Visualized Statistics and Students' Reasoning Processes in A ...

Processes in A Post Truth EraJörgen Nissen and Linnéa Stenliden

The self-archived postprint version of this journal article is available at Linköping University Institutional Repository (DiVA):http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-170247

N.B.: When citing this work, cite the original publication.Nissen, J., Stenliden, L., (2020), Visualized Statistics anTruth Era, The Journal of Interactive Learning Research, 31(1), 49-76.

Original publication available at:Copyright: Association for the Advancement of Computing in Educationhttp://www.aace.org/

1

Visualized Statistics and Students’ Reasoning Processes in a Post-True Era

JÖRGEN NISSENLinköping University, Sweden

[email protected]

LINNÉA STENLIDEN Linköping University, Sweden

[email protected]

In these times, described as an ´post-truth´ era where we are faced with information overload, it is a challenge to help students find relevant information, support their knowledge building and engage them in thinking critically about information and knowledge from different perspectives. This study investigates how a visual analytics interface (with a dynalinked view of an interactive map together with interactive graphs) and students interact to solve tasks in secondary schools’ social-science classrooms. Teachers are probably better able to support their students if they know more about the translations and the patterns that emerge when students try to engage in interactive graph reading. We have distinguished three patterns that emerge in the interactions. These patterns, decoding, manoeuvring and incorporation of prior knowledge, is supportive in elucidating the students’ visual and analytical reasoning processes. Insights about those reasoning processes is important since earlier research has highlighted the centrality of considering the problems of the complexity of interactive graph reading and thus dealing with issues concerning students’ abilities to read and interpret such graphs, when presented as part of interactive information visualization technology.

Introduction

In this ‘post-truth’ era, students (and others) face a situation of information overload, both

accurate and fake information is constantly being generated, and they operate within vast

influxes of information available through digital media (Stokols, 2017). Hartshorne, Heafner,

and Thripp (2019) argue that it becomes incredibly challenging in these times to help students

find relevant information, support their knowledge building and engage them in thinking

critically about information and knowledge from different perspectives. One challenge is the

2

circumstances in which objective facts become less influential in shaping people’s opinions

than appeals to emotion or personal belief (Del Vicario et al., 2016; Lewandowsky, Ecker, &

Cook, 2017). In such a society, power may lie with those who are most vocal and influential

on social media: from celebrities and big corporations to botnet puppeteers – fake online

personas through which a small group of operatives can create the illusion of a widespread

opinion (Bu, Xia, & Wang, 2013). Therefore, it is increasingly important to teach students

how to build their knowledge through relevant facts and how to critically discuss and listen to

different perspectives (Joordens, Kapoor, & Hofman, 2019). As an example, schools can

introduce specific technology for data management and presentation and train students to

master these tools. The centrality of employing technology for visual data presentation,

interacting with such data and the reasoning around it in schools has been emphasized by

many (Jern & Stenliden, 2011; Lundblad & Jern, 2012; Stenliden, 2014). It is often

underlined that, the more accessible data becomes, the more visual it is; and the more

comprehensively it can serve to inform teachers and students in 21st-