Rhetoric and Feminist Critiques of Data and Visualization
Infrastructure: The Inventive Possibilities of Arrangement
Recent work by digital rhetoric and writing studies scholars demonstrates growing engagement with what Douglas Eyman (2016) mapped as one of the emerging directions in digital rhetoric: a socio-cultural and technological engagement with infrastructure, including "the [often blackboxed] software tools that allow nearly anyone to expertly remix content across multiple modes and media." Key to Eyman’s definition of infrastructure is an understanding of and fluency with the ways different kinds of technologies work and the social and institutional knowledge of the infrastructures in which these technologies are housed and employed. Rhetorical concerns with infrastructure have manifested in multiple special issues in digital rhetoric journals, like Present Tense's special issue on the Rhetoric of Platforms (Edwards & Gelms, 2018a), which argued that theorizing and practicing digital rhetoric is inseparable from considerations of infrastructure, or "a meeting point of hardware, software, and culture" (Edwards & Gelms, 2018b). Such research points to the way in which digital infrastructures create unique and constantly shifting constraints and possibilities for composition and agency. As Jim Ridolfo and Dànielle Nicole DeVoss (2009) observed, the infrastructure of Web 2.0 allows for the rhetorical velocity of texts and necessitates a digital update to the canon of delivery. John Gallagher (2017) revealed how the infrastructural layer of algorithms creates new media audiences that participate in the remix and remediation of digital texts and data. Within the growing body of writing studies literature focusing on infrastructure, the shape data and visualization take is not controlled solely by the researcher, nor is agency enacted solely by users, but co-constructed by the technologies and multiple situated contexts that become involved in communication and analytic technologies.
Seeing the Rhetoricity of Data, Databases, and Visualization through Feminisms
Not only are these technologies active players in our digital research practices and resulting knowledge products, but Kevin Brock and Ashley Rose Mehlenbacher (2018) argued that "code might produce and reproduce certain norms, values, and social actions" (p. 2). These authors also argued that "the 'digital,' as a collection of computational machines and the logics and experiences involved in using them, remains obscured for many" (p. 5). Access issues are deeply tied to technological and social infrastructures, as Aaron Beveridge (2018) noted by identifying several barriers for rhetoric and writing studies' engagement with data analytics and literacy, especially involving social media data. Equally as important as physical access, Beveridge emphasized the importance of access to "current research methodologies" (p. 245) for the study of large data sets often generated through streaming APIs like Twitter's. Although scholars like Laurie Gries (2015, 2017) and Beveridge (2018) are both advocates of creating data analysis and visualization tools especially for the specific research needs of rhetoric and writing studies scholars, and indeed both have projects underway to do just that, existing tools provide opportunities for rhetoric and writing studies to develop such methodologies. Gaining experience with data analysis and visualization technologies will allow us to push against their affordances and perhaps contribute insight into the ways in which future tools might be designed in more just and rhetorically informed ways.
Rhetoric and writing scholars have not been the only ones to emphasize the rhetoricity of data and the ways in which data and visualization structure knowledge. Feminist and critical technology scholars (D'Ignazio & Klein, 2016; Frost, 2016; Haraway, 1988; Hill et al., 2016; Leurs, 2017; Petersen & Walton, 2018; Tufekci & Wilson, 2012) argue against the presumed objectivity and neutral qualities often ascribed to data, pointing out that data is neither objective, nor neutral, but constructed, often many times over, through the process of collection, analysis, and visualization. In "Raw Data" Is An Oxymoron, Lisa Gitelman (2013) pointed out that "the seemingly indispensable misperception that data are ever raw seems to be one way in which data are forever contextualized—that is, framed—according to a mythology of their own supposed decontextualization" (pp. 6–7). The seemingly abstract, decontextualized qualities of data are perhaps what allows it to be such a convincing, though often imperceptible, vehicle of dominant cultural logics and social values. Indeed, Wendy Chun (2009) argued that "race and technology impact each other’s logic and development" (p. 8). In some academic and public circles, however, data is often presumed neutral and objective, fact-like, and unexamined. This neutrality, Gitelman (2013) observed, along with the "aggregative quality of data helps to lend them their potential power, their rhetorical weight" (p. 8).
