Critical Approach to Data Visualization

Overview

This workshop led by UC Davis’ Data Feminism research and learning cluster unpacks the subjective process of data visualization and its relationship to concepts of diversity, equity and inclusion. We’ll critically explore how data can be used to uphold and perpetuate, or quantify and demonstrate structural oppression. Through this workshop learners will practice the technique of “data visceralization,” the process of experiencing differences in data and understanding them viscerally.

Introduction

Learning Objectives

After this workshop, participants should be able to:

  • Critically read a dataset and its documentation
  • Use the principles of data feminism to visualize data responsibly
  • Consider visceralizing data in addition to visualizing them

Before We begin

This workshop provides strategies for thinking critically about your data and help you visualize your data in more responsible and responsive ways. It will also help you consider how to present your data in meaningful ways that demonstrate impact and evoke emotion. This workshop will not provide technical instruction in data visualization. You can apply the frameworks described here to create data visualizations via any technology or medium.

This workshop will be interactive - you will work through case studies with your colleagues, brainstorm together, and share out what you have uncovered.

This workshop arose from efforts by the UC Davis DataLab’s Data Feminism research and learning cluster, established by Drs. Merchant, Poirier and Reynolds in fall of 2019 with support from the Feminist Research Institute.

You are welcome to use this workshop “reader” and the materials herein for your own instruction, provided that appropriate attribution is given.

Critical Approaches to Data Visualization

Step 1: Read Your Data Carefully

To build a responsible data visualization of a variable in a dataset, you should first critically explore the data. The following types of readings of your data provide a helpful framework:

  • Denotative reading - What is the technical definition underlying this statistic? You can often find this information in the data dictionaries and metadata (when available).
  • Connotative reading - What is the cultural history of this definition? What controversies are there? How has this definition changed over time? Data collection documentation and grey literature can help you dig into this.
  • Deconstructive reading - What is included by this definition? Perform critical analyses of the definitions, categories and structures to begin to understand who and what is being left out.

To illustrate this process, let’s look at a published visualization of population by sex from the U.S. Census:

A careful read of the data, its documentation, and surrounding literature behind in this visualization tells us:

  • Dennotative: Sex classification is either male or female.
  • Connotative: Prior to 1950, no definition was provided except M for male and F for female; the classification was done by enumerator inference and was assumed to be obvious. Starting in 1950, enumerators were provided with guidelines to use name or relationship to head of household to guide classification first, and to only ask if that information did not make it obvious. This indicates that “sex” may actually indicate gender prior to 1960.In 1970 sex was self-reported as either male or female. The 2020 census defined sex as “biological sex.” This is problematic for resolving with historical census data. The National LGBTQ Task Force’s Guide to the 2020 Census advised: “self-identify here in the way that feels most comfortable to you… Your answer to this question does not need to match what you have on official documents… answer this question in whichever way feels best to you.”
  • Deconstructive: Transgender, non-binary, and gender nonconforming individuals are marginalized, excluded, and not accurately characterized in the dataset.

Step 2: Visualizing Data Responsibly

Using what you learned above, you can now start to think about creating a more responsible data visualization. To do so, think about how you could:

  • Bring back the bodies - be clear on who is included, and who is excluded.
  • Show your work and give credit where due - cite your sources, methods, etc.
  • Acknowledge your standpoint - it is impossible to create perfectly objective visualizations.
  • Use appropriate text - titles, labels, captions, etc. can help readers better understand the complexities, and caveats, of the data.

Using those principles, we can improve the visualization:

Step 3: Visceralize Your Data

“One death is a tragedy, a million deaths a statistic” (attributed to Stalin, 1947). Now that you can create more responsible data visualizations, how can you demonstrate the importance of a message it conveys? One way to do that is visceralize your data to evoke emotion.

These are two examples of data visceralizations. How do these compare to a statistic? How do they make you feel?

Case Studies

This workshop will involve some group work in breakout rooms. In our first breakout session, you will be able to choose the group you want to join. If possible, please join one that has no more than seven other members (eight people total). Make a note of the group number so that you can return to the same group for subsequent breakout sessions. Please work with the dataset assigned to your group:

Groups 11 and 12: Medical error deaths

Jamboard for Breakout Sessions

Exercises

Step one: Read data critically

In this section, you will learn the difference between denotative, connotatative, and deconstructive readings of a dataset; how to do each; and why all three are necessary in order to analyze and visualize your data accurately and responsibly.

In your breakout room: Using the links provided above, explore the dataset assigned to your group. Choose one variable to read denotatively, connotatively, and deconstructively. Select one person to report on these readings to the workshop as a whole.

Step two: Visualize your data responsibly

In this section, you will learn how to apply your denotative, connotative, and deconstructive readings to standard data visualizations. We will go over some of the principles of feminist data visualization — such as acknowledging your standpoint, showing your work, and giving credit where due — and discuss how to use them in practice.

In your breakout room: How would you visualize the variable you analyzed in the previous step? How would you incorporate the principles of feminist data visualization? What kind of argument would your visualization make? Choose one person to report back to the workshop.

Step three: Visceralize your data

In this section, we will discuss strategies for going beyond standard data visualizations to consider ways of presenting data that might evoke the appropriate emotions in your audience.

In your breakout room: If you were not constrained by standard data visualization software (or your own artistic abilities), how would you visceralize the variable you have been working with? Choose one person to report back to the workshop.

Resources

Check out our curated Data Feminism reading list.

D’lgnazio, C., and Klein, L. 2020. Data Feminism. Cambridge, Massachusetts: The MIT Press.

Cifor, M., Garcia, P., Cowan, T.L., Rault, J., Sutherland, T., Chan, A., Rhode, J., Hoffmann, A.L., Salehi, N., Nakamura, L. 2019. Feminist Data Manifest-No. Retrieved from: https://www.manifestno.com/.

UC Davis affiliates are welcome to join DataLab’s Data Feminism research and learning cluster.