Introduction

On November 16, 1990, NAGPRA, the Native American Graves Protection and Repatriation Act, was passed by Congress. This Act provides a framework for museums and federal agencies to handle repatriation requests for human remains, funerary and sacred objects, and other items of cultural significance from lineal descendants or culturally affiliated Native groups (National Park Service, National NAGPRA). Importantly, this legislation does not affect the way that repatriation of indigenous American cultural artifacts is handled through national museums (ie. Smithsonian Institutions), but rather applies to all other museums, federal agencies, and cultural heritage institutions. After an institution receives a repatriation request from an individual or a tribal community, they must decide whether or not they feel that the repatriation of the requested items is in fact the right decision. Once an institution agrees to repatriate an item or group of items, they must file a Notice of Intent to Repatriate with the National Park Service, through NAGPRA. These records are available to the public, beginning in 1995 and running through November 2016, through the NAGPRA Notices of Intent to Repatriate Database. The goal of the following visualizations is to elucidate any trends or patterns in American cultural heritage repatriation practices over the past 21 years.

There are a few important notes regarding this data. For certain Notices that were filed, two or more organization were listed on the form. In this instance, I plotted the coordinates for the first institution listed, while maintaining both names in the “Institution” field of the spreadsheet. When a user clicks on a point on a map, they will see the names of all institutions that were responsible for a given filing. If no department is listed for a repatriation filing at a university, the address of the admissions building was used. For repatriation filings from national and state forests and parks, the address of either the ranger’s office or forest supervisor’s office was used. Finally, it is important to clarify that this dataset is only reflective of American Indian, Native Hawaiian, and Alaska Native tribal repatriation requests made to museums, cultural heritage institutions, and federal agencies excluding Smithsonian Institutions. Indigenous American repatriation requests submitted to the Smithsonian are handled through the National Museum of the American Indian Act Amendment of 1996, and are not filed with the National Park Service. Relatedly, repatriation requests from other countries are also funneled through different government agencies and are not represented in this data.

Process + Design

The NAGPRA Notice of Intent to Repatriate dataset was quite messy, and I was forced to do quite a bit of cleaning before the data was functional. This included text trimming in OpenRefine (to remove irrelevant information from the “institution name” column), manual searching for the addresses of 256 unique institutions, and identifying the correct latitude and longitude coordinates for each organization. Once all of this was done, I uploaded the clean data to both Carto and Tableau Public.

Initially, I was under the impression that my dataset included information on the American Indian tribes and communities who had requested the repatriation of items. I had planned on creating a directed network showing which organizations were repatriating items to which communities. When I determined that this information was in fact not included in the dataset, I had to rethink my methods. Instead, I decided to create a map depicting which organizations had filed Notices of Intent to Repatriate. The temporal nature of my data made this a bit of a challenge, since any map would need to have an interactive component in order to effectively display 21 years worth of information. I created multiple visualizations in both Carto and Tableau, which were updated to reflect the feedback I received from my user testing.

My design choices were largely influenced by our readings, notably the work of Stephen Few and Edward Tufte. I chose to use graduated symbols for all of my maps, which seemed to be the most effective way to visualize the quantity of filings per institution. I also attempted to utilize colors that would not cause any issues for people with color perception challenges, though since most of my visualizations only contain one color, this was not that great of a concern. Finally, I was also very deliberate in my legend and filter naming, and attempted to be as concise and as clear as possible. I think this dataset has the potential to come across as slightly intimidating at first pass, and the results of my user testing made it quite clear that my titling was not as explanatory as it should be.  I reworked all of my labeling to more effectively explain the type of information that was being visualized.

Methods + Results

I began in Carto, and attempted to create a map with layer filters for each year of the 21 years of data. I quickly realized that Carto would not allow me to add more than 7 layer filters at a time. I then created four separate maps and evenly segmented the 21 years of information into five or six year increments. I was fairly disappointed with the results, especially the fact that users needed to move back and forth between four different maps in order to view one dataset. My user testing revealed that others shared this opinion, and found moving back and forth between multiple maps to be extremely confusing. User testing will be discussed in greater detail below, but this bit of feedback confirmed my own suspicions. I ultimately ceased work on this series of maps and invested more time in the other versions.

My next approach was to try using the “Time Series” widget in Carto. This enabled me to animate my map based upon the “date” column from my spreadsheet. While the animation worked well, it wasn’t very effective at conveying meaningful information. Users were unable to identify any patterns in the data, such as which regions saw a greater number of repatriation requests than others, or whether certain years saw more repatriation filings than others. The “burst” of activity, which represented each instance of an Intent to Repatriate filing, ended up blurring together in a frenzy of animated explosions and actually obscured any temporal or geographic trends in the data. Ultimately, this type of animation was fairly ineffective, and I again chose to focus more time on my visualizations in Tableau.