At a recent Visualization for the Digital Humanities conference workshop, Catherine D'Ignazio and Lauren Klein (2016) advanced several feminist tenets for visualization and design, including the concept of challenging "claims of objectivity, neutrality and universalism, emphasizing instead how knowledge is always conducted within specific subject positions" (p. 2). For D'Ignazio and Klein, in part, this involves making the rhetorical choices of the researcher/designer and the affordances of the technology explicit, as well as disclosing particular experiences and background factors that may influence these choices. While positionalities are always in flux and have the possibility of transformation (Chaput, 2000; Massumi, 2002), rhetorical feminist data analysis and visualization methodologies should incorporate ways of explicitly understanding how positionality constructs data analysis. In "Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective," Donna Haraway (1988) redefined objectivity away from its conventional association with neutrality and instead emphasized "understanding how these visual systems work, technically, socially, and psychically, ought to be a way of embodying feminist objectivity" (p. 583). Rather than aspiring toward conventional notions of objectivity, the critical perspective Haraway advocates for is an embodied feminist objectivity that accounts for human and nonhuman decision-making and influence, particularly emphasizing how visualization works to create certain kinds of perspectives on data, often seemingly totalizing and conclusive.
In Haraway's wake, feminist research has troubled notions of objectivity by, in part, breaking down the binary structure between categorizations that structure our social and data-driven worlds. In writing studies, communication, and the digital humanities, feminist contributions have complicated distinctions between categories, such as the gender options available to Facebook users (Bivens, 2017; Dadas, 2016), notions of public and private (Penney & Dadas, 2014), approaches to distant and close reading (McPherson, 2014), and concepts of vacant and nonvacant housing (Knigge & Cope, 2006). Furthering this structural critique, feminist critiques of binary and other forms of classification point to the importance of understanding the way in which data is structured for analysis. These scholars argue that taken-for-granted norms in structuring data participate in reifying the kinds of binaries like those mentioned in the work above and participates in a larger history of classification that creates and replicates social hierarchies and power dynamics.
To combat this replication, D'Ignazio (2017) urged researchers and designers to critically examine our data structuring choices and the choices made for us by APIs and software. Rather than accepting the structure and content of data as it flows through infrastructures to become what we'd term a collection, she emphasizes the need to reflect an awareness that those collections will always be missing data and that existing data is shaped in particular ways. She further argues that these structures and gaps should be acknowledged and perhaps figured into our data analysis and visualization. Although this may indeed prove a less orderly process, Elizabeth Losh, Jacqueline Wernimont, Laura Wexler, and Hong-An Wu (2016) argued in "Putting the Human Back into the Digital Humanities: Feminism, Generosity, and Mess," that to recognize and address the "human" in digital humanities,
it is vital to attend to how corpora composed of supposedly neutral and transparent databases and tools may obscure the many ways that objects of study are positioned in relationship to human—and race, classed, and gendered—constructs of discovery, revelation, display, exhibition, desire, curation, witnessing, and bearing witness it is necessary to evaluate carefully and with discretion the type of datasets that the digital humanities employ. (para. 26)
Such scholarship urges us to interrogate our datasets and formulate intersectional feminist methodologies that de-center the positivism that underlies many assumptions of data; instead, we should strive toward deeper ways of knowing digital data that make apparent its very human, cultural characteristics. Perhaps most importantly, we should strive to attend to how the power hierarchies that structure our social world play a role in structuring our data, data analysis, and digital projects. By doing so, we may end up with methodologies and projects that are far "messier" but that better account for data's relationship to the human and social world.
