Agency in Action

Abstract

In 2014, Rawlins and Wilson proposed a typology of agential interactions between users and designers of interactive data displays. This article tests that typology by studying 20 users working with three different types of interactive data displays and answering questions, which were coded by verb and actor and analyzed for themes. The authors show that rhetorical agency is marked by thoughts, actions, and language. Affordances by the designer open a shared rhetorical space where user and designer are coparticipants. As interactivity increases, participants see themselves as rhetorical agents in a community of rhetorical agents rather than as conduits of information.

Keywords

agency interactivity Internet graphics rhetorical agency

Introduction

Rawlins and Wilson (2014) explored the agential relationships between the designers and users of interactive data displays (IDDs). This article extends the conversation on IDDs and agency by observing how users act as agents in a controlled setting. Using the different levels of IDDs from the Rawlins and Wilson (2014) typology, we presented users with tasks to accomplish for a fictional client using a specific IDD. We then captured user behaviors and attitudes through usability methods and interview protocols in an attempt to gain better understanding of online rhetorical agency. Our small study (20 participants over 3 experimental conditions) focused on one central research question, which has two subparts for which we collected data:

Do different types of IDDs elicit different behaviors and user experiences? Can we say that these differences represent markers of agency?

a. How do users understand and discuss the control they (and the IDD designers) have over the rhetorical situation?

b. How do users understand the usefulness of the IDDs and their affordances to make content useful to the audiences?

Through this study, we hoped to make behavioral traces of agency visible in real-world applications. We found that rhetorical agency was marked in our study by different thoughts, actions, and language in user responses. These different reactions by the users contained markers of agency that provide insights into how agency functions in practical settings. In particular, we found that affordances by the designer opened up a shared rhetorical space where the user and designer could become coparticipants in solving problems, answering questions about the data, and making decisions about how to influence others. In addition, the study showed that as the interactivity of the IDD increased, participants began to see themselves more as rhetorical agents in a community of rhetorical agents rather than as simple conduits of information between the designer and the client. Our findings provide practical application and additional depth to academic conversations about rhetorical agency and to discussions of the implications of interactivity in data displays.

Background

Modern Internet graphics are, for the most part, built on the idea that more interactivity equals greater user engagement and interest in both the graphic and the Web page on which the graphic resides. The interactive elements of the graphics allow for more user input and encourage users to spend time understanding, digesting, interpreting, and creating many different kinds of data (Wojdynski, 2015). Mirel (1998) reported that “visualization experts agree that the more directly users manipulate displays, the more they are able to solve complex problems” (p. 500). As the desire to encourage greater user engagement with Internet sites has increased and the technology for creating graphics has improved, the number and variety of IDDs have exploded in the past two decades. Now these IDDs are an expected part of the online experience, whether in simple online polls, news graphics, or complex big data visualizations.

While IDDs have become commonplace for Internet users, they still represent a largely unexplored territory for researchers. Adding interactivity to data displays raises important questions of agency, credibility, usability, and effectiveness and how the creators of IDDs use interactivity to influence users' beliefs and actions. While there are well-established concepts in technical communication for determining whether graphics are efficient and clear (Tufte, 1990) or ethical (Barton & Barton, 1993; Kienzler, 1997), Rawlins and Wilson (2014) asked whether we need other concepts to understand interactivity. Online interfaces allow users to create their own versions of graphics, with user data intermingled with larger data sets to form unique visualizations. While it seems that these newly created graphics would be the intellectual and rhetorical product of the user, Rawlins and Wilson propose that users and designers inhabit a “shared rhetorical space” (p. 315) where agency is an important additional element, particularly in relation to:

what data can be considered or incorporated into a graphic,

what design features can be manipulated or incorporated in the graphic, and

what the final rhetorical message of the graphic will be (p. 309).

Through these questions, the designer and the user coparticipate in the process of invention, arrangement, and delivery in a shared rhetorical space. While the user has the illusion of control over the graphic, the actual opportunities for agential user action are limited and controlled by the design and functionality of the IDD. The “extent of control” (Mirel, 1998, p. 501) that designers afford their users creates opportunities for users to become both problem-solvers and cocreators of the data and the data visualizations. Withagen, de Poel, Araújo, and Pepping (2012) argue that affordances by the creators of information, data, or displays invite agency on the part of the users or consumers.

Rawlins and Wilson (2014) applied these ideas of control, affordances, and agency to modern graphics by proposing a typology of IDDs that includes five broad categories (static, zoom-and-pan, constrained, shared, and unconstrained) that can be arranged on a continuum where designers have controlling agency on one end and users have controlling agency on the other end. In between those extremes, designers and users share differing degrees of agency. In the static category, designers control the data, design, and rhetorical message of the IDD. As we move across the continuum, however, the user gains control over movement in the display (zoom-and-pan), a limited part of the data and some of the design features (constrained), and then much of the data and the design, but not necessarily the rhetorical message (shared). At the far right of the continuum, the user controls the data, design, and rhetorical message of the IDD, while the designer simply provides interactive tools for the user. This typological framework for understanding designer/user agency in IDDs provides an important tool for studying and critiquing the rhetorical effect of online graphics.

Discussions of agency often involve complex and abstract critical theory, but the Rawlins and Wilson typology suggests strategies for recognizing agential behavior in specific settings. While Herndl and Licona (2007) describe agency as “a potential for action” (p. 141), Rawlins and Wilson (2014) found that in IDDs, the “potential for action” represents

cooperation, collaboration, co-creation, or even co-authorship between the designer and the user. The user, who is an interpretive agent in static displays, becomes a creative agent who actively makes decisions and participates in the creation of the data, design, and rhetorical message of the IDD. (p. 308)

Therefore, IDDs provide a specific, practical opportunity to observe how individuals “use available means of persuasion to effect change in [their] spheres of influence” (Wolford, 2016). Our study provides technical communicators and technical communication scholars with the beginnings of a more accessible and practical exploration of agency.

Methods

Study Overview

In this institutional review board-approved study (IRB approval #504977), we used the definitions set out in Rawlins and Wilson (2014) to identify a “static graphic” (p. 309), a “zoom-and-pan” IDD (p. 310), and a “constrained-agency” IDD (p. 312) that were constructed to help the public understand and use information from a 2014 United States Department of Agriculture (USDA) report titled “Expenditures on Children by Families, 2013.” We chose these graphics because they represent real data being used in different ways by different governmental and private entities. The data, along with the graphics that illustrate the raw data, show how the United States government is inventing its relationships with its citizens by defining a range of choices related to family planning. As such, these graphics become excellent examples for examining how IDDs define and limit the opportunities for agential action by the user.