Finally, my third approach to visualizing this data was through Tableau Public. I initially created a simple line graph that clearly displayed the aggregate of repatriation filings for each year from 1995 through 2016. This ended up being an extremely effective way to display the total number of Notices of Intent to Repatriate per year. I added a trendline to help make the overall trend of the data extremely clear, and I think this was quite successful. Since there is a significant amount of fluctuation across the 21 years of filings, the trendline helps to illuminate the general direction of the data. I also created a page with the names of all institutions and the number of filings they submitted. I used this to create a filter for the line graph; when a user clicks on a specific year, they are able to see which institutions filed Notices of Intent to Repatriate that year, and the number of notices they filed.

While this line graph was effective at visualizing the temporal aspect of the data, it really failed at conveying geographic information. I created a third map, this time in Tableau Public, with the hope of creating a more successful filter than I had in the Carto maps. I sized the institution points on the map according to total number of filings over the 21 years, and added a filter for year. When users hover over a point, they see the name of the institution as well as the number of repatriation filings they have filed. If users have filtered by year, the institution’s plot point will automatically resize to reflect the sum of the years selected, and the “hover” information will be updated to match.

I believe, out of all of the maps I created, that this one is definitely the most effective. The year filters still allow users to interact with the data and to filter based upon their preferences. The fact that the resulting image is static, as opposed to the “Time Series” map in Carto, allows users to spend as much time as they would like digesting the information. It is far easier to internalize the information presented on a given set of years in a static image than in the animated timeline map, where the information is being presented so rapidly that it is nearly impossible to glean any meaning from it.

UX Testing

My user group was diverse in their interests and familiarity with the concept of cultural heritage repatriation. Of the six people that I recruited, only three gave feedback in a timely fashion. These users were fairly diverse in their professions (a children’s librarian, an engineer, and a social worker-turned full-time parent), gender identities, and geographic locations, though they were all of similar ages (late twenties to early thirties). I would have loved to have user tested with a larger group, but, as always, time was a constraint.

Users were sent a link to each of the three maps (the series of four maps discussed above was treated as one visualization) and asked to spend roughly 5 minutes with each visualization. Users also sent a link to a Google Survey which contained a variety of questions regarding each map. I used a combination of multiple choice questions, which enabled quantitative ranking, and short answer questions, which allowed the users to voice their personal opinions regarding style choices and effectiveness of each visualization. Perhaps most illuminating was the section where users were asked to briefly describe what they believed each visualization was displaying. The feedback I received was quite mixed. I initially had challenges with creating legends in both of my Carto maps, and clearly saw the confusion that the lack of labeling caused the users. People were at a loss as to what either of the maps were conveying without accurate labeling, despite the introductory paragraph I provided which contextualized the visualizations.

Perhaps most interesting was the dissonance between people’s responses to the animated timeline map, the Tableau map, and their ultimate ranking of the three maps.

 

As you can see from the above results, two thirds of respondents stated that the Tableau map was “pretty good”, while two thirds rated the animated timeline map as “excellent”. Yet in the section where users were asked to rank the three visualization by order of preference, two thirds rated the Tableau map as their first choice. Rather than making their preferences clearer, I think this portion of the survey was most useful for helping me to recognize the challenges of establishing a successful user test.

Revisions + Future Directions

Based upon the feedback I received from my user group, it seemed clear that better labeling and more effective filters would help to make these visualizations more successful. As mentioned above, I enhanced the legends on each of the visualizations, and this alone seemed to make the maps much more approachable and easier to understand. I also updated the styling of the Tableau map, removing the stepped coloring that was initially intended to represent each year’s worth of data, and replacing it with one uniform color for each point on the map. Rather than having color act as an identifier of temporal information, I added a “year” filter to the map, which enabled users to filter by time as they saw fit.

In the future, I would love to determine which American Indian, Native Hawaiian, and Alaska Native communities requested the repatriation of items reflected here in the NAGPRA database. I believe this information would lead to far stronger visualizations and, as mentioned earlier, would lend itself well to a directed network in addition to mapping. I would also love to have expanded the interactivity of these maps, including creating filters for region and type of institution (university, state or national park, museum, etc.). Ultimately, I was fairly pleased with the way the combination of the Tableau line graph and map turned out. I think the combination of these two visualizations really helps to elucidate trends in both the temporal and geographic aspects of this data.