Although the orderly form of the spreadsheet is nearly indispensable to data analysis and visualization, its rhetorical form and function are masked by its seemingly straightforward, orderly appearance. Johanna Drucker (2017) observed that "spreadsheets are rhetorically powerful encodings of values and belief whose graphical features participate in their appearance of neutrality—of 'unmarkedness'—through the way they graph their relations of power" (p. 909). The power Drucker referred to functions through the seemingly neutral organization of spreadsheets, and by extension, data. This rhetorical arrangement, Drucker noted, "enacts its power through an effective erasure of the structuring apparatus" (p. 909). To illustrate her point, Drucker dove into the form and function of data visualizations, arguing that they operate through established and familiar "conventions for communicating information in graphs, charts, or diagrams [to] make them readily legible. Perhaps, too readily, since the use of these familiar formats often allows them to be consumed without hesitation—or critical consideration" (pp. 906–907). Far from a neutral process, she argued that these familiar conventions mask how visualizations, and the spreadsheets that underlie them, are created through "authored activities," which include processes like the naming of columns and rows and the way items are classified and produced by humans and increasingly by technologies that feed researchers information in established categories that are then replicated in spreadsheet form (p. 908).
Data construction and analysis involve various actors and computational processes at many levels, including the collection and aggregation of this data into quantitative forms meant to identify trends and patterns in large-scale distant readings of the data. This distant reading method, according to Klein (2018), means that "these subject positions are too easily (if at times unwittingly) occluded when taking a distant view" (para. 9). Klein's observation of the distant view of data analysis mirrored Haraway's (1988) observation of most visualizations, that they create the visual perspective of a "conquering gaze from nowhere" and by doing so disembody both the data and the viewer (p. 581). Even interactive visualizations, Drucker (2017) pointed out, provide only "[t]he illusion of control or agency, even of authorship" (p. 912). This "illusion" of agency, as Drucker put it, "is produced by the interaction with the features, the dials, input boxes, and so forth that all speak me as they speak to me. You, they each say, are the one who does this, and not that, or that, and then this and so on" (p. 912). In this brief quote, Drucker explicated how visualizations and their functionalities work rhetorically as a form of direct address. Following Drucker, visualizations interpellate users into a particular perspective/ideology by a seeming lack of subjective authorship and the illusion that the viewer has control over the display and, perhaps, the underlying data. Through these controls, items like buttons and toggles that control aspects of the visualization, users develop a false sense of agency and control. This illusion has been predicated on the interactive possibilities already determined by the affordances of the particular technology and the many influences involved in data production, collection, organization, and analysis.
Given these critical feminist examinations of data and infrastructure, what are the possibilities of working toward more rhetorically informed data analysis and visualization? Although Klein (2018) critiqued the distant view involved in data mining, analysis, and visualization, she also pointed out that power inequalities, "like sexism or racism, are also problems of scale but they require an increased attention to, rather than a passing over, of the subject positions that are too easily (if at times unwittingly) occluded when taking a distant view" (para. 9). Far from rejecting large-scale data analysis and visualization, Klein argued that "we need to assemble more corpora—more accessible corpora—that perform the work of recovery or resistance" (para. 11). In "What Would Feminist Data Look Like?", D'Ignazio (2017) advocated for developing ways to design visualizations that talk back to data, partly by finding ways to explicitly include subjectivity and dispel the colonial, omniscient perspective Haraway observed. To accomplish this, D'Ignazio further argued for including dissenting perspectives and provides a concrete example through creating visualization labels. She cited a decades-old academic/community research and visualization project, The Detroit Geographic Expedition and Institute of the 1960s (kanarinka, 2013), which used titles that reflect a dissenting perspective rather than one of objectivity, like the map entitled, "Where Commuters Run Over Black Children." Using this example and others, D'Ignazio pointed out that feminist visualizations should strive to push against data's disembodied fragmentation and situate data in geographical locations and bodies. This subjectivity might also include attending to the positionality of the researcher. Although positionality has been theorized in its relationship to ethnography, pedagogy, and cultural rhetorics, positionality—identity markers that influence how we construct and understand our world—is far less attended to in relation to digital methodologies, especially around data analysis and visualization. By attending to positionality and incorporating subjectivity into data visualization's design, we disrupt the presumed neutrality of data and the empirical positivist perspective embedded in much data visualization.
