Deathly Dataviz Talk Materials

Dispersion vs Disparity: Hiding Variability Can Encourage Stereotyping When Visualizing Social Outcomes

Collaboration with Cindy Xiong Bearfield, exploring the impact of variability on attribution and stereotypes.

Polarizing Political Polls: How Visualization Design Choices Can Shape Public Opinion and Increase Political Polarization

Collaboration with Cindy Xiong Bearfield, exploring how social normative influences are triggered by charts.

When people think of “good dataviz” they think of John Snow’s maps, because good dataviz is supposed to be intellectual and enlightening. It works by guiding us toward smarter, more rational decisions. But there are other ways that data can influence us, and a surprising amount of it relates to social psychology.

Public dataviz (e.g. dashboards published by big institutions like government health agencies) can also be influential, but not always in the ways we expect. Conventional ways of visualizing social outcome disparities are an example where conventional data visualization approaches can backfire.

They blame people, not systems.

Social cognitive biases can interfere with viewers' perceptions of dataviz. For example, health disparity charts may be accurate and easy to read, while still biasing viewers' causal explanations toward blaming the groups being visualized, instead of external factors like social determinants of health.

Conventional health disparity charts that promote misattributions for social outcome disparities pose downstream risks such as undermining political urgency for change and reinforcing disparities in treatment within the healhtcare system.

Visualizing people as monoliths makes them seem monolithic.

When visualizing social outcomes, highlighting within-group outcome variability can reduce misattribution and stereotyping. This implies that it's possible to design charts that are less toxic, and still effective for their original communication goals. It also implies that data designers have a choice, and a responsibility, in how they visualize social outcome disparities.

They undermine health outcomes for everyone. Conventional health disparity charts that promote harmful stereotypes pose downstream risks such as increased cardiovascular mortality and reduced support for programs like Medicaid.

Positively influential public dataviz starts with clear, complete communication goals. Data communication strategy should explicitly cover both equity and efficacy.

Identify and apply the equitable data design principles targeted toward equity communication goals. These can be used as inspiration while developing a chart, and as a checklist for evaluating the work.

The geo-emphasis chart in the "Must Be A Tuesday" study was designed with the following principles in mind:

• Blame Systems, Not People
• Highlight Within-Group Variability (and Between-Group Commonality)
• Defiantly Definite
• Visceral Value Judgements

Since dataviz is contextual and task-dependent, the best way to understand a design and be confident that it works is to test it out.

Conventional charts may feel familiar, but there’s no intrinsic advantage to them. In our "Must Be A Tuesday" study, we found that conventional bar charts and the alternative geo-emphasis charts were both effective for influencing health risk perception, behavioral intent, and policy support. This demonstrates the possibility that health institutions can change the way we visualize social outcomes, without sacrificing our original communication goals.

Data design choices, such as like assertive titles and showing variability, can play an active role in correcting health misbeliefs and harmful stereotypes.... or they can make them worse. Since these misbeliefs are associated health outcomes, it suggests making more equitable design choices when visualizing public health may indirectly benefit the public's health.