All three data representations conveyed the same data from the report—the cost of raising a child in different parts of the country—but each graphic required the user to interact in different ways to get the information (see Figures 1 to 4). We treated each of the graphics as a separate experimental condition.

Figure 1.

The static graphic provides a simplified set of information that summarizes cost data from the USDA report and provides no affordances for users to interact or contribute their own unique data to shape the final graphic. The initial graphic is the final graphic.

Figure 2.

This graphic, from http://www.babycenter.com/cost-of-raising-child-calculator, illustrates zoom-and-pan functionality. There is a slider bar at the bottom (not shown) that permits users to change the age of the child in question to produce a different pie chart. The text box shown in the right panel is the type of contextual information that pops up if users click on one of the pie chart slices. There are limited affordances for user input with zoom-and-pan IDDs; mostly, users are able to manipulate and reveal shallow information already embedded in the graphic. IDDs = interactive data displays.

Figure 3.

The user first encounters the input screen on the left that generates the bar chart and tabular information on the right. Rawlins and Wilson (2014) describe this opportunity for user input of unique information as providing more agency than static graphics or zoom-and-pan IDDs. USDA = United States Department of Agriculture; IDDs = interactive data displays.

Figure 4.

This is the same IDD as the zoom-and-pan shown in Figure 2 but with an additional input screen enabled for the participants so they are contributing more of their own information to the final graphic. IDD = interactive data display.

For this small exploratory study, we assigned each participant an experimental condition. Two researchers were present for all test sessions. One facilitated the test by interacting with the user to explain the study procedures and answer questions. The second researcher observed the test and performed the posttest interview. Each participant received a prompt asking them to use the assigned data representation to prepare to advise a client. Before the task, we collected demographic data. During the task, we used Morae usability software to record user audio feedback and screen capture of the user completing her or his task. After completing the task, each participant was presented with a survey of Likert scale questions and open-ended interview questions administered by one of the researchers. At the conclusion of the study, we transcribed the interviews. The usability and quantitative findings we gathered will be the subject of another article. Here, we present an analysis of the posttest interview responses that pertained to our research questions on control of the rhetorical situation and usefulness of the IDDs and their content.

Stephens, DeLorme, and Hagen (2015) propose user testing as a method for “codefining the parameters of the interaction space or the space of shared agency” (p. 334) with interactive tools. We explored how agency could be productively studied with usability methods. Usability methods are designed to observe representative users performing representative tasks (Barnum, 2011; Nielsen, 1993). This means, ideally, that the users are not necessarily subject-matter experts and that they approach the task with different goals (Albers, 2004). Usability methods take into account the various goals of the users as they test how well the information products help the users in meeting their goals (Albers, 2011; Shearer, 2011). In this study, we pivot from testing how well users meet their goals to defining how the interactivity opens a space for agential action. While the potential for agential action is difficult to measure quantitatively, we hoped that the participants' experiences with the different levels of interactivity in the IDDs would create qualitative differences in their answers to interview questions.

Participants

Because we were using data on family costs and asking participants to imagine themselves in a role advising a family, our original plan was to recruit majors from the department of Human Development and Family Studies in a large Southwestern university. These students are generally numerous in the undergraduate technical communication service course. When a direct recruitment appeal to the department yielded no participants, we expanded the recruitment efforts to include other students in these technical communication courses. We recruited 20 volunteer participants for the study, including 13 females and 7 males between the ages of 18 and 26. We capped the initial study at 20 participants to test the methods and prepare for a larger future study. Each of the participants was a full-time student with no children of his or her own. The participants represented a wide range of majors, including human development and family studies; community, family, and addiction services; soil science; speech pathology; computer science; biology; geophysics; hospitality; business; public relations; advertising; communications; and sociology. Each participant was compensated with a $5 coffee shop gift card for attending the study session.

Data Collection

We scheduled participants at 30-minute intervals in a university usability lab. Participants were first asked to sign a consent form and then to complete a pretask questionnaire about demographics and familiarity with graphics software. Following the pretask consent and questionnaire, the participants were seated in front of a Mac Pro desktop running on Windows XP and using Morae software. They were presented with either an onscreen PDF of the static graphic (Figure 1), an Internet browser open to the zoom-and-pan IDD (Figure 2), or one of two constrained-agency IDDs (Figures 3 and 4). At the beginning of the study, we were using just one constrained-agency IDD and were expecting to have 15 participants. After presenting preliminary findings at the 2015 Association for Teachers of Technical Writing (ATTW) Conference, we added five other participant sessions with a second constrained-agency IDD (see Figure 4) that had a graphic interface more consistent with the graphic look and feel of the zoom-and-pan IDD. This second constrained-agency IDD was the same zoom-and-pan IDD but with an additional input screen enabled for the participants. During data collection, the experimental conditions were rotated so that data from all conditions were collected during the early, middle, and late phases of the study. The data collected from the second constrained condition did not differ noticeably from the first constrained condition, so we combine data from those two conditions when we display or discuss our findings.

Once they were in front of the computer, the participants were given a prompt (see Appendix A for full prompt) that cast them in the role of a financial consultant working with parents. In part, the prompt reads:

You are a family financial consultant who works with parents to help them plan for raising their kids. Your client is a single parent who earns $50,000 per year, has two children ages 5 and 9, and hopes to send them both to college. Your client lives in the Southern US but has an opportunity to relocate to the Midwest, earning the same income. Your client has come to you hoping to better understand the costs involved in raising these two kids.

Participants were asked to respond to the prompt by using the static graphic or IDD to make a plan to advise their client. During the task, we recorded the audio and screen capture of their task interactions. After the test, we analyzed dwell time, number of inputs, and time on task. When the participants completed the task, they were given an online Likert scale survey to gauge their experience using the graphic. At the completion of the survey, we conducted an interview using a protocol of open-ended questions (see Appendix B). The audio of these interviews was recorded using a smartphone application. While the participants were not necessarily experts in family finances, or students in a related area, we asked them to imagine acting as a consultant for a client family so that they were using the static graphic or IDD to make an argument to an audience. If we had asked them to simply consider the graphic for their own purposes, the rhetorical aspects of experimental conditions would have been difficult to tease out. And, as discussed in the study overview, expertise with the task is not always assumed with usability studies (Albers, 2004).

At the conclusion of the study, we transcribed the audio recordings of the interviews using an online transcription service. There was a subset of questions that asked specifically about terms and concepts related to our research questions (e.g., agency, control, and usefulness), and those specific questions are discussed in this article.

