The Datatext: A Multilevel-discursive Theory For Improved Public Health Data Visualizations

Abstract

In the run up to the COVID-19 lockdown in March 2020, Prime Minister Boris Johnson challenged the British public to “squash the sombrero,” and so save thousands of lives in the event of the pandemic overburdening an already stretched National Health Service. There was a jarring sense of incongruity between this tabloid metaphor, and the minimalist line-graph to which the prime minister was referring. Best practice in infographic design may be well-suited to the communication of data amongst scientists and other literate audiences. But today matters of public health are subject to debate between citizens who are actively engaged in creating and circulating knowledge amongst wider publics with variable levels of literacy. Here a different epistemic approach, and different assumptions about design, are required. When conceiving of the infographic in public health as a multilevel discourse containing visual arguments mutually re-enforced by combinations of words, numbers and images, what I call a datatext (after W.J.T. Mitchell), it may be possible to design more effective communications. In this paper I set out a theoretical approach to infographic design drawing upon image schema theory, as well as conventional best practice. I conclude with recommendations for designing effective datatexts for optimal biocommunicability.

Keywords:

• data visualisation
• standards
• best practice
• discourse
• public health

Introduction

Towards the conclusion of the UK government’s first official Coronavirus (COVID-19) pandemic press conference on March 23rd, 2020, Prime Minister Boris Johnson challenged the British public to “squash the sombrero.” This allusion, delivered with characteristic tabloid alliteration, proved popular in sections of the press. It added a touch of levity (and no little cognitive dissonance) to the formal communication of a profoundly serious public health crisis. Stopping the pandemic could, apparently, be fun, rather than merely a burdensome chore. The sombrero requiring squashing in this instance, was a minimalist line-chart to which the Prime Minster and his advisors made reference throughout the briefing (HM Government 2020). Uncharitable commentators will no doubt attribute the incoherence between this sober, minimalist chart and the “fun” metaphor, as being in keeping with a wider sense of chaos in the UK government’s communications strategy, during the early stages of the pandemic. Those more inclined to charity may instead see these two very different approaches as a pragmatic means of getting the government’s message across to a diverse public, who are used to acquiring their news in a diversity of forms, from a diversity of sources. Infographics and data visualisations (I will use these terms interchangeably throughout, as synonyms or near-synonyms) have, it is argued, the potential to save lives (Bergstrom 2020). But how they may be harnessed most effectively towards achieving optimal public health outcomes, particularly in times of crisis, is another matter (Kennedy 2020). In this paper I present an approach to infographic design that addresses the disjuncture at the heart of this public health communication. The effectiveness of infographics designed using this approach may be measured in terms of their memorability, rather than in terms of actions or material outcomes. Factors that inform public responses to the information communicated, such as in the context, for example, of tensions between economic and health policies, are not included in the purview of this proposal. Instead, I will argue that by accommodating the public’s habituation and ritualistic familiarity with certain highly aesthetised graphical forms within a standardised best-practice approach, a more effective biocommunicability of risks is possible. By biocommunicability I mean an ideological regime that bounds the communication (comprising production, circulation and reception) of health and medical information as a process in governmentality (Briggs and Hallin 2007).

I will set out to explain the mismatch between how visually rich designs (including pictographs) are viewed in the sciences, and how they are viewed in the public sphere. I will then set out a theoretical outline for the datatext – a graphical form that accommodates normative standards such as accuracy, and other criteria, including metaphors, within a theorisation of the data visualisation as a text comprising a multilevel discourse.

Standards and Aesthetics in Graphic Design Best Practice

The minimalist charts used in the UK government’s COVID-19 briefings, are typical of infographics shared between scientists, and other white-collar professionals, and more broadly between those familiar with standards of best practice in the field. These are minimalist, neutral, standardised and (seemingly) self-evident forms. They are far removed from visually embellished graphical forms, such as pictograms, that are often unloved in histories and textbooks in the field of infographic design. Criticisms of pictograms fall broadly into three categories, concerning their association with propaganda (De Foville 1887; Funkhouser 1937; Chizlett 1992), problems with techniques used in their design (De Foville 1883; Brinton 1939) and matters of taste (Tufte 1997). They are, whether as; pictorial charts or pictographs; whether proportional or unit-based, often either omitted from indexes (Friendly 2006; 2008a; Tufte 1983; Wainer 2006), or otherwise outright impugned; as chart-junk (Tufte 1983), as an innovation “perhaps best forgotten” (Friendly 2008b, 30), or as a form suitable only for children (Few 2010).

