Your Pol(l)s May Be Lying to You

How text visualization can help in understanding politics.

Image by MidJourney (v4).

Design has always been affected by politics. Take the example of Robert Moses, an urban planner from the 1930’s who notoriously designed bridges and archways that restricted access to public vehicles to certain areas of New York City that disproportionally affected minorities. There are several instances throughout history leading up to the present day which have been studied and well-documented in the field of design. However, when it comes to visualizations of public opinions, the effects of politics on these designs have been largely ignored. In a recent work published at the EuroVis 2022 conference, Baumer and a team from Leigh University and University of Massachusetts took it upon themselves to remedy this gap.

When a public figure such as a politician desires to know the opinions of the masses on a topic, it is done through a variety of means, surveys, opinion polls, and interviews being the most prominent among them. The data gathered is then processed by analysts who then visualize it to have it make sense to politicians and the public alike. The most common methods involve civic text visualizations. One of them is shown in the figure below.

Example of a civic-text visualization.

Public leaders in positions of power use civic-text visualizations to make decisions regarding policy. However, there are inherent issues with themse visualizations that are often overlooked. These issues stem from the notion that visualization is a completely analytical and logical process that is devoid of a political angle. On the contrary, visualization is a political process, one that grants representation and authority to some voices while potentially marginalizing others. Below are some of the inherent challenges in visualizing civic text that Baumet al. identifies in this paper.

• Poor collaboration support: Bureaucrats who are responsible for decision-making often outsource the analysis of public input to analysts. While they often employ multiple analysts to avoid individual bias, the tools and techniques used by these analysts are designed primarily for an individual. The functionalities they provide are not necessarily suitable for collaboration.
• Lack of expertise: People who are trained in visualization are often not included in the design process. Instead, the people who end up creating the visualizations often have local political knowledge and deep subject matter expertise, but little data visualization and analytics expertise.
• Balancing complexity and accessibility: It is difficult to strike a balance between visualizing data that is representative and yet easily understood. When the data has a rich detail, the visualizations become difficult to interpret and thus inaccessible to but a few experts. On the other hand, simplifying the visualizations leaves out the underlying text that is critical to understand the subtleties of a topic.
• Ensuring representation: These days, digital devices are the primary way of collecting data from the public. However, not everyone has access to technology which might leave them from voicing their opinions.
• Biases: The analyst’s personal bias may creep into the data while they are trying to visualize them.
• Process issues: Creating data visualization is a linear process that involves several people in different roles. Due to its linearity, there is an inherent disconnect between everyone involved; see the following figure.

The process of creating a civic-text visualization.

How do we recognize and combat the politicization of our public opinions? Instead of changing the process of how our visualizations are done, we might account for it by being mindful of a few polar concepts while we create them. The below figure illustrates these concepts.

Polar concepts that we need to be mindful of when we design Civic-Text Visualizations.

1. Data and metadata and provenance and paradata: Apart from synthesizing the opinions we receive from the public, it is also imperative to understand where we have received the data from and the methods we have used to collect it.
2. Prescription and interpretation: Visualizations should strike a balance between being prescriptive and open to interpretation.
3. Singular user and multiple relations: Visualizations are often designed with a single user in mind without considering the entire audience. While they might be useful in certain circumstances, it is also important that they are readable by a wide variety of people.
4. Complexity to inclusivity: Complex visualizations may be rich in detail and add value. However, only a few experts may have the knowledge to interpret them causing them to be inaccessible to others. Thus, there should be a balance between complexity and simplicity.
5. Aggregation to articulation: Aggregating public opinion and forming visualization gives a great idea of the bigger picture, but runs the risk of leaving out the subtleties of the topic and the opinions of minorities. However, articulating every opinion makes it impossibly hard to understand the data and thus a balance must be found based on the purpose of the visualization.

We have seen how politics are an inherent trait of visualizations of public opinions and have learned about certain techniques that analysts use to recognize and combat it. However, as a reader, the next time you look at a public opinion visualization, Baumer and team tell us to think about what the graphic isn’t telling you.

Reference

• Baumer, E. P. S., Jasim, M., Sarvghad, A., & Mahyar, N. Of Course, it’s Political! A Critical Inquiry into Underemphasized Dimensions in Civic Text Visualization. Computer Graphics Forum, 41(3):1–14, 2022.