Analysis and Coding

The three researchers each initially reviewed three of the transcripts (one from each condition) to identify where and how participants spoke about our three concepts of interest (e.g., agency, control, and usefulness). After coming to agreement as to what constituted participant language about the three concepts, two of the researchers then separately reviewed the remaining 15 transcripts. For technical reasons, recordings for Subjects 12 (zoom-and-pan) and 13 (constrained) were lost, and transcriptions were not possible. These two researchers identified the passages where participants discussed our key concepts in the existing transcripts.

Questions 2, 7, and 10 on the open-ended interview protocol asked specifically about terms and concepts related to our research questions (i.e., usefulness, control, and agency). Because we were interested in how participants experienced and discussed agency, we had used our terms of interest directly in the questions. Unsurprisingly, participants discussed our concepts of interest almost exclusively in response to these questions.

Q2. What was the most useful information in preparing your answer for your client?

Q7. As you were interacting with the graphic, how much were you able to control the information and the experience?

Q10. Do you think the person who designed this tool/interface had an impact on the advice you gave?

The two researchers collected all passages discussing each of the three concepts and looked for themes and patterns in how participants discussed their experiences depending on the experimental condition.

In addition, the researchers reviewed the complete answers for all 18 transcripts in response to Questions 2, 7, and 10 and coded the verbs used in those answers as either abstract (e.g., think, feel, understand) or concrete and observable (e.g., make, pick, help). Researchers also coded for which agent is associated with each verb (i.e., user, client, graphic, or designer). We had added Question 10 to the interview protocol after feedback from colleagues at the 2015 ATTW conference and after 9 of the 20 interviews had been conducted. Because only 11 of the 20 participants answered Question 10 and we lost data for two of those participants, we do not discuss values for the verb coding for Question 10 here. The data summaries and average verb counts for Question 10 were not robust enough to make comparative statements with findings for Questions 2 and 7. We do include passages from Question 10 answers in our qualitative analysis of themes because they provide information and nuances that help explain participant attitudes.

In the broader collection of themes and patterns, we focus on a qualitative discussion of usefulness, control, and agency that shows key differences in how the participants perceive their roles in using the IDDs. Coding the verbs for Questions 2 and 7 answers, on the other hand, provides us with some evidence of the types of actions discussed by the participants and their assignation of agency to both human and nonhuman actors. Taken together, the intersection of these two analysis schemes provides a window into how rhetorical agency looks in action during practical tasks.

Findings and Discussion

In this section, we discuss the agency-related themes and patterns identified in the transcript and the verb coding. We specifically discuss data that align with the subparts of our research question: (a) How do users understand the control they (and the IDD designers) have over the IDDs? and (b) How do users understand the usefulness of the IDDs and their affordances to make content useful to the clients? In the subsequent discussion section, we connect the findings back to our larger thematic research questions: Do different types of IDDs elicit different behaviors and user experiences, and can we say that these differences represent markers of agency?

We conducted open-ended interviews with participants following the online Likert questionnaire. The interview protocol can be found in Appendix B. We wanted to capture information about user experiences with the IDDs and the ways in which the users discussed those experiences. In this article, we are focusing on our qualitative data and analysis, particularly on the users' discussion of the tasks. The Likert questionnaire and usability data gathered will form the basis for a separate article.

In our analysis of the participants' answers to Questions 2, 7, and 10, we coded each verb in those responses as either involving abstract or concrete action. We also coded the actor for each verb as user (the participant), graphic (the IDD itself or information it contained), client (the fictional client for whom the participant was preparing information for), designer (the person who designed the IDD), or unknown (for undiscernible actors or actors outside these primary categories). Table 1 shows the average number of verbs per participant answer broken down by question and experimental condition, and the ratio of abstract-to-concrete verbs for each experimental condition by question. Table 2 shows the percentages of verbs attributed to each type of actor by question and experimental condition.

Table 1.

The Average Number of Verbs Per Participant Answer.

		Total verbs	Abstract	Concrete	Abstract– concrete ratio
Question 2	Static	20.60	14.00	6.60	2.12
	Zoom-and-pan	7.38	5.38	2.00	2.69
	Constrained	12.22	6.78	5.44	1.25
Question 7	Static	8.30	6.20	2.10	2.95
	Zoom-and-pan	6.38	3.88	2.50	1.55
	Constrained	8.06	6.78	4.89	1.39

Table 2.

The Percentage of Verbs Attributed to Each Actor by Question and Experimental Condition.

		Average verbs per answer	User %	Graphic %	Client %	Designer %	Unknown %
Question 2	Static	20.60	30	52	17	0	1
	Zoom-and-pan	7.38	32	41	14	0	14
	Constrained	12.22	59	28	10	1	2
Question 7	Static	8.30	87	10	4	0	0
	Zoom-and-pan	6.38	67	20	6	0	8
	Constrained	11.67	50	23	18	4	4

This verb coding is aimed at answering our overarching research question about the ways IDDs might elicit different behaviors and user experiences, revealing potential markers of agency. We are less interested here in what was said in response to the questions and more interested in how the participants' discourse reflects the agential space they operated in while completing the task with the assigned IDD. Although we are not aiming in this exploratory study (with only 20 participants) to generate numbers that meet standards of statistical significance, we argue that there are suggestive and informative patterns in the verb counts in Table 1 that we can use to begin our discussion of participant agency. Future researchers may want to revisit our approach to making agential behaviors visible with more robust numbers.

We can see for Questions 2 and 7 that the number of verbs varies by experimental condition, which we expect to see if Rawlins and Wilson (2014) are right that these different types of IDD represent different agential relationships. We coded for abstract (e.g., think, feel, understand) versus concrete and observable (e.g., make, pick, help) verbs, thinking the former would be found more often when participants were focused on passive and internal agency with a narrow scope of interactivity. We expected that the latter type of verbs would be found more often when participants focused on active and external agency with a broader scope of interactivity. The ratio of active to concrete verbs in the fourth column shows that for the static condition, abstract verbs were used 2 or 3 to 1 for Questions 2 and 7. For the constrained condition, this ratio is closer to 1 to 1 for both questions. The results for the zoom-and-pan condition are uneven—closer to the static ratio for Question 2 and closer to the constrained ratio for Question 7. That may be due to a bad choice for our zoom-and-pan example or it may be that as a middle step between static where Rawlins and Wilson expect low participant agency and constrained where they predict higher participant agency, zoom-and-pan may present participants with options they can take or leave, skewing markers of behavior in different directions.