Yet despite this historic condescension, embellished infographics endure. Today there is a growing body of research in the fields of risk and health communications, that advocates for their use, particularly amongst low-numerate social groups (Nelson et al. 2008; Hawley et al. 2008; Tait et al. 2010a; Fagerlin, Wang, and Ubel 2005; Galesic, Garcia-Retamero, and Gigerenzer 2009; Tait et al. 2010b). Recent research suggests that they may be particularly useful in communicating health concerns, such as antimicrobial resistance (Walker 2019), and even that their use, albeit in certain circumstances and aligned with modest outcomes, may be effective in influencing public opinion regarding health policy goals and initiatives (Reynolds, Pilling, and Marteau 2018). While it has been found that too many pictograms can distract (albeit in a costless way), nonetheless, they may aid recall during demanding tasks, because they encourage closer inspection of the data (Haroz, Kosara, and Franconeri 2015). How, then, to account for this seeming discrepancy between normative best practice on one hand, and practical clinical outcomes on the other?

Common Assumptions about Data Visualization in Best Practice Literature

The dominant approach amongst scientific and social-scientific texts in the data visualisation literature, may be described as logico-semiotic, and may be said to be bounded by three biases that I will address in turn. These are:

1. The notion that designer and audience are, first and foremost, rational actors.

2. The perceived importance of separating information from persuasion, and;

3. The emphasis (implicit or explicit) on a transmission theory of communication.

Firstly, the idea that designers of data visualisations adhere to logic and reason in their design-making decisions (and that audiences do too, in their consumption) is rooted in the logico-semiotic origins of the field (Bertin 1983 [1965]). And yet there is more to effective communication (whether in transmission or in reception) than reason alone. Cognitive, affective, and behavioural effects upon audiences may contribute towards establishing the communicative efficacy of certain media forms (Ball-Rokeach and DeFleur 1976), and hence adverse outcomes in any of these measures may limit or discourage audience engagement. The rational actor model has been subject to criticism since the mid-1970s. An over-reliance on heuristics and cognitive biases has been found to account predictably for errors of judgement (Tversky and Kahneman 1974). Here it may be helpful to revert to an old anthropological concept; namely that human beings are less rational animals, than they are symbolic animals (Cassirer 1944). Experiences that publics bring to their interpretative vantage point may shape their preparedness or willingness to engage with data presented in certain forms, and these are factors that may in turn inform optimal design.

Secondly, regarding the classical positivist notion that seeks to separate explanation from interpretation; wherein facts are able to simply speak for themselves. Certainly, to the specialist, data visualisations (when created according to accepted standards) represent evidence; proof of a verifiable fact (or facts) about objective reality. When people in these groups communicate visual data with each other, they often do so in minimalist terms; stripping out unnecessary decoration and focussing on the data; safe in the knowledge that their audience is sufficiently familiar with the conventions employed. But these assumptions do not necessarily hold true for wider publics, and indeed today some scholars in graphic design have begun to question the extent to which there may be a neat, and convenient dividing line between fact and argument (McCoy 2000). It may therefore be sensible to assume that, in order to communicate public health objectives optimally, with a view to informing demonstrable change in public attitudes and behaviours, merely communicating the facts may not be enough. In keeping with Otto Neurath’s public health pictograms, they may need to comprise “visual arguments,” in order to be optimally effective (1944).

Thirdly, with regard to the (often implicit) focus on communication as a process of transmission, within the literature. This perspective, whether explicit or implicit, tends to prioritise individual agency (of the designer, or the audience) at the expense of wider social and communal communicative cues, and contexts such as sharing, participation, association, and fellowship. In the field of communication, particularly amongst scholars informed by the British Cultural Studies of the 1970s, over-reliance upon this model has long been contested. Alternatively, a pragmatic approach concerned with communications to and amongst publics, might instead conceive of communication as comprising both modes of transmission and modes of ritual (Carey 2008). In this approach it may then be helpful to draw upon signs, symbols and icons familiar to particular publics, and to base an appeal upon these phenomena, an approach Neurath made use of, in developing the Isotype picture language (1944). When analysing the effectiveness of data visualisations in health communication, it may be helpful to conceive of this process in terms of a different paradigm, one perhaps better suited to analysing the sphere(s) in which publics acquire and circulate knowledge about public health.

