Sections

1. Introduction
2. Methods & Processes
3. User Experience Study
4. Findings
5. Future Directions

Introduction

Originally, the project objective was to supplement a larger project on online violence against women in politics. It was going to focus squarely on the literature review and visualizing information about it. Throughout the course of research and design, for reasons detailed in subsequent sections, I decided to prioritize width, an overview of gendered research on online violence, over depth, drilling down into political aspects. 

As a result, the project as it stands today is a soft-launched pilot, or dry run, of a data visualization framework. This topline view of the state of the research on online violence is a series of connected dashboards on Tableau Public.

Motivations

The impetus behind this project is philosophical. The twin challenges of finding and/or collecting feminist data can at times be harrowing. I’m determined to alleviate some of that pressure by making dozens of datasets more accessible and modeling open access. A truly feminist act. This is what we mean when we say “do the work.”

It’s also a result of the success I’ve had utilizing data visualization, based in qualitative information, for advocacy and policy changes. The dashboard is the first of its kind (as far as I can see) to be designed for public use with these methods, on a gendered topic, in this sphere.

Defining features include:

1. Gendered datasets are requisite. At a bare minimum, they must be disaggregated by gender.
2. Including a variety of sources highlights existing research and iterations, plus gaps and opportunities.
3. The information is rendered more accessible by means of individual dashboards that are organized according to facets of the issue. The whole body of work is an intersectional look at the phenomenon. 
4. Visual representation of data and the ability to freely download the dataset are both valuable research and analysis tools in their own right. It is extremely rare that they are deliberately combined in order to be simultaneously accessible, which is what I’ve pursued here. (For example, economist data has the visuals but not the data, Pew has the data but not the visuals)
5. Independent research, in my case, affords greater creativity and input, not bound by organizational demands, timelines, or budgets. I’m able to maintain academic rigor while expressing results in relatively unorthodox ways. A novel approach often gleans novel findings.

Outputs

1. Visual representations of the data in dashboards
2. Twenty-one accessible datasets properly cited and collated, formatted for easy manipulation
3. Accompanying report and analysis

My bibliography and literature review used for this project are accessible as a Zotero collection and can be considered an output.

Methods & Processes

Materials & Software

Software used to create the visualizations was Tableau Public (Desktop) v.2021.1. I’m running a Windows 10 Environment, primarily using Google Chrome, subsequently G Suite. Zotero was used to manage research, coolors.co to manage colors. I used Lucidchart to create the diagrams in this report, which is hosted on a Pratt-managed WordPress site. The datasets are linked above and available to view within the dashboard.

Dataset Selection

The primary organizing principle for the dashboards is rooted in analysis. My primary audience is researchers pursuing their own analysis on online abuse/harassment. As my research shows, the amount of empirical information about this phenomenon has increased dramatically since 2014. Longitudinal studies are still rare, and I only found a couple by Pew Research Center, but some organizations like the The National Democratic Institute, Anti-Defamation League, Data & Society, among others, conduct surveys with questions that can be longitudinal in order to examine trends. I have yet to find a study that compares with Pew’s quantitative finesse, and don’t anticipate finding one soon. As such, in several instances I could have included more nuanced information from Pew that supports deep analysis, therefore relevant to the target audience, but I typically omitted those in favor of providing a breadth of sources and geographic variety.

Then, running into a familiar problem, I eliminated several potential datasets because despite novel findings, they were not even disaggregated by gender. Often, the most pertinent datasets are buried in paragraphs, nested in PDF’s. It takes time to parse, extract and organize that information into records with dimensions, variables and targets. (Note: I will be reaching out to several researchers for data in a more accessible format, we’ll see how far I get). 

Design Aesthetics

In the Overview v.1,5, I used redundant coding to redesign the original dashboard that this project is an expansion of. I also changed the graph and chart types to more accurately reflect the information.

This slideshow shows v.1.0 (in green and yellow) next to v.1.5 too illuminate the dramatic changes made. The UX study informed many of the decisions to change my designs.

By adding a shape and color to the Online Experiences graph, the eye can better distinguish between the overlapping points.

