References
- Nielsen’s 10 heuristics for evaluation
- Few, Stephen. (2009). Now You See It: Simple Visualization Techniques for Quantitative Analysis.
- Macdonald, Lindsay. (1999). Using Colour Effectively in Computer Graphics.
- Segel, Edward & Heer, Jeffrey. (2010). Narrative Visualization: Telling Stories with Data.
Wikipedia and Secondary Resources Pages