Chartjunk, Tabloidization and Communication in the Public Sphere

In attempting to formulate a coherent critique that might encapsulate a range of design sins (from misleading cues, to aesthetic faux pas), Edward Tufte (1983) coined the term “chartjunk.” Central to this concept lies a tension in infographic design, a balance between maintaining the fundamental integrity, and accuracy of visual display, and the need to engage the viewer (or audience) in the data (Dick 2017). Poor standards and bad taste were not the only bounding factors of this concept, however:

Lurking behind chartjunk is contempt both for information and for the audience. Chartjunk promoters imagine that numbers and details are boring, dull and tedious, requiring ornament to enliven … Credibility vanishes in a cloud of chartjunk; who would trust a chart that looks like a video game? … Disrespect for the audience will leak through, damaging communication. (Tufte 1997, 34)

The notion that visually embellished graphical forms lack credibility may be seen, from the field of communications, to have something in common with a wider debate about standards of political discourse in the mass media, that has come to be defined as tabloidization. Indeed, the two concepts, chartjunk and tabloidization, are historically contingent. The rise of infographics as a popular (and routine) form in the UK may be traced to the 1950s most popular (and most lavishly resourced) mid-market newspaper, The Daily Express, a title with strong associations of propaganda (Dick 2015), aimed at a mass audience, in a commercial approach that Martin Conboy has called the “attractiveness of orchestrated variety” (2002, 75). The concept of tabloidization is a slippery one. It is identified as having at least three usages, namely, (1) newspapers and broadcasters who prioritise private lives over politics in their coverage, (2) a medium-specific manifestation of this process of re-alignment, and (3) changing in the boundaries of matters of taste and decency in the media (Sparks in Sparks and Tulloch 2000, 10–11). As a process, it has been suggested that tabloidization (under definition 2, above) leads to the dissolution of categories of private and public life, to the integration of politics into many areas of life, and hence to a shift in political communications towards the visual in a performative sense, moreover, towards an “aestheticisation of the public sphere” (Holly 2008, 324). The manifestation of chartjunk as a form of tabloidization may conceivably be observed in any of the above definitions, however it is perhaps most clearly associated with matters of taste, discernment, and credibility.

Tabloidization has both its detractors and defenders, as well as those who recognise distinctions between conventional “high” and “low” definitions as ideologically loaded by the context from which they arise (Hartley 1996). To some it represents a “dumbing down” of journalism's public function, and a squandering of its fourth estate legacy. On the other hand, it may be countered that such hostility merely represents snobbery “grounded in a conventional and long-standing hostility to popular culture itself” (Turner 1999, 63). The logic of tabloidization as a process that seemingly locks growth in readership to growth in advertising income in a self-perpetuating cycle, may well lead to yet further tabloid colonisation of the mainstream media, but it leads too, crucially, to increased democratisation (Randev 2014).

At the root of many criticisms of tabloidization there seems to be a concern about the efficacy of visual material (or images) being increasingly used to supplant words in the media’s political discourse. Historically, for adherents to the Critical Theory tradition within the Frankfurt School, spoken and/or written languages have been seen to be oppositional to images, the perceived linear nature of language comprehension and the potential in grammar to convey complex ideas, perceived as standing in opposition to the simultaneous, or holistic interpretation required of images; the former seen as being key to political discourse, the latter essential to propaganda. In a contemporary articulation of this idea, summarising the views of Robert McChesney (2007) on tabloidization in newsprint, David Rowe explains: “if there was a corresponding increase in the size of visual material (especially so-called ‘impact’ photographs, as well as diagrams, tables and cartoons) it might … be claimed that textual depth is being sacrificed in the interests of visual attraction and, perhaps, distraction” (Rowe in Allan 2009, 353). Such criticisms have a long antecedence. For example, the incorporation of photography into news culture, during the mid-nineteenth century represented a “threat to reason, and the journalistic institution’s Enlightenment heritage.” (Becker in Dahlgren and Sparks 1992, 129). The subsequent rise of the sensational form of imagery we now associate with the tabloid press emerged during the 1920s in the US (Becker in Dahlgren and Sparks 1992, 132), that in turn led to a change in public perception, namely: “ … the leading role photography was playing in the tabloids’ abuse of press credibility made it increasingly difficult to see the photograph as a medium for serious news” (Becker in Dahlgren and Sparks 1992, 134).

