Representing Marginalized Populations: Challenges in Anthropographics

Anthropographic visualizations usually do not include identifiable details of the individuals being represented. Rather, most examples use a small set of human-shaped templates which are repeated to represent larger groups of people. The use of similar human-shaped icons and visuals to represent distinct marginalized populations erases the demographic differences between these groups and treats marginalized populations as interchangeable. Although as McCloud argues  [17], homogeneous human shapes are abstractions that can be understood as representing people generally, the use of generic human shapes to represent marginalized populations with different demographic characteristics can obscure distinctly different cultures and lived experiences. Furthermore, representing marginalized populations as interchangeable can influence how they are publicly perceived, which might affect the implementation of policies directed towards them  [5, 13].

For instance, in Figure 1, the choice to represent migrants from Southeast Asia and migrants from the Middle East using the same shapes [19] can confound the demographic and cultural differences between these two groups. Representing these populations as interchangeable hides the unique experiences of marginalization, causes of death, and humanitarian issues each face. Similarly, in Figure 2, the use of similar visuals to represent many demographically-distinct victims of the COVID-19 pandemic erases important differences such as social class, race, and gender [2]—despite the fact that many of these demographic differences also likely influenced their COVID exposure and cause of death  [5, 13].

The availability of data about populations can also influence which anthropographic projects they are included in. Because data about marginalized populations may be incomplete or require special considerations for privacy and anonymity—demographic data and other contextualizing information about these groups is often sparse. Meanwhile, the data that is available about marginalized populations (including groups like refugees, unhoused people, or abuse victims) often focuses exclusively on the issues affecting them—such as conflicts, death, and suffering. This can result in visualizations that repeatedly and exclusively focus on marginalized groups through the lens of victimhood. For instance, data stories about past and ongoing conflicts around the world have sought to use anthropographics to raise awareness about the state of marginalized populations. Yet, designers of these data stories have often used graphic images of death and starvation alongside anthropographics, charts, or maps—showing these populations in demeaning positions and reinforcing stereotypes about their marginalization [7]. One recent example is an article by The Guardian that reports on famine resulting from current conflicts in countries such as Ethiopia, Somalia, South Sudan, Yemen and Afghanistan that includes graphic images of starvation juxtaposed with charts showing data about affected populations [7].

Issues surrounding marginalization as an inclusion criteria are not unique to visualization research. In fact, there has been recent interest in journalism in re-examining how persistent depictions of discrimination and suffering (instead of positive stories about educational, social, and political achievements) can normalize dehumanizing narratives of marginalization [20].

The relative lack of visualization designers and researchers from marginalized groups also influences how data about marginalization is understood within complex social, historical, and political contexts [15]. In particular, designers without sufficient cultural context may risk reinforcing stereotypes or caricatures of marginalized populations. One example is a data physicalization project titled “Slave Voyages: Reflections on Data Sculptures” that shows the journey of enslaved people from Africa to the Americas (Figure 3). In this project, the researchers used colored beads to represent data about enslaved people—including whether or not they survived the transatlantic journey [14]. However, the researchers were unaware that glass beads were used to barter for enslaved people and were significant for their Black viewers until after they completed the project [14]. Because of the lack of cultural background and domain knowledge, there was little consideration for the significance of the chosen material and how viewers might perceive the design choices in the context of historical and familial connections to slavery.

Similarly, the data journalism article from the Associated Press seen in Figure 4 uses a color palette ranging from gray to dark brown to show the distribution of Afghan refugee resettlment across the US [16]. This color palette had been used extensively in other data stories not involving marginalized people, but in this specific context the graphic’s beiges and browns can easily be misinterpreted as skin tones. This choice of color palette is problematic considering how non-White immigrants are often stereotypically perceived in the US. Specifically, darker tones might reinforce racist stereotypes about immigrants changing the racial demographic composition of a state to be less White. Here, an awareness of racial equity, as Schwabish and Feng emphasize [23], would have been useful to consider how the choice of color palette might be received in the context of the broader history of immigration to the US.

Most current anthropographics do not convey the demographic differences between marginalized groups. Yet, representing these demographic differences positively can help address issues surrounding homogeneous anthropographics. The process of creating representative, contextualized, and humanizing anthropographics of marginalized populations would almost certainly benefit from input from the populations being represented. We thus advocate for approaches such as participatory design that include marginalized populations in the design process of anthropographics. More broadly, designers and researchers should also seek to understand how marginalized populations perceive homogeneous anthropographics and how (or if) they wish to be represented.

While non-homogeneous anthropographics have the potential to provide more equitable representations of marginalized populations, we also acknowledge concerns around the practicality and usefulness of such representations. Often researchers and designers do not have access to demographic data or good visual mappings for encoding it. Even where demographic data is present, creating humanizing and representative anthropographics can be challenging. If executed poorly, such approaches can even result in misrepresentation, harmful stereotypes, and caricatures. Creating distinct anthropographics for every demographic group being represented can also be time consuming for larger population-scale datasets such as the dataset about COVID-19 deaths in the USA (Figure 2). Despite these challenges, we emphasize that the creation of non-homogeneous anthropographics is a worthwhile endeavor to equitably represent the diversity among marginalized populations.

Anthropographics have largely been used in projects that have attempted to provoke prosocial feelings towards marginalized populations. However, there are other potential uses of anthropographics that have yet to be explored—including supporting better understanding of complex demographic datasets, helping people to see themselves or people like them in the data, and creating opportunities for storytelling and engagement. Anthropographics have also not been evaluated for unintended side effects. For instance, homogeneous anthropographics could possibly alienate viewers from the people being represented or further harmful stereotypes about marginalized populations being interchangeable. More positive effects of anthropographics could include ensuring equitable representation of marginalized populations and dispelling stereotypes.

That said, the use of representative, non-homogeneous anthropographics may be limited to projects that involve adequate demographic data. These projects require considerations for the privacy of individuals being represented, particularly when anthropographics are realistic. Additionally, the types of data used might impact where anthropographics might be most useful, how these visualizations are evaluated, and what side effects are observed. On the other hand, demographic data is used in sectors such as financial services, healthcare, education, housing, immigration, and video games, among many others [1]. This opens up new avenues for exploring creative ways of designing non-homogeneous anthropographics for diverse uses.