But the closer ratio of abstract-to-concrete verbs in the constrained condition may indicate participants talking about (and perhaps experiencing) agency in more active ways.

Table 2 shows how participants for each question and in each experimental condition attributed action across different types of actors, which could be seen as one of the markers of agency. For Question 2, which specifically asks about the usefulness of the information, there is a strong attribution of action to the graphic for each condition. But in the constrained condition, the attribution of action to the user is highest by nearly 30 percentage points. The constrained condition affords more interactivity for the user, so this higher number may signal a link between higher interactivity and the user discussing their own action more. For Question 7, attribution of user action drops as affordances and interactivity increase with the zoom-and-pan and constrained conditions. Likewise, as participants answer this direct question on agency, they attribute action more widely across the possible actors.

Table 3 shows examples of coding data for each of the three experimental conditions. The left column shows the details of the example response; the middle column shows the quoted answer with verbs we coded as abstract italicized and those we coded as concrete bolded. The right column shows how we coded attribution of action (agency) in the participants' responses. Because we are coding verbs in transcriptions of natural language, often we group the verb with surrounding language to help us determine whether the verb was abstract or concrete and to whom the agency is attributed.

Table 3.

Examples of Coding Data.

Details	Answer coding (abstract, concrete)	Attribution (parentheses denote implied actor)
Participant 11 Static condition Question 2	I like this part the most. Of course just because it gives you numbers. Like it gives you good numbers to look at and so you're just like “Okay this is where this is; and this is where she would fall if” – just having numbers makes it a lot easier ‘cause it specifies more. So it helps her really narrow stuff down. [8/5]	User: I Graphic: it, it, this, this, this, (just having), it, it, it Client: she, her Designer: Unknown:
Participant 1 Static condition Question 7	I felt pretty in control. I mean, everything you need is pretty much on here. So as long as you pay attention to what you're reading, I feel (garbled) could be really helpful for someone who's planning to move with children, and keeping the future in mind. [8/1]	User: I, I, you, pay, you, you, I Graphic: everything, (garbled) Client: someone Designer: Unknown:
Participant 18 Zoom-and-pan condition Question 2	I think the total cost, the big picture I guess. What's gonna be the most important thing is total cost, and that'd be me. [5/0]	User: I, I, that Graphic: what, thing Client: Designer: Unknown:
Participant 5 Zoom-and-pan condition Question 7	I feel like I was in control of it. You get to choose your options at the bottom with the little slider bar and then you can go back and forth between the South and Midwest, which is what they need to know. [4/2]	User: I, I, you, you Graphic: (which is) Client: they Designer: Unknown:
Participant 3 Constrained condition Question 2	I think the numbers are really just—it makes you really realize how much everything costs. And so seeing how much you really need to spend on your education, it just makes you really realize how much it costs, I guess. And so the numbers, I'd say. [8/2]	User: I, I, I Graphic: numbers, it, it Client: (seeing), you Designer: Unknown: everything, it^a
Participant 15 Constrained condition Question 7	I thought I was, I mean, for the most part pretty much all in control. I mean, you plug in your information and they give you options of course. But it's your information that you're controlling. So I think they just pretty much help you clarify what housing expenses are, and food expenses, and then the areas of course. [7/4]	User: I, I, I, I, you, you, I Graphic: it Client: Designer: they, they Unknown:

Action is seemingly being attributed to conditions in the world outside of information contained in the graphic.

The six responses in Table 3 are not perfect examples of every trend we observed in the data. The abstract-to-concrete verb ratios shown in brackets at the end of each response are more typical for the static responses than for the constrained responses. As reported from Table 1, we saw much higher ratio of abstract-to-concrete verbs in the static responses (as shown here) and more even ratios for the constrained responses where the IDD offers more interactivity. These examples do align closely with the findings from Table 2. Static responses for Question 2 attribute action more to the graphic, and responses for Question 7 attribute action more to the user. For the constrained responses, action is attributed across a wider set of potential agents.

Rawlins and Wilson (2014) devise their typology to describe the ways users and designers “share agency in the creation of the graphic” (p. 305). When we put people in front of three versions of IDDs from their typology, we see differences in verb use that might be attributable to the different ways agency is shared in these three IDDs. The different affordances for interacting with the IDD (i.e., different degrees to which participants can enter and manipulate data/information) might be directly eliciting the differences in verb use we see. If so, the increased prominence of concrete verbs shown in Table 1 may reflect participants talking through and verbalizing their agency as they discuss completing the task with the assigned IDDs. Likewise, the broader attribution of action across more actors seen in response to Question 7 might be a verbal sign of increased interactivity (i.e., the user isn't just thinking about doing more herself but operating in a space of shared agency with more actors). By examining each of these discursive behaviors, we are trying to connect with the way Wolford defines agency as how individuals “use available means of persuasion to effect change in [their] spheres of influence” (Wolford, 2016) . We are looking for markers that show what discursively enacted agency looks like.

Patterns in Answers to Question 2 on Usefulness

We asked specifically about usefulness in Question 2: “What was the most useful information in preparing your answer for your client?”). Analysis of these answers provides us with insights into how much the participants valued the information and, thus, how they would use it to interact with a third-party client.

The idea of how useful a graphic is connected with the value of the information and the potential applications of the graphics in other contexts opens a window into user agency. By recognizing the usefulness (or lack of usefulness) of graphics and discussing plans to use the graphics with clients, the participants were stepping into the agential roles determined in part by the designers of the graphics. The extent to which the participants could imagine themselves as agents or more active users of the graphics depended in large part on the level of interactivity in the graphic. In other words, the user's ability to inhabit an agential role in both the current task and in future tasks was determined by the interactivity built into the graphic.

We found that more interactivity coincides with a shift toward participants talking about themselves as actors rather than the graphics. In the static category, the graphic was the most prominent actor, but the graphic and the user were balanced in the zoom-and-pan category, and the user was the most prominent actor in the constrained category. More interactivity encouraged a greater discussion of the participants' own roles in using the information.

In our qualitative analysis, there was a shift in the way participants' answered to Question 2. Most of the users expressed that the graphics were useful in preparing answers for their clients. With each experimental condition, however, the answers focused on different aspects of usefulness, ranging from a focus on the data (static), to a focus on the tool itself (zoom-and-pan), to a focus on individual opportunities for action (constrained).

For the participants who viewed the static IDDs, the information or data were identified as most useful. For example, Participant 1 answered

I mean, it goes over the cost in each section so you can kinda pinpoint where they are and where they're going and let them know that the cost is going to be a little bit higher where they're moving.