In light of this debate, and given the contentious origins of highly embellished infographics, one might ask, can these forms ever be a fitting medium for communicating serious, trustworthy news, or information? To answer this question, one need only visit the online editions of our daily newspapers. Today interactive infographics are designed to the highest of standards and are mobilised towards providing different ways of understanding some of the most serious, and complex of news stories, as well as some of the most personal, and trivial (Dick 2014). What may traditionally be considered a manifestation of tabloidization, is today capable of expressing both serious and trivial news, with little awareness of the classificatory distinctions between news “types” often expressed in the literature. Today’s interactive graphics engage news audiences on serious matters by a range of means that textual or verbal forms might struggle to achieve (Dick 2014). The Daily Express’ embrace of “chart junk” during the mid-1950s may be considered to represent the emergence of “mass culture” (an aesthetically appealing, but nonetheless debased rendering of “popular culture”). However, it would be false to presume that highly embellished infographics are incapable of contributing meaningfully, and effectively towards the communication of health information merely because of what has come before. Where we conflate medium and message, we risk reducing complex matters to tropes, and so give ground to cognitive biases, such as the bandwagon effect, that may in turn cloud our judgement.

Situating Infographic Design within the Public Sphere

Scientific data in the form of diagrams and models have long played an important role in shaping public understanding of complex phenomena, as well as in furthering research and the production of new knowledge (Nikolow and Bluma 2008). The success of these forms seems to depend on an inherent tension in the appeal to mass audiences, between “semantic flexibility and identities for different groups of recipients” (Nikolow and Bluma 2008, 34) on one hand, and a general robustness, coherence, or for want of a better term, an appeal to pragmatic common sense, on the other. The history of the infographic is, moreover, intimately connected to the progressive pursuit of political realignment. Queen Victoria and Albert were moved by the graphical forms of Joseph Fletcher (statistical maps) and Florence Nightingale (polar area diagrams, and other experimental charts) (Dick 2020). From the late eighteenth- to the mid-nineteenth centuries, these forms often seemed to serve as a means of facilitating the brokering of a new political pact, between ancient power, and a new increasingly powerful, and sharp-elbowed improving middle class (Dick 2020).

Understanding what precisely infographics may contribute within the public sphere, and to what effect, depends on how we define the public sphere, whether (for example) in the “folk” sense (in terms of being concerned with public space), as “agonistic” (after Hanna Arendt); as classical liberal (after John Stuart Mill) or as relational (in terms of comprising various groups with various identities and group-attachments) (Wright 2008; Kollar and Wodak 2008). Here I will draw upon a theorisation of public health information in the public sphere developed by Briggs and Hallin (2010), based on the concept of biocommunicability. This idea stands apart from earlier approaches that conceive of the politics of the circulation and creation of medical knowledge based either on a medical authority model, or on a patient–consumer model. In Briggs and Hallin’s approach, people are conceived of as active citizens, rather than docile patients or transactive consumers. These publics actively engage in the production, evaluation, and circulation of health information in complex networks, interspersed with supplementary information from mass media, journalists and experts. Within this context and removed from the scientistic assumptions that inform much of the best practice literature in infographic design, an alternative philosophy of data visualisation design may be situated.

The Politics of Interpretation

If we remove ourselves from some of the suppositions that underpin much of the best practice literature in data visualisation, it may be possible to derive an alternative theoretical perspective; one that accommodates important components of conventional standards and best practice (concerning, for example, accuracy and proportionality in design), but that balances these factors against other important factors, including; audience needs, the embrace of argumentation in particular contexts, and not least the embrace (rather than the forcing apart) of transmission and ritual modes of communication. Such an alternative approach may be identified in the work of art historian W.J.T. Mitchell. In a wide-ranging critique of 1990s visual studies, Mitchell (1995) observed that too often theorists had become absorbed in the minutiae of social codes concerning matters such as spectacle and surveillance, without bothering to set out a robust conceptualisation of the objects of their study - images, and the words associated with them. Dismissing traditional distinctions between image and word in visual culture as an arbitrary relic, Mitchell proposed instead that we conceive of such forms as imagetexts; as collections of words and images whose meanings are inseparably encoded, and whose scope of interpretation is far removed from the reductive strictures of normative literacies. In this approach, imagetexts are expressed in language, which is in turn conceived of as a “medium rather than a system, a heterogenous field of discursive modes” (Mitchell 1995, 97).

The theory I presents here is situated within a multi-perspectival epistemology that acknowledges the operation of linguistic (and visual) signs as operating on different levels, simultaneously. It starts from the premise that the data visualisation is a text comprising a multilevel discourse (Eco 1976). In any data visualisation it is possible to discern four discrete (albeit contingent) discourses, where the form is conceived of as Method, as Aesthetic, as Tool, and as Ideology (hence the TATI model of multilevel discourse). The categorical emphasis between these discourses is a function of variation in style, and between these (at times) competing discourses the outline of different forms, genres, or classes of data visualisation may be observed. So, for example, conventional scientific and social scientific data visualisations are, by degree of emphasis, a form of evidence or proof primarily, and a method for exploring some factual aspect of the world. Alternatively, Neurath’s Isotype served, first and foremost, as a form of ideology; while the infographic designs of, for example, David McCandless, are, by degree of emphasis, works of aesthetic expression, first and foremost (Dick 2020).

Data visualisations designed for biocommunicability may be epistemically situated within a multilevel discursive theorisation of the data visualisation as a tool, or as a means of achieving change (ibid.). This approach is informed by John Dewey and Sidney Hook ([1916] 2008) and others in the pragmatist tradition. According to cultural historian Walter Ong, “Writing and print and the computer are all ways of technologizing the word.” (1982, 80). In similar terms, and from a pragmatic perspective, infographics and data visualisation may be conceived of as ways of technologizing, or instrumentalizing numbers. Best practice literature that expressly conceives of data visualisations as a tool, would therefore represent an optimal starting point (Cairo 2012, 2016).

The approach I set out here encompasses an aesthetics and a rhetoric of data visualisation. Taking after W.J.T. Mitchell’s imagetext, I conceive of data visualisations (comprising shapes, designs, labelling, and accompanying text, and other “page furniture,” whether static or interactive) as datatexts. In this conception, words and images (regardless of medium) both supplement and support each other, contributing towards the meanings communicated, in a multilevel discourse whose site of emphasis is situated within the pursuit of pragmatic outcomes.

The Datatext and Embodied Metaphor

Just as all infographics serve as visual metaphors (Wainer 2006, 30), so too do datatexts. Here I use the term metaphor in a relatively broad sense – as the application of figurative language in any form. This is consonant with earlier, broad definitions of visual metaphors within the arts, concerned both with sites of production, and reception (Aldrich 1968, 1971). This approach is in turn predicated on the empirical reality that figurative discourse (whether as images or words) seem to work in psychologically similar ways (Danesi 1995, 266). I therefore start with three fundamental assumptions in relation to the metaphors contained within datatexts, namely that:

1. Some are abstract and implicit (for example, line graphs, where up is more, and down is less, according to cognitive metaphor theory),

2. Some are representational, and explicit (for example, pictograms), and;

3. Some contain both.

Metaphors are more than simply ways of talking, they help us to situate ourselves within, and come to understand reality, too (Lakoff and Johnson 1980). Certain metaphors in data visualisations seem to take the form of gestalt-like image schemas, that are grounded in universal human experiences of being in the world, informed by our interaction and engagement with space, and with objects in the space around us. In this conceptualisation, Lakoff’s Spatialization of Form hypothesis explains how image-schemas emerge out of the process of metaphorically mapping experiences of physical reality into mental space, and comprehension. In turn, particular conceptual structures are associated with certain image schemas, which John S Rich has set out and mapped to different graphical forms, as follows:

Categories take the form of CONTAINMENT image schemas (and so may ideally be expressed in the form of, for example, Scatter plots, Pie charts, Venn diagrams, Stacked bar graphs, and Histograms).

Hierarchies take the form of PART-WHOLE and UP–DOWN image schemas (and so may ideally be expressed in the form of, for example, Tree diagrams, Ring charts and Sunburst diagrams).

Relationships take the form of LINK image schemas (and so should ideally be expressed in Matrix charts, Node-link diagrams, Word clouds and Alluvial diagrams)

Radial connections take the form of CENTRE-PERIPHERY image schemas (and so may ideally be expressed in the form of, for example, Flow maps, Density maps, Cartograms, and Heat maps).

Foreground-background connections take the form of FRONT-BACK image schemas.

Linear scales take the form of UP–DOWN and LINEAR ORDER image schemas (and so may ideally be expressed in the form of, for example, Scatter plots, Polar area diagrams, Time series sequences, Timelines, and Line graphs) (Risch 2008, 7–8).

Representational Metaphor and Affective Reasoning in Design

Studies in the field of affect indicate that emotion and identity are key components in online engagement, and in sharing (Papacharissi 2015), a finding that may have some bearing on biocommunicability, given the assumption that citizens are actively engaged in creating and sharing knowledge about public health and risk online. A datatext that engages emotion and identity may therefore be shared more freely within the wider architecture of social media, taking advantage of affective connectivities in social networks. These networks may in turn mediate affective relationships people have online, contributing to the “soft structures of feeling that may potentially sustain and mediate the feeling of democracy” (Papacharissi 2015, 32). For a datatext to have optimal biocommunicability therefore, it must present metaphor in a way that engages different publics’ affective reasoning.

But what purpose do metaphors serve in data visualisation? In his “ritual” approach to communication as culture, James Carey sets out a seeming dualism in metaphor, that may inform our understanding of the cultural significance and resonance of representational (and embodied) metaphors:

A blueprint of a house in one mode is a representation “for” reality: under its guidance and control a reality, a house, is produced that expresses the relations contained in reduced and simplified form in the blueprint. There is a second use of a blueprint, however. If someone asks for a description of a particular house, one can simply point to a blueprint and say, “That's the house.” Here the blueprint stands as a representation or symbol of reality: it expresses or represents in an alternative medium a synoptic formulation of the nature of a particular reality. While these are merely two sides of the same coin, they point to the dual capacity of symbolic forms: as “symbols of” they present reality; as “symbols for” they create the very reality they present. (Carey 2008, 23).

In substituting “a blueprint” for “a datatext,” and “a house” for “empirical data,” a difference in the kinds of meanings and knowledge that may be expressed as data in graphical form, becomes apparent. The datatext stands both *as* and *for* data that empirically expresses natural phenomena. On one hand, abstract (and universal) metaphors, such as up is more, and down is less, tend to present reality “as’ is; whereas representative (and locally resonant) visual metaphors contribute towards the normative construction of the realities they embody. The former instrumentalizes a particular way of knowing; the latter helps to shape what is known. In its embrace of bodily and representational metaphors, the datatext may be thought of as looking back to an ideal of “truth to nature” in the visual representation of data, to a way of knowing that would eventually be superseded by scientific objectivity (Daston and Galison 2007). In truth, of course, datatexts may (and often do) embody both modes; “as” and “for.” Where they do, they may become an ideal vehicle for conveying important public information in a compelling way, during a crisis.

The Datatext in the Data

The theorisation I set out here is informed in part by empirical findings from Borkin et al. (2013). The consequences of these findings for best practice in infographic design, are here summarised concisely by John Wihbey:

1. Visualisations that are memorable “at-a-glance” are often the most memorable even after longer viewings — i.e. there is something instant, enduring and intrinsically powerful about memorable graphics

2. Titles and text are key elements in a visualisation and help recall the message

3. Human recognisable objects (e.g. pictograms) help with the recognition or recall of a visualisation

4. Redundancy helps with visualisation recall and understanding. (Wihbey 2015)

These findings speak both to the limitations of the logico-semiotic theorisation of data visualisation, and to the potential of a multilevel discursive approach, namely:

1. Infographics should be simple to be memorable, but they need not necessarily be multivariate, and hence need not conform to Tufte’s “Principle of Graphical Excellence.”

2. If both text and image in infographics are factors in recall, this suggests that they are interdependent variables in the communicative message (much as they are conceived of in Mitchell’s imagetext, and hence in the datatext).

3. If human recognisable objects aid recall, then representational metaphors are important, regardless of aesthetic considerations.

4. If memorable charts include additional repeated information, then the centring of efficiency in design (as expressed in Tufte’s data-ink ratio) may be less efficacious than concepts like impact, or engagement.

Borkin et al. (2013) illustrate their findings, with a series of three galleries of visualisations, presented left-to-right (Borkin et al. 2013, 2306). These galleries show (first) the 12 most memorable visualisations (right-to-left, top to bottom), followed by the 12 most memorable visualisations that do not feature human-recognisable iconography, and then lastly, the 12 least memorable visualisations in their experiment. The most memorable visualisations (the 12 on the left) tend to blend rich (and vivid) representational metaphors with an adherence to spacialisation of form conventions (such as UP–DOWN schemas, or LINK schemas). On the other hand, those least memorable (the 12 on the right) are visually inconspicuous, are devoid of representational metaphor, and in some cases (particularly those that involve plotting data on negative axes) appear to subvert image-schematic conventions (particularly UP–DOWN schemas). Infographics that subvert these image schema conventions (such as up is more, down is less) may therefore be thought of as being unintuitive to read, and hence less memorable.

It is important to note that there are certain limitations in using memorability as a proxy for effectiveness, in this approach. Memorability is, of course, no proxy for comprehensibility, and so chart types that do not map clearly to image-schematic conventions should not be abandoned entirely. Nonetheless, as Borkin et al. (2013) observe, using memorability as a proxy for effectiveness may serve as a useful steppingstone towards future, more sophisticated questions about infographic design, particularly concerning the nature of, and variable levels of, engagement.

There are, inevitably, additional critical considerations to take into account when designing data visualisations. For example, if they are to have an optimal impact at the local level, they must take advantage of locally organised structures of knowledge, where appropriate, so that important context is not lost, or distorted (Merry 2021). Similarly, metaphors and other models of knowing can constrain approaches in the assembly, production and opperationalization of policy as it arises from big data (Halpern 2015). Nonetheless, where a working practice that embraces local knowledge is maintained, this need not represent a fundamental obstacle. Where infographics are conceived of as “quantum media” (Wernimont 2019), additional concerns may arise concerning their capacity to directly impact peoples’ lives, rather than being merely the product of passive observance. Such concerns must, of course, be taken seriously.

I will now conclude, presenting a definition, and some suggestions for how to operationalise the datatext in future empirical studies, before setting out further theoretical directions that might help to refine the concept of the datatext, towards optimising its biomcommunicability, in future studies.

Conclusion

Building upon previous calls in the public health literature (Featherstone 2014), I have attempted to set out here a theoretically coherent approach towards justifying the use of visually embellished infographics as datatexts that may contribute towards optimising biocommunicability, a particularly pressing concern for pandemic publics. Matters of style in data visualisation must move beyond the stigma associated with highly visual forms, particularly where they are concerned with public health campaigns. Some infographic forms may seem patronising, or childish, or tasteless to whomever the arbiters of these things may be. And yet, even assuming that this is true, what need a temporary sense of indignity matter, if a life-critical message is more effectively conveyed? Some may find it easy, during a period defined by information disorder (Wardle and Derakhshan 2018), to dismiss persuasive infographics as little more than propaganda in the culture wars. But this ignores the fundamental difference between propaganda and persuasion in the public sphere (Jowett and O’Donnell 2014 [1986]). When developing communications and literature for crisis campaigns, public bodies should be aware that certain publics may respond differently to certain chart forms, and that for some publics, particular graphical forms may be more effective than others, not necessarily due to literacy, but perhaps due to affective issues (Dick 2014). This decision-making process in turn requires a sound empirical basis. It is necessary to move beyond hunches, and towards reliable and replicable methods for designing the most effective formats towards achieving optimal public health outcomes.

The Datatext: A Definition

The datatext is a visual argument that takes the form of images and words, that combine to express numerical data with the aims of optimising memorability, and affective connectivity, while at the same time adhering to standards of best practice in terms of accuracy. In public health communications, the datatext may achieve optimal biocommunicability where it adheres to embodied metaphorical conventions, along with culturally resonant representational metaphors and design components (including suitable colours, words and other design features). Datatext design takes into account the social identity and group dynamics within the publics they are designed for. The datatext should be clear and vivid, and should contain reliable information, displayed accurately.

