Introduction

The idea for this project stemmed from my Digital Media internship at the Whitney Museum of American Art. I wanted to create an interactive visualization that would allow users to explore data related to the Whitney’s collection. The kind of data I wanted, however, isn’t publicly available (if it has been compiled at all). So, I decided to compile my own data, using Whitney Museum of American Art: Handbook of the Collection as a starting point. I decided to only include women artists for several reasons. Importantly, representation of women in Art History has been a topic of interest for decades. Additionally, narrowing the focus made the dataset more manageable, both for myself as I manually compiled it, and for the eventual users deciphering information in a busy dashboard.

There were a few key themes that surfaced in “the handbook”, that I wanted to explore further:

1. As their name implies, the Whitney was founded with the intention to focus on “American” art. In her essay for the handbook, former Curator of the Permanent Collection Dana Miller asks “By what criteria is an artist considered American?” She goes on to explain that many of the artists included in the handbook were born outside of the U.S., but were still accessioned. I knew a map displaying this trend would be a significant way to represent the Whitney’s identity.
2. The museum’s Director, Adam D. Weinberg, also wrote an introductory essay for the handbook, in which he acknowledges that by the 1950s it had become evident that the Whitney’s acquisitions were largely realist pieces. An acquisitions committee was formed to “fill many serious omissions in the permanent collection and do fuller justice to the vastly expanding field of contemporary American art.” I was interested in exploring how these corrective measures affected the shape of the Whitney’s current collection holdings.
3. Weinberg’s essay also details how collection strategy has varied over the years in regard to the notoriety of a given artist, as well as the importance placed on acquiring “masterworks.” The Whitney as an institution has evolved based on the original motivations of its founder, Gertrude Vanderbilt Whitney, who focused on supporting living, emerging artists. But at some points in the institution’s history, various directors and donors have emphasized acquiring more prestigious pieces. I wondered if the effects of these strategies would be visible by exploring the relationship between how many works of art are held in the collection for a given artist versus how many times the artist has been deemed worthy of exhibition. Theoretically, works by already well-known artists would be more costly, and therefore more difficult to obtain. But that work might appear in multiple exhibitions as a “crowd favorite.”

I used the handbook to compile names, birth and death dates, and birth place. The Whitney’s website lists the total number of works by each artist, as well as exhibitions that have featured the artist dating back to 2006. To compile the methods used by each artist and their artistic style(s) I used the askART database.

Methods and Design

I used three separate CSV files to store my data. The “Master” file contains one record for each artist, which includes:

• Artist Name: I used Last, First format for easily alphabetization, but also First Last format, as an easy-to-read option for the Tool Tips in some visualizations.
• Birth and Death Dates: I ultimately decided to format these as a span of time (YYYY-YYYY) – again for ease of reading – and treated them as string data.
• Number of Works
• Number of Exhibitions
• Birthplace: I formatted this data as Latitude and Longitude, after uploading each artists’ birthplace to a geocoder. This step was necessary because Tableau did not recognize most cities outside the U.S. when they were formatted as addresses.

The “Method” and “Style” data was originally stored in the master file. The data was separated by commas while I was still in the process of compiling it. In order to use it in Tableau, however, I moved the two categories to their own spreadsheets, and cleaned them using Open Refine.

Once I had the three spreadsheets in Tableau, I connected them using two full outer joins. This step was necessary for the filters to work together in my final dashboard, but it did create one problem, which I describe in more detail below.

I created the following worksheets to be included in my dashboard:

• A table of each artist’s name, alphabetized by last name, and their lifespan. This worksheet includes a Wildcard Match filter to allow users to search the dashboard for information by artist.
• A map of each artist’s birthplace. The points are set to 30% opacity, and as they “stack up” on top of each other they become darker. This allows users to identify areas where several artists were born. Originally I wanted to provide an inset map for the New York area, but it was not feasible due to space constraints. Tooltips provide only the artist’s name.
• A scatter plot showing the relationship between the number of works for each artist and the number of exhibitions they have been included in since 2006. The full outer joins caused artists with multiple methods and/or styles to have multiple records that reflected their works/exhibitions. I had to adjust the default aggregation to AVG (and edit the number of decimal places displayed in the Tooltips) to accurately reflect the data. The points in the resulting scatter plot are categorized by intensity of color for number of exhibitions, and size for number of works.
• Packed bubble charts show both Method and Style, with the number of artists that fall into each category reflected by the size of the bubble.