Besides disrupting objectivity and neutrality, feminist methodologies can incorporate rhetorical subjectivity by becoming more attuned to the silences and gaps within scholarship and within one's individual research archive (Graban, 2013; Lay, 2004; Royster & Kirsch, 2012) and by participating in recovering stories of activism. Emily January Petersen and Rebecca Walton (2018) argued that attending to silences and gaps in our research methodologies leaves room to find what is missing in our data, methods, and methodologies: "filling gaps and silences are an essential goal of feminist research. One way to find these gaps is by using research methods designed to recognize them. For example, feminist content analysis is a way of generating themes from textual data, allowing themes to emerge based on 'close readings, coding, and memoing'" (p. 427). In this brief quote, Petersen and Walton connected content analysis to the feminist practice of recovering the voices and contributions lost in the silence and gaps of both dominant narratives and computational quantitative analytics.
Although technical communication and professional writing are often viewed as apolitical and outside the realm of social justice work, in "Apparent Feminism," Erin Frost (2016) made the case for a resurgence of feminist-oriented research, which, as she saw it, is especially important due to the public nature of technical communication and the way that such communication is often consumed as value-neutral and objective. For Frost, one way to combat this is through storytelling: "Apparent feminists can take up this call to tell such stories and, thus, insert feminist apparency into the public sphere. The first step in doing so is to pay attention to activists operating in the public sphere" (p. 8). Discussing the transformation of the rhetorical situation from the objective observation of Lloyd Bitzer's model to Richard Vatz's, which Frost reads as more attuned to the role of perception in creating rhetorical interpretation, she stated, "Indeed, when considering the urgent need for apparent feminism, we see that the situation is highly rhetorical and that particular kinds of utterances, texts, and rhetorical acts create public biases—or, we might say, crises—which in turn demand an apparent feminist response" (p. 12). Natasha N. Jones, Kristen Moore, and Rebecca Walton (2016) also argued for a closer relationship between technical and professional communication (TPC) and activism: "Although TPC scholars have long been exploring issues of inclusion, the collective contribution of this work has gone largely unnoticed, (over)shadowed by the dominant narrative that technical communication is most concerned with objective, apolitical, acultural practices, theories, and pedagogies" (pp. 211–212).
To conclude this section, I offer the following framework as a way to highlight shared principles, along with methodological and design suggestions synthesized from the writers represented above. This framework is not meant as a step-by-step recipe, but instead as flexible principles that also elucidate my method-in-practice for studying gentrification encounters through a large scale Twitter data collection. It's my hope that others might find this framework useful, and I welcome contributions to this methodology that explore the possibilities and limitations of large data collection, analysis, and visualization.
Framework for Feminist Rhetorical Data Visualization Methodology
- Incorporate an awareness of the way in which data and visualizations are authored in unseen ways by humans and technologies. Ask how we might
- augment and challenge existing structures,
- bring awareness to the way in which data and classification both ignores and creates difference, and
- apply a critical lens to and make apparent the way in which framing and other visualization features create a form of direct address, structure certain kinds of argument, and position viewers as passive receivers of certain kinds of knowledge.
- Attune to silences and gaps and recover what might be missing from the archive or data collection. This might include
- locating and emphasizing what's overlooked by quantitative measures,
- creating visualizations that counter dominant perspectives,
- asking what isn't reflected in the resulting visualization, and
- coding the data using grounded theory to include themes that might be missed using other means of analysis.
- Attend to subjectivity and consider incorporating dissent in the representation of data and conventional features of visualizations. This could manifest in numerous ways, such as through
- choosing categories and structuring data,
- making decisions about labels and naming practices of visualizations, and
- including communities in the shape of research and the visualizations produced.
- Create visualizations that put big data in conversation with small data and other ways of knowing. This might be achieved through
- situating the data in embodied and geographic contexts (D'Ignazio, 2017; Haraway, 1988),
- incorporating other data types like interviews and participant communities, and
- using data visualization to tell stories of activism.
The following pages of this webtext demonstrate some of the ways in which I've practiced the tenets of this framework. Although some sections correspond more closely to certain elements of the framework than others, often emphasized through headings, many of the elements of this framework are interconnected and part of a recursive process, one that I'm continuously striving to better actualize by gaining feedback and input from scholars and communities alike.