Most of the participants who viewed the static IDDs had similar answers to questions about the most useful features of the IDD: “Probably this cost by region” (Participant 2); “The most useful information was definitely cost by region” (Participant 7); “I mean it's clear, but I would just say it's a lot of information and—it's all useful information” (Participant 17).

While it is primarily an issue of trust, the credibility of the data was a key component of the usefulness of the graphics for the users in the static group. In other words, the static information displayed in the graphic was useful because it was considered credible. Several of the participants in the static group specifically mentioned the credibility of the numbers. One participant noted: “I could also see where they're coming from; it's not just numbers that they're making up. Because this looks like it's a very credible source considering that it's from the United States Department of Agriculture” (Participant 11).

While they appreciated the credibility of the numbers, some of the participants in the static group also noted the frustration of not being able to adjust the information: “there's not enough information on here to actually formulate all these things to where you can come up with an exact sum” (Participant 7). Their frustrations seemed to stem mostly from the inability to customize data for their client. They wanted specific answers, and while they trusted the information source, they wanted more ability to serve their client. For example, Participant 17 said, “It's all useful information, but I guess just to give them scenarios, it's just kinda—if you're using it for a particular scenario, it could be overwhelming.”

The participants in the static group focused their attention on the information, its credibility, and the shortcomings of the IDD in customizing data. The participants in the zoom-and-pan group, however, focused more of their attention on discussing the IDD tool itself rather than the data. In many of the interviews, this focus on the IDD as a tool delved into its specific functions, abilities, and designs. Several of the participants focused particular attention on the slider bar as being central to the IDD's usefulness:

Participant 2: Honestly, this kind of threw me off, this thing is, like, not in the middle, you know, and so it was really confusing for me, 'cause it doesn't move along there. …[Interviewer: So the slider margin doesn't line up with ages.] Yeah. …I just feel like it would look better and be, like, more, um, more like, um, like intuitive.

Participant 5: I would say the fact that it gives you, it's broken down into the total expense for the year and it's got it—I really like this part, the age thing. [Interviewer: The slider bar and the—] The totals. …It works pretty easy, and it's listed out pretty nicely. I didn't have to go search and find what I needed to; it's written right there, everything is.

Participant 8: I guess just that it was easy to move this by the age, and then it tells you all the numbers exactly, like with housing.

This focus on the IDD as a tool in the zoom-and-pan group seemed to come at the expense of discussing the data. The participants were far less concerned with the credibility and the usefulness of the data than with what the tool could do to adjust the data. This overall trend continued into the constrained groups, where the participants focused their discussion of usefulness on their own actions and their opportunities to adjust the data and customize the display. The constrained groups took a much more active role in working with the data and focused on their expanded roles in their interviews. One participant discussed what he would do to improve the IDD display and functionality:

Participant 9: I guess my impression would just be that it shows, it compares really well, and I don't know, you can put 'em side by side and tell where things are more or less in certain areas, so I guess that would be useful for them to know. …I would say maybe put them side by side and able to see, comparing 'em side by side and maybe change some things and see how the numbers change with the change that you maybe put in or whatever, but I think the, like I said, I think the tool's really useful. I think maybe just adding a side-by-side comparison would be good.

Another participant even exported the data to Excel to conduct additional calculations:

Participant 6: Yes. I didn't really like how this was laid out simply because I couldn't quickly go back and forth between the results from the Midwest and the South, so I exported both to Excel, and I was trying to compare like that. Ideally, I wanted the two windows side by side, but Excel was only allowing me to open one widow. I printed them out and compared like that.

Ultimately, we found that each of the participants in the constrained groups was more focused on themselves as agents rather than on the data as the agent (static group) or the tool as the agent (zoom-and-pan group). The difference between the levels of interactivity suggests a difference in the participants' inhabitation of agential roles, both in their conceptualization of their roles and in their discussions of their roles after the task.

There were conceptual shifts moving from static to zoom-and-pan to constrained, where the participants were talking about the data, the tool, and then themselves as agents. Most importantly, the more interactivity built into the graphic, the more the participants inhabited their roles as agents. This was also shown in linguistic shifts in the participants' answers to questions. In the static group's answers to questions about usefulness, the most common subjects used in each sentence were “it” (referring to the information) and “the information.” In the zoom-and-pan group's answers, the most common subjects of the sentences referred to the IDD tools as “this,” “it,” and “this thing.” In the constrained groups' answers, however, the most common subject of the sentences was the personal pronoun “I,” indicating strongly that the participants who viewed the constrained IDDs considered themselves as the agent or initiator of the action, while those in the static and zoom-and-pan groups considered the information or the graphic as the primary agents.

Patterns in Answers to Questions 7 and 10 on Agency and Control

We asked specifically about agency and control in interview Questions 7 and 10. Question 7 was in the original interview protocol and was asked of each participant in the study. We added Question 10 starting with Participant 10 based on feedback from a 2015 ATTW presentation. The phrasing of the questions was as follows:

Q7. As you were interacting with the graphic, how much were you able to control the information and the experience?

Q10. Do you think the person who designed this tool/interface had an impact on the advice you gave?

Question 7 was intended to elicit statements on participant's direct experience of agency while completing the task, while Question 10 was intended to elicit statements on a participant's experience of the agency exercised by the designer (as built into the graphic or IDD). Designer and User were the two agents discussed in Rawlins and Wilson (2014).

For Question 7, participants consistently talked about the information as if they were merely passing it along as intermediaries between the static graphic and the hypothetical clients. In one strong example of this, Participant 17 responded, “I guess I personally didn't feel like I was in control of the information 'cause it's someone else's information, and I'm just using it.” Across answers, the respondents talked about reading or interpreting the information for the client but not expressing much ownership of the information nor the process of mediating between the graphic and the client. Participants in the static group overall gave shorter, less engaged answers to our interview questions.

When replying to Question 10, about the impact the designer of the static graphic might have had on the agency enacted by the user, respondents understood that there was a lot of agency in the hands of the designer but still did not express much ownership of the consulting process or the rhetorical choices. For example, Participant 11 responded,

Like I said, the specific numbers that they have for costs, they obviously invested in this a lot. Or invested in this situation and did a lot of research for them. So I feel like, they would definitely make—they definitely helped make an impact for that person to give them all the options that there are.

Similarly, even while talking about herself as the person acting, Participant 17 describes her actions as passing information from the designer to the client, “Yes, because I used all the information. I just took their information and gave it to my client. So yes.”