Users can search by artist name, or click on an artist name, to filter results. This provides an intuitive starting point. The Methods and Style charts are also designated as filters for further exploration.

To fine-tune my visualization I did three rounds of user experience testing:

• Round One: Design Flaws
  • User A: Young visual media professional
    • Observation
    • Familiarity with Technology: High
    • Feedback on design principles and responsive layout
• Round Two: Intended User Group (Museum Visitors, Professionals in the Arts and/or Education)
  • User B: Educator and frequent museum visitor
    • Phone Interview
    • Familiarity with Technology: Moderately High
    • Feedback on language, visual appeal, and usability
  • User C: Young museum professional with an Art History background
    • Email Interview
    • Familiarity with Technology: Moderately High
    • Feedback on visual appeal and usability
  • User D: Museum professional, formerly a New York resident, familiar with the Whitney’s collection
    • Email Interview
    • Familiarity with Technology: Moderate
    • Feedback on visual appeal and usability
• Round Three: Final Details
  • User E: SAS programmer familiar with Tableau
    • Email Interview
    • Familiarity with Technology: High
    • Feedback on overall design

Experienced with a variety of visual media, User A provided feedback on the overall design of my visualization before I sent it out to my Intended User Group (represented by Users B, C, and D). I addressed the feedback from Round Two before sending the visualization to my final user. As someone familiar with Tableau, User E had the experience to provide feedback on the “finalized” design.

The final visualization is hosted on Tableau Public.

Findings

The dashboard facilitates exploration of key themes from the handbook.

The map makes up a small part of the dashboard as a whole, which does make it more difficult to view densely populated areas. Users have to zoom in several times to parse out the stacked dots in the New York area. The intended message of the map, however, is that “American” has been broadly defined. This fact is made clear through the default view of the map, when users can see artists’ points distributed across the world.

The Whitney took steps to balance the holdings of its collection beginning in the 1950s, at a time when non-representational art was gaining popularity. The success of these efforts is evident through the Style chart. Most obviously, the large “Abstract” bubble dwarfs the “Realism” bubble (although it’s worth noting that the “Realism” bubble is still large enough to accommodate a label, unlike other less well-represented styles).

When viewed without filtering by artist, style, or method, the Works vs. Exhibitions scatter plot is best at conveying information about outliers. Some important information is easily accessible: Georgia O’Keeffe’s renown makes her work more difficult to acquire, but also makes her a desirable artist to include in exhibitions.

Users enjoyed the interactive approach – but still wanted more.

All three of the Intended Users said they enjoyed that the visualization was interactive.

User C wrote:

“I’m a huge fan of infographics, so I really like the multiple charts and graphs with the gradations and proportioned spheres. It all connects really well. I like that I can pick an artist and the graph and charts adjust and show me everything to do with that particular person.”

User B wanted more ways to explore. She said she wished the visualization could link to images of each artists’ work.

User E (not technically part of the Intended User Group) wanted short bios through Tooltips in the Artist Name list. After exploring, she realized that a lot of the information she wanted was available through the other charts, and ultimately decided that the additional text she originally wanted might end up being excessive. Despite this conclusion, I think it’s important to note that there is still so much more this visualization could do, in the right setting.

Users appreciate color, and need it for visual appeal even if it does not distinguish specific categories in a dimension.

Initially, all of these visualizations were done in greyscale, or black and white. I did this for two reasons. First, the Whitney’s web style guide does not typically include color. The logo, navigation etc. are always done in black, white, or shades of grey, with color featured in installation shots or high res images of art. Second, each of my dimensions had too many categories to be distinguished by color. Since I couldn’t use color to “mean” anything, I felt it would be best to eliminate it entirely to avoid confusion.

Despite my best intentions, all Round Two Users mentioned using color to make it more visually appealing.

After Round Two, I spent a lot of time playing around with color. I knew it was important for my users to interact with a visually pleasing dashboard, but I was still wary of causing confusion. I finally decided on two separate earth-toned color palettes for the Methods and Style charts, setting them both to 75% opacity for a softer effect. The colors are arbitrarily assigned to the categories, and are repeated throughout the chart. I colored select text across the rest of the dashboard to provide a little balance.

Next, I had User E look over the visualization, without soliciting any feedback on the colors specifically. In her email response she said: “The colors are good.” This comment, coming from a Tableau user, made me feel more comfortable with the decision to incorporate them.