Visual Hierarchy

Creating each dashboard was not a linear process. Starting with the data itself, I had to decide what my criteria was for inclusion. The Impacts & Experiences dashboard took the most energy to organize and collate. A main goal is to collate different answers to the same question: what is the impact of online abuse and harassment?

Organized according to facet of the topic – the whole body of work is an intersectional look at the phenomenon. An individual dashboard investigates an aspect of online violence. Visual hierarchy of each dashboard is organized by related dimensions of the aspect. I took mapping and constraints quite seriously during the organization design process.

To ensure I’m compliant with copyright, as well as transparent and clear about where the data is coming from, I chose to display detailed citations and survey information for each statistic.

Color & Whitespace

Color contrast was important, as well as cohesive color schemes. Throughout, I took care to design figures that do not need a legend. I chose contrasting colors for accessibility (and am still concerned about how a screen reader would interpret my results). More minimalist than most of my work, I maintained a consistent color scheme of only purple, black and red. Figures were shades and hues of purple, headings in red, and everything else plain, aside from the pie charts. It would be too busy and distracting to have many colors, many figures, and many titles on one dashboard.

User Experience Study

The UX study was conducted primarily in order to support an iterative design process.

UX Study Goals

To better understand the target audience in terms of two broad categories.

1. General research methods when examining the same types of data included in the dashboard (i.e. gendered datasets, empirical evidence).
   • How they use, perceive and interact with data visualizations.
2. Gain insight into how target audience interacts with a dashboard I designed previously
   • Understand conceptual models of data visualizations in the sphere of international gender advocacy and policy.

UX Study Methods

A formative evaluation was conducted with two participants, individually, via Zoom. 

Study Parameters

• Type: Moderated test
• Interaction: Face to face (remote) individual conversation
• Participants: 2-3
• Two parts: 
  • Pre-task interview
  • Task completion think-aloud, while screen sharing
• Estimated time to complete: 30 minutes

Structure

Study Participants

Three women in my network fit my ideal participant profile: potential audience members, subject experts and researchers using similar data. Two of them responded: One is a subject expert on online violence against women in politics. The other is an expert in gendered data at the international level.

UX Study Findings

1. Research priorities
   1. When searching for data to use, whether it’s downloadable as a .csv or .xlsx is important
   2. Intergovernmental organizations like the World Health Organization, the Sustainable Development Goals or International Labor Organization were cited as the top places for reliable, fast, gendered data. 
   3. Some organizations that consistently publish one or two datasets via a map or interactive visualization, like the Women in Power Index, achieve their specific goals well; no organizational competing interests.
2. Data visualizations in the space
   1. The participant who is a mixed methods researcher makes and uses data visualizations in her work. The other participant considers herself a qualitative researcher primarily. She likes to use/reference data visualizations, but hires additional capacity to do quantitative work.
   2. There are very few people or institutions approaching the gender equality space from an information perspective, broadly defined.
3. System image and conceptual models
   1. My assumption that users didn’t need signifiers to discover the tooltip hover function was proven incorrect. (Gulf of execution)
   2. Participants expected there to be more ability to drill down into the data
   3. My ideas about participants’ use of data visualizations in their work were proven to be true. Note that this is topline.
4. Dashboard design
   1. The dashboard did not render properly on one participant’s screen (Windows 10, latest version of Chrome installed that day). Neither font nor layout was preserved, to my dismay.
   2. Context was lacking, it’s unrealistic for a user to have parallel screens with the report and Tableau up at the same time.
   3. Colors were not likeable
   4. Surprising findings could be highlighted better
   5. Prominence of the “source” text on the dashboard added legitimacy and value immediately, as both participants stated
5. Unexpected findings
   1. Both participants expressed the value that a dashboard like this will add. Given the little information I provided about my grand ambitions, I was mildly surprised. My research was also validated: both these women have much longer careers than me, and neither of them knew of a project like this.
   2. One was surprised to know that there was no institutional sponsorship or grant driving the project (aside from Pratt). She offered to share it with her networks towards that end, in order to expand the exposure and eventually scope.
   3. The other was excited most by the conceptual contribution the project represents, specifically the link between data visualization and open research. Notably, she reminded me that this issue is not taken seriously yet: recognizing the problem and creating a coherent summary is the work that needs to be done.
   4. The participants vary in seniority, skills, background and research priorities, which was most poignantly reflected in how they expressed enthusiasm and support for this project, the unexpected and candid part of my study. My choice in participants was further validated.