Despite discussing the choices and adjustment possible in the zoom-and-pan IDD, those respondents to Question 7 likewise recognized they were acting as intermediaries between a designer with substantial expertise and a client. For example, Participant 2 stated,

I just kinda printed it out, and it's just already in the graph form, so it was just given to me and I basically gave it to them. I passed on the information, I felt like, and kind of helped them interpret it, but that's pretty much it.

We would have coded much of the language in this answer as agency self-attributed by the user (i.e., I printed, I gave, I passed, I felt). The user is talking not only about her own involvement in the task but also about the agency of being an intermediary experienced with the IDD.

The users of the constrained graphics offered longer and more complex answers to both Questions 7 and 10. Instead of seeing themselves as intermediaries, their responses show more user engagement with the information, consideration of the topics, information, and usage of the tool more than previously discussed experimental conditions. Here are two representative answers to Question 7:

Participant 9: Control the information? None, because it was calculation and then experience. Is that what you asked? Experience. Experience probably none neither just because I don't have kids, so I don't really see those numbers play out in real life, although I guess, for my parents I could see where they spent and, but other than that, probably not very much on either of those.

Participant 15: I thought I was, I mean, for the most part pretty much all in control. I mean, you plug in your information and they give you options of course. But it's your information that you're controlling. So I think they just pretty much help you clarify what housing expenses are, and food expenses, and then the areas of course.

The participants were thinking about task and information in a more conceptual way. They show engagement (or perhaps agency) in the level of their embeddedness and capacity for self-reflection. Unlike participants working with the static and zoom-and-pan graphics, constrained-graphic users showed rhetorical thinking (in terms of considering their own ethos [Participant 9: “I don't have kids”] or discussing shared strategies for arrangement [Participant 15: “But it's your information that you're controlling. So I think they just pretty much help you clarify what housing expenses are, and food expenses, and then the areas of course.”]). Participants in all three experimental conditions were given an invention tool and a rhetorical task, but in this study, only the constrained-graphic users show rhetorical engagement beyond being an intermediary or conduit in the assigned task.

When constrained participants were asked if the interface designer had an impact on the advice they planned to give, some of the participants indicated a sense of operating (as Rawlins and Wilson, 2014 predicted) in a shared rhetorical space with the interface designer. We can see this in Participant 15's answer to Question 10.

Yes, I think they knew what they were talking about. I think they—I mean, it just helped me provide more information—it would help me provide more information that I didn't have previously. And it shows comparisons, and I really like that. So they kinda do the work for me so I can just put it into simpler terms and then give it to them.

Participant 15 is speaking and thinking in complex ways about the information and how it becomes a persuasive message for the hypothetical client. In the next section, we will discuss what conclusions we might draw from the coding and thematic reading of our participants' responses.

Conclusions and Implications

This exploratory study was designed to find traces of rhetorical agency in real-world applications. Initially, this exploration focused on testing whether Rawlins and Wilson's (2014) predictions about agency in IDDs would be borne out by practical user testing. In particular, we questioned whether different levels of interactivity would elicit different responses and whether we could identify markers of agency in participants' actions and responses. This study has expanded on Rawlins and Wilson's predictions and provided a richer view of agency in action, particularly of how designers and users can occupy a shared rhetorical space in an IDD, and how that space can be identified by both language and action. This richer understanding of agency can best be discussed in terms of affordances and markers of agency.

Affordances

One of the key insights in this study was how user agency is affected by the affordances that interactive graphic designers provide. Withagen et al. (2012) argue that affordances by the designer invite agency on the part of the user. Our study shows that affordances by the designer open up a shared rhetorical space where the user can participate in problem-solving activities, take ownership of the data, and make greater use of the data to influence other people. In other words, affordances allow problem-solving, which in turn invites the user to inhabit an agential role by creating what Herndl and Licona (2007) call “the temporary and contingent conditions of possibility for rhetorical action” (p. 138).

Throughout the study, we saw a clear progression of rhetorical agency across the different experimental conditions and a shift in the focus of that agency from data, to tool, to participant who corresponded with the interactive affordances provided by the IDD designers. In the static group, participants viewed themselves as simply intermediaries between the designer of the graphic and their client. They perceived more external control of the experience and therefore did not take ownership of the consulting task. They were simply relaying information, not synthesizing it for their clients. The participants in the static group focused their shorter, less engaged answers on the data. In particular, they attributed agential power in the task to the data, not to themselves, the designer, or the IDD.

For participants in the zoom-and-pan group, however, the interactive tool was the primary agent. While these participants spent about the same amount of time on task (as measured with Morae software) as those in the static group, they were more engaged with the interface. This engagement became particularly clear in the discussions of usefulness, where these participants focused on the parts of the IDD (such as the slider bar) rather than the information in the tool or on their own role. Even with the additional engagement, however, the zoom-and-pan group still saw themselves as intermediaries between an expert who compiled the information and a client who would use the information.

The participants in the constrained groups showed the most rhetorical engagement. Rather than seeing themselves as intermediaries or conduits, they saw themselves as active agents in creating specific information for their clients. They created their own strategies for working with the information and engaged in more meaningful ways with the assigned task. These participants spent more time on task, gave longer and more complex answers during the posttask interviews, and thought about the task and the information in a more conceptual, self-reflective way. The participants in the constrained groups also discussed usefulness in terms of their own actions and opportunities rather than in terms of the information or the IDD.

In terms of the participants' actions and their responses to interview questions, our results matched up with our initial assumptions that as users were afforded more agency, they would become more active participants in the shared rhetorical space. But this increased agency did not translate to the information being more persuasive. In fact, we found that the participants in the static group viewed the information in the graphics as more persuasive than those in the other groups. There are several possible explanations for this result, however:

In static graphics, the information has been compiled by an exterior expert (in the case of our study, the USDA). This perception of a credible source, combined with the lack of opportunity to manipulate the data, may lead to a greater recognition of persuasiveness. In the zoom-and-pan and constrained groups, the ability to manipulate data for a client made the data more engaging but less persuasive.

The limited opportunities for agency afforded by the designers of the static IDDs may have led users to inhabit a less agential role. In other words, because they were only acting as passive intermediaries, they did not have the opportunity or afforded ability to doubt the information's persuasiveness. As users took more active roles in the zoom-and-pan and constrained IDDs, they also became more active in criticizing the data and the tools.

The persuasive elements in static graphics are overt, leading users to recognize them as being persuasive. The tools that provide more interactivity and engagement in the zoom-and-pan and constrained IDDs may obscure the persuasiveness of the information and the graphic.