A responsive layout is not always ideal for information visualizations.

Initially, I tried an automatic layout, in the hopes that my final visualization would be responsive to different screen sizes. User A provided ongoing feedback as I experimented with the automatic layout, comparing his 24” desktop screen to my 15” laptop screen (and eventually comparing mobile screens as well).

After a lot of trial and error, it became clear that the automatic layout frequently led to distorted charts (especially Works vs. Exhibitions), and excess white space around Methods and Style. I moved on to exploring fixed layouts, testing various landscape formats before ultimately creating a custom layout at 1150 x 600 pixels which seems to work well on varying laptop and desktop screens. It is not a mobile-friendly visualization.

Any text that appears in a visualization must be absolutely clear.

Before I sent the visualization to my intended user group, I was aware that the “Keep Only” and “Exclude” commands popped up, but I guessed that users would simply ignore them.

My assumption was incorrect. Users B, C, and D all voiced their confusion over the commands. User C wrote: “I was a bit confused with what to do with the “Keep Only” and “Exclude” but I figured it out after playing with it some.”
Interacting with these commands was not an intention for my design, so even though User C felt like she eventually “figured it out,” I knew I needed to do the work to eliminate the confusion. The procedure to remove the commands was not intuitive for me, but after consulting Tableau’s online help discussions I was successful.

In addition to this trouble, my instructions in the top left of the dashboard caused confusion. Originally, “click the bubbles to the far right” did not include the word “far,” so User E clicked on the map thinking it would filter.

Important interactive controls can’t be buried.

User A helped me pick out small design flaws. One of my concerns before I even began testing was that Tableau was defaulting to the rectangle selection tool, rather than the pan tool for the map. User A quickly realized where the pan tool was buried, but I knew I would need to get feedback from less tech-savvy users as well.

In my discussion with User B I realized she thought that she wasn’t “allowed” to move the map at all, because she didn’t even know to look for the pan tool in the first place, much less where to look for it.

After a lot of trial and error, I realized that to make the pan tool the default for the map, all I had to do was make sure it was currently selected when I uploaded the dashboard to Tableau Public.

Conclusion

Academics, artists, and museum professionals enjoy learning more about museum collections – especially using interactive and responsive tools. Established institutions, with the resources to do so, should explore ways to make these experiences possible.

However, it shouldn’t be assumed that just because these users enjoy interactions in a digital setting they are familiar with advanced software/applications. Instructions need to be succinct but clear. Superfluous text should be omitted, or only offered as an easily dismissed option that doesn’t clutter the experience.

There is a lot of evidence that people in general are increasingly using their mobile devices to interact with digital content. Emphasis has been placed on responsive design, but highly informative and interactive content may not be best experienced on a small screen. In these instances, institutions should make content accessible via appropriate channels, and make it clear to users how they can best enjoy the experience.

Information professionals need to keep in mind that users who want to interact with art related content are not necessarily going to appreciate some of the established visualization best practices. A visually appealing dashboard will take precedence over one that technically communicates the information more directly.

There is a lot of potential for a visualization like this to expand in the right context. I had to consult multiple sources to manually compile the data for this project. In contrast, some institutions at the very least publish full CMS records online, while others provide tidy datasets through GitHub. By compiling the right data, and allowing it to be available to the general public, the Whitney could create an interactive portion of their website that includes charts similar to the ones created for this project (albeit much more extensive) that include all artists in the collection and all exhibitions since the inception of the museum. To make information easier to digest, or to highlight important messages, the interactions could be divided into sections on a webpage, i.e. “Political Art”, “Abstract Expressionism at the Whitney”, “Emerging Artists” etc.

To provide more information, the Whitney could incorporate or link to artist records. For example, the Works vs. Exhibitions chart would be more interesting if users could do more to explore why artists fall where they do on the chart (Why is Kiki Smith so well represented overall? Do some of the artists fall in the middle ground because the Whitney patronized them before they became well-known?) As one user suggested, the Artist table could become more interesting with Tooltips that include a brief biography and/or images, and the option to visit an official record page for more information.

It’s becoming common practice for world-renowned cultural institutions to provide large amounts of collection data online, many times in an interactive context. When compared to museums like Rijksmuseum, National Gallery Victoria, Tate Gallery, Brooklyn Museum etc. the Whitney’s online collection data can seem a bit sparse. With the right attention and tools, however, their visitors could enjoy exploring the collection in more detail.