By the People, For the People

New visual languages for showing human-related data.

Photo from Unsplash taken by Tom Barrett.

Anthropographics. A big word, but we’ve all seen the products: anthropographics are human-related data representations using visual forms. They are often used to provoke action and empathy. In the paper “Showing Data about People: A Design Space of Anthropographics,” published at IEEE VIS 2021, a team of French researchers led by Luiz Morais established a framework to facilitate researchers and practitioners to communicate, design, and evaluate anthropographics. They examined 105 collections of visualization that share data about people to induce seven design dimensions for anthropographics.

Anthropographics are data visualizations or infographics about people.

Fig. 1: “Cancer is not always the end,” tells the story of 13 women who had breast cancer, represented by 13 flowers. The petals of the flowers convey information related to the symptoms or treatment. Source: Oncoguis

However, there is currently no explicit or clear language to communicate about anthropographics between practitioners.

We lack the system to categorize, evaluate, or choose the visual representation for the data to present the anthropographics we care about.

As a result, Morais et al. proposed a framework to help people organize anthropographics to guide our decisions regarding the patterns and trade-offs. They talk about the reference population and reference dataset as the subject of the visualization message and the dataset for that message, respectively. They also discuss how the attributes of the individuals in this reference population who are actually visualized, either through names, labels, or photographs. This yields a design space for what is shown in an anthrographic visualization and how it is shown:

Fig. 2: The seven dimensions of the anthropographics design space introduced by Morais et al. CC-BY 4.0. Source: osf.io/2zhjt/.

What is shown consists of four dimensions: the granularity (how many people per visual item), specificity (the distinctiveness of attributes), coverage (how many persons are visualized), and authenticity (the proportion of genuine attributes).

How is is shown, on the other hand, comprises three attributes: the realism (how much do the visual marks resemble actual people), physicality (the degree to which a visualization’s marks are embodied in physical objects as opposed to shown on a flat display), and situatedness (how close spatially is a mark to the physical person).

Based on these seven dimensions, the authors identify six common families of visualizations. The first three are non-anthromorphic visualizations where the marks do not resemble human beings: (1) Statistical charts present patterns rather than individual stories with a low granularity; (2) Information-poor unit charts show only a few attributes to convey several people with maximum granularity and low information specificity; and (3) Information-rich charts convey detailed information about each person with many attributes, maximum granularity, and intermediate to high information specificity.

Then there are the anthropomorphic visualizations where marks actually do resemble human beings: (1) Proportional wee-people charts that use graphic marks to represent a fixed set of people with intermediate realism and intermediate granularity; (2) Unit wee-people charts where each graphic mark represents one person, but the information specificity may vary, the realism is intermediate, and the granularity is maximum; and (3) Face charts that use photographs, portraits, or people themselves to represent and serve as data item marks with high realism and high information specificity.

Fig. 3: The overview of visualization families and the corresponding dimensions provided by Morias et al. CC-BY 4.0. Source: osf.io/p265q/.

This framework also helped classifying less common types of visualizations, such as (1) Embellished charts charts with partial authenticity, where designers might make up synthesized attributes, (2) Single-person charts of rich information about one individual and is often autobiography, (3) Example-driven charts that only selected a few people to tell a story, (4) Situated visualizations: that show information about people who are physically near the visualization, and (5) Physicalization that uses physical objects to present the persons. These uncommon types of families may contain irregular approach to present data. Some of these might even appear in art galleries or performance theaters for artists to promote public awareness on human subjects.

In sum, this framework offers us a new way to think about how graphs are used to move us. If you are a practitioner inviting people to “care for others”, maybe referring to this framework would help you make better visualizations. If you are a passionate reader, that is a better tool to help you examine your feelings and whether it’s “visual-driven”!

If you are interested in Morais and his works, visit https://luizaugustomm.github.io to see more projects!

References

• Morais, L., Jansen, Y., Andrade N., & Dragicevic P. (2021) Showing Data about People: A Design Space of Anthropographics. IEEE Transactions on Visualization and Computer Graphics, 2022, 28 (3), pp.1661–1679. 10.1109/TVCG.2020.3023013. hal-02931257v2f. https://hal.archives-ouvertes.fr/hal-02931257v2