Findings

Dataset Accessibility

The data itself was not easy to find, and I did a lot of reformatting. I was surprised at how inaccessible it truly is to scale when dealing with different sources that format information differently. A lot of interesting information is in paragraphs in PDFs, and not easy to find. It’s not included in databases, because it’s formatted as part of a report and not a statistic.

Terminology

The terminology around the phenomenon of online violence against women is technical at best and frivolous at worst. Different agendas will use different terms to pursue advocacy goals, which is necessary to an extent, especially in policy and law on the rare occasions that survivors pursue justice.

My research involved investigating all these terms. Worth noting is that cyberbullying is largely applied to contexts with younger people and children.

Other common terms include incivility online, toxic content, and non-consensual image sharing (AKA “revenge porn,” which we no longer use). Sometimes cybersecurity and cybercrime surveys encompass questions about harassment, abuse or other violence, too. The International Center for Research on Women has a great explainer on technology-facilitated gender based violence, and PEN America defines many of the behaviors associated with this online violence.

All these terms are encompassed by the concept of the continuum of violence. This concept has been defined and applied in a number of ways by feminist scholars. For the purposes of this report, it can be understood as a useful tool to interpret the varied impacts, experiences, spaces and times that women are subjected to subversion. The continuum of violence recognizes that frequency and type of incident of violence do not adequately capture the scope of events that constitute violence, nor the ongoing and varied experiences of survivors. It’s the idea that a violent act does not occur in a vacuum, with a neutral victim and perpetrator, but that there are many socio-political, personal and other issues at play.

Glaring Gaps in Data

Datasets that are adequately disaggregated, meaning that they afford intersectional analysis, are extremely rare. Italy has by far the most comprehensive information about online violence, with Denmark and France in close second. A lot of my research was done using statista.com for expediency. That said, many case studies exist, and a lot of those focus on non-Global North countries. The statistics I found and used were limited to wealthy, predominantly white nations, which is unsettling.

There were few datasets disaggregated by gender, and even less surveys that allowed a space for anything but male or female under “gender.” Self-identifying characteristics are tantamount to analyzing the continuum of violence. So far, that information has yet to be collected and published at scale. It is promising to see some studies like ADL and OpenSociety ask about sexual orientation; Pew did recently as well. Given time constraints these were not included, but will be in future iterations.

Openwashing

The datasets from social media platforms are completely openwashed. Similar to greenwashing, where for example a fast food chain switches from plastic to paper straws to then brand itself as “green” and “eco-friendly” or “climate conscious,” openwashing is equally deceitful. It occurs when an entity (i.e. the government, a corporation) makes data available to the public that is convenient, uncontroversial, or very topline, to subsequently brand itself as “open access.” It’s also been applied to information transparency concepts, like Freedom of Information laws. 

Given this fact, I decided not to prioritize platform policies and responses, but left it for future work.

Future Directions

Revisions

I’m certain that the methods I used to connect dashboards is not the best way to navigate across them. It’s not elegant, involves a lot of manual formatting, and because Tableau publishes a workbook, all the sheets and data sources are included, which can be messy and off-putting for those unfamiliar. 

Expansion

Scaling up the project is a grand plan. Ultimately, there would be proper UX studies that do not include only experts in the field such as mine. I would include young women, men and trans folks, with all ethnicities, education levels, of varying ranks and stages in their careers. Feminist networks are fantastic at spreading calls for input, and at providing it. It’s a space focused on participation and I would tap into that. All this lays the groundwork for a large, admittedly quite ambitious, study on online violence against women and LGBTQI+ people. Rooted in iterative, participatory design.

Explore the outputs

1. Visual representations of the data in dashboards
2. Accessible datasets properly cited and collated, formatted for easy manipulation
3. Accompanying report and analysis

• My bibliography / literature review in Zotero.