All three of these explanations provide some possible insight into the perception of persuasiveness in the information and the graphics across the experimental conditions. The findings support Kimball's (2006) argument that “readers find graphics implying relative clarity more convincing than graphics emphasizing complication” (p. 355). Although Kimball was explaining readers' easier acceptance of static infographics that “imply a more transparent view of reality” (p. 354), our findings perhaps similarly imply a user's easier acceptance of information from static interactions over more demanding interactions.

This finding complicates our view of affordances and agency. While additional affordances offered by the designer do create a shared rhetorical space where user and designer cocreate an IDD, additional affordances also decrease the overall persuasiveness of the information. In more interactive IDDs, therefore, the user is asked to inhabit a larger agential role not just in deciding how the information is displayed in the graphic and presented to others but also in deciding how much to trust (or doubt) the information.

This small study provides only a glimpse into the mechanisms of these perceptions, and the findings may not hold up in larger studies. In particular, the information we chose for our study was deliberately straightforward. We did not want to have the graphics to manipulate a certain outcome for the users. However, many IDDs are designed with the intent to persuade or to provide a specific outcome regardless of user input (see, e.g., the Carbon Footprint Calculator described in Rawlins and Wilson, 2014). If users were working with graphics that had an obvious agenda or dubious information, their perceptions of their agential roles and the persuasiveness of the graphics may shift.

Markers of Agency

The other key insight from this study of IDD users is that we can identify (at least preliminarily) markers of agency in practical tasks. As noted earlier in this article, most discussions of agency are complex and abstract. In this study, that abstract theory is connected to a set of concrete tasks, where users had different opportunities for agency. In our initial planning for the study, we tracked agency through quantitative means: time on task, number of mouse clicks, and so on. While this did yield useful information (which will be the subject of a different article), we present here the findings from the qualitative examination of interview responses. We found markers of agency in our broad analysis of the participants' answers and in our more specific analysis of the verbs they use. While there are many ways that markers of agency can be classified, this study shows two primary categories of markers.

First, there are markers of action, which are primarily verbs that indicate a more active role by the user in the shared rhetorical space. Our study showed that participants who viewed static IDDs used more abstract verbs, while those who viewed IDDs with more interactivity balanced concrete and abstract verbs. Participants who were given the constrained IDDs were also more likely to take additional opportunities for action to manipulate and take control of the information, like the participant who exported the data to Excel to create better materials for his client.

Second, there are markers of actors, which include both verbs and other parts of speech that identify who has the opportunities for agential action. In this study, we found not only that participants attributed more action to themselves with more interactive graphics but also that they attributed more action to others. In other words, they had a greater recognition of the shared nature of the rhetorical space and were more self-reflective about their own role in creating the graphics. As the interactivity of the graphics increased, the participants saw themselves more as rhetorical agents in a community of rhetorical agents rather than as simple conduits of information between the designer and the client.

These two markers of agency, the action and the actor, can provide a more concrete foundation for further tests of agency in applied contexts. Larger studies may identify additional markers that can be used to classify and (in ways) quantify the ethereal concept of agency.

Implications

Returning to our research questions, we have shown that different types of IDDs do elicit different behaviors and user experiences and that these different behaviors contain markers of rhetorical agency. We have shown that users perceive control of information differently and assume different roles in between designers and “clients,” depending on the type of IDD. We have shown that users may alternately focus on the usefulness of data, tool characteristics, or their own actions or opportunities when working with different types of IDDs. These findings have benefits for practitioners and academic researchers, offering insight into how users respond to, and perceive information delivered in, static versus different interactive formats.

Understanding the ways that IDDs facilitate shared agency can help us plan for better technical communication and address audience needs (Stephens et al., 2015, p. 337) . While some IDDs function for simple informational purposes, these interfaces are increasingly being used to guide “real-world” decision-making (Stephens, 2015), so understanding how users interact with different formats is valuable. Practitioners designing data displays could choose different IDD formats depending on whether they want a passive or active user. Designers might also choose different IDD formats with an eye toward the type of interface that will best meet an audience's informational needs or the type of audience and shared rhetorical space that best suits the goals of the designer.

Articles on the role of technical communicators in the workplace might also usefully map onto the type of interactions we observed. If we see the users in this study as technical communicators, we might understand their limited or enhanced agency as similar to the variable categories of technical communication work described by Johnson-Eilola (1996) or Slack, Miller, and Doak (1993). Specifically, we could use aspects of symbolic analytic work (Johnson-Eilola, 1996) or articulatory technical communication (Slack et al., 1993) to better understand the shared rhetorical space that enhanced IDD affordances create for agency. In addition, we could consider the implications when users of IDDs with more interactivity are less persuaded by the numbers and more aware of their own roles.

Ultimately, the findings in this study may lead us to new ways of thinking about rhetorical agency. The participants in our study experienced agency differently depending on the affordances provided by the designers of the IDDs. But those differences in agential experiences also led to differences in actions, thoughts, language, and reflection by the participants. Rawlins and Wilson (2014) attempted to flexibly map a theory of agency across the different shared rhetorical spaces featured on their IDD typology. This could be read as suggesting that maximizing agency is always the best course and, by extension, that tools that afford the most agency are better or less manipulative. However, Wolford (2016) offers a feminist view of agency that does not assume that one theory (or experience) of agency will explain all interactions.

Our findings show that different levels of interactivity, with the accompanying affordances and opportunities for agency, create different outcomes. While we can track the differences with markers of agency, we can't necessarily identify one level of interactivity as being the best, or most persuasive, or most useful. Rather, our findings indicate that different types of IDDs and different levels of agency may each have their place in accomplishing different technical communication goals. Therefore, just as there are multiple, complex, and sometimes contradictory theoretical approaches to agency, there are multiple, complex, overlapping, and contradictory approaches to IDDs. And in both agency and IDDs, the complexity provides for rich academic inquiry and creative practical application.

Limitations

This exploratory study seeks to identify discursive markers of agency in an applied task that models common interactions with different types of IDDs. We borrow usability and qualitative methods and tools in an attempt to make visible behaviors and attitudes that are largely approached only from a theoretical standpoint in technical communication and rhetoric. Our accomplishments are imperfect, but hopefully informative. If future researchers see promise in our coding and counting, then larger sample sizes are warranted for more definitive claims. If future researchers find our questioning too direct or leading, then more subtlety is warranted, but perhaps we have identified what to search for in those more diffuse responses. As an exploratory study, seeking markers of intangible or ineffable phenomena, we are not generating unassailable facts or making fast claims of validity or statistical significance. But we hope to create a space where smart approaches can make more progress toward those goals.