Applying the Datatext in Biocommunication

For communicating information relating to fixedness (within space), small multiples of pictographic datatexts, rather than line graphs, may be used to indicate rates of infection, and death in a given locality, over a particular period of time. Backgrounds to these charts may feature recognisable fixed local icons and landmarks (such as statues, buildings, stadia, and natural landmarks). These may be presented in colour, in greyscale or in outline, depending in part upon the potential for such images to clash with the pictograms used, and in part upon the potential for the effectiveness of such images, based on the metaphor’s level or richness (from connection, through similarity, to opposition, low to high), weighted against the metaphor’s visual complexity (from juxtaposition, to fusion, to replacement, low to high) (Phillips and McQuarrie 2004). Alternatively (or perhaps additionally), given that pictures of people tend to significantly affect the memorability of photography (Isola et al. 2011), it may be helpful to use images of recognisable local or regional figures to accompany such designs. In turn these datatexts may be published in news media or social media consumed by, or in public spaces routinely accessed by low-numerate, or politically disengaged publics.

Alternatively, for communicating comparative information relating to movement, (through space), such as for example the transmission of infection across communities, circular ribbon charts within (or superimposed over) dynamic local transport network maps, or locally recognisable bridges, may serve as good metaphors for expressing the dynamics of connectedness, conforming as they do to the gestalt-type LINK schema.

Some problems in the communication of the pandemic arose at sites of reception, amongst publics unfamiliar with certain data formats. Later in the first lockdown, government and media outlets opted to express case load growth rates using logarithmic, rather than linear vertical scales in their charts, flattening the dramatic nature of the rise in cases as they did so. This approach to scaling is standard practice in fields such as epidemiology, where the identification of shifts in the growth rate of data are the focus of analysis. However, the wider public are not epidemiologists, and many seemed to struggle to comprehend data presented using this scaling, a situation with clear public policy consequences (Romano et al. 2020).

Other problems arose at sites of production. Relatively early into the UK's first national lockdown, in April 2020, Public Health England started publishing a weekly surveillance report for England, in a highly visualised format. These scrolling infographics were used to convey a range of key metrics; from regional variations in incidences and deaths (both absolute and relative); to national cases (and deaths) by age range and sex; to rates of hospitalisation. However, though these charts were clearly designed for sharing on social media, the graphical choices, and the design principles employed (broadly objective, impersonal, and abstract) were sub-optimal. For example, to achieve a more memorable effect for a wider public, vertical bars (comparative or stacked) could have been used to compare male and female cases and deaths, by age range, rather than the horizontal mode of comparison of the sexes used, which appears to confound the LINEAR ORDER image schema (Public Health England 2020).

Future Theoretical Directions

Drawing upon the theoretical domains of media effect (and in particular selective exposure and selective retention) (Klapper 1960); and on affect theory (Tomkins 1984), future empirical research may help to develop replicable datatext forms, templates and exemplars, drawing upon the field of cognitive heuristics (Jones 2015), in order to accommodate the affective reasoning of the audience in tailored, targeted campaigns (Kreuter and Wray 2003). Such research may experimentally test the suitability of certain combinations of words, images and metaphors, toward establishing a datatext visual grammar (or gestalt) for crisis communications. The potential in phatic visual language, or graphic phaticity (Hazaparu 2015) that may be used to set the relational bounds between a medium (for example, a highly visual public health datatext) and its intended public, may also be worth exploring. The ways in which suitable images are selected for use in datatexts may be informed by future empirical research into the cultural determinants of different publics, whether, for example, they adhere to high or low context communication (Hall 1976). Similarly, findings from empirical studies concerning the uses and gratifications of datatexts may be informed by Fishbein and Cappella’s (2006) intervention variables in the field of health communications, concerning the form, function, and extent of media effects.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Murray Dick

Murray Dick lectures in Multimedia Journalism in the department of Media, Culture, Heritage, Newcastle University, UK. His current research is concerned with cultures of data visualization, journalism historiography, and the public sphere. Email: [email protected]