Discussing Bruno Latour's use of the term articulation, Wilson (2013) wrote,

He [Latour] discusses Pasteur's articulation of the lactic acid ferment or the soil researcher's articulation of the clod of dirt in the pedocomparator in terms of bringing an object of research forward so that propositions can be made. Those propositions are transformative in that they allow researchers to argue that these objects are something else and not just themselves. Latour defines this act as “predication,” the need to define a term in other words to avoid a tautology.

In technical communication scholarship, we have only had tautological ways of discussing agency. We have had theories that more or less tell us, “agency is agency.” But this article has pressed to bring methods into a description of the behavior of agency, to bring agency forward as an object so that we can begin to make statements about what kind of thing agency is in a technical and professional setting. The approaches here are raw and limited in their resemblance to robust research on longer standing objects of study.

Appendix A

Participant Prompt

Scenario

Here is your scenario for this test. You are at personal finance consultant working with parents to determine the best conditions for raising a child (conditions including locations, income level, age of the child/children, and future educational goals). Using this data tool, you will perform a set of tasks to determine different scenarios for the parent to consider. Do you have any questions?

Task

Your client is a single parent and earns $50,000 per year, has two children, ages 5 and 9, and hopes to send them both to college. Using the data graphic, how would you advise this parent?

After you have analyzed the data, please tell the facilitator how you would advise the parent?

Appendix B

Open-Ended Interview Questions

Tell me about the plan you developed.

What was the most useful information in the graphic?

What general impressions do you have about your experience with the graphic in this task?

What steps did you have to take to prepare your answers?

How did you use the graphic in preparing your answers?

Thinking about your final answers for your client, what percentage of your final answer did each of the following contribute?

a. Your expertise

b. Information from your client

c. Information from the graphic

d. Other knowledge

As you were interacting with the graphic, how much were you able to control the information and the experience?

What were the shortcomings of the graphic?

Is there something you wanted to do that this graphic wouldn't let you do?

Do you think the person who designed this tool/interface had an impact on the advice you gave?

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Gregory D. Wilson is an assistant professor in technical communication and rhetoric at Texas Tech University. He teaches courses in technical communication, rhetorical theory and criticism, and the rhetoric of science. He researches agency, technology, and culture.

Jacob D. Rawlins is an assistant professor in the Linguistics and English Language department at Brigham Young University, Provo, UT, where he teaches courses in editing, publishing, and grammar. His research interests include professional communication pedagogy, workplace myth-building, and interactive data visualizations.

Kate Crane is an assistant professor in technical communication at Eastern Washington University. She teaches courses in technical communication, usability, and information design. Her research interests include user experience, user-centered design, usability, and information design.

References

Albers

M. J.

(2004) Communication of complex information: User goals and information needs for dynamic web information, Mahwah, NJ: Lawrence Erlbaum.

Albers

M. J.

(2011) Usability of complex information systems. In: Albers

M. J.

Still

(eds) Usability of complex information systems: Evaluation of user interaction, Boca Raton, FL: CRC Press, pp. 3–16.

Barnum

C. M.

(2011) Usability testing essentials…ready, set, test!, Burlington, MA: Morgan Kaufmann.

Barton

B. F.

Barton

M. S.

(1993) Modes of power in technical and professional visuals. Journal of Business and Technical Communication 7(1): 138–162.

Herndl

C. G.

Adela

C. L.

(2007) Shifting agency: Agency, kairos, and the possibilities of social action. In: Zachry

Thralls

(eds) Communicative practices in workplaces and the professions: Cultural perspectives on the regulation of discourse and organizations, Amityville, NY: Baywood, pp. 133–154.

Johnson-Eilola

(1996) Relocating the value of work: Technical communication in a post-industrial age. Technical Communication Quarterly 5(3): 245–270.

Kienzler

(1997) Visual ethics. The Journal of Business Communication 34(2): 171–187.

Kimball

M. A.

(2006) London through rose-colored graphics: Visual rhetoric and information graphic design in Charles Booth's maps of London poverty. Journal of Technical Writing and Communication 36(4): 353–381.

Mirel, B. (1998). Visualizations for data exploration and analysis: A critical review of usability research. Technical Communication, 45(4), 491–509.

10.

Nielsen

(1993) Usability engineering, Mountain View, CA: Morgan Kaufmann.

11.

Rawlins

J. D.

Wilson

G. D.

(2014) Agency and interactive data displays: Internet graphics as co-created rhetorical spaces. Technical Communication Quarterly 23(4): 303–322.

12.

Shearer

(2011) An activity-theoretical approach to the usability testing of information products to support complex use. In: Albers

M. J.

Still

(eds) Usability of complex information systems: Evaluation of user interaction, Boca Raton, FL: CRC Press, pp. 181–206.

13.

Slack

J. D.

Miller

D. J.

Doak

(1993) The technical communicator as author meaning, power, authority. Journal of Business and Technical Communication 7(1): 12–36.

14.

Stephens, S. H. (2015, July). Interactive data visualization for risk assessment: Can there be too much user agency? In Gossett, K. and Mallory, A. (Eds), Proceedings of the 33rd Annual International Conference on the Design of Communication. New York, NY: ACM

15.

Stephens

S. H.

DeLorme

D. E.

Hagen

S. C.

(2015) Evaluating the utility and communicative effectiveness of an interactive sea-level rise viewer through stakeholder engagement. Journal of Business and Technical Communication 29(3): 314–343.

16.

Tufte

E. R.

(1990) Envisioning information, Cheshire, CT: Graphics Press.

17.

U.S. Department of Agriculture. (2014). Expenditures on children by families, 2013. Retrieved from https://www.cnpp.usda.gov/ExpendituresonChildrenbyFamilies.

18.

Wilson

(2013) Arguing for torture: Using articulation theory to understand George W. Bush's military commissions speech. Enculturation: A Journal of Rhetoric, Writing, and Culture 16. .

19.

Withagen

de Poel

H. J.

Araújo

Pepping

G. J.

(2012) Affordances can invite behavior: Reconsidering the relationship between affordances and agency. New Ideas in Psychology 30(2): 250–258.

20.

Wojdynski

B. W.

(2015) Interactive data graphics and information processing: The moderating role of involvement. Journal of Media Psychology 27(1): 11–21.

21.

Wolford

(2016) When a woman owns the farm: A case for diachronic and synchronic rhetorical agency. Enculturation 22. Retrieved from http://enculturation.net/when-a-woman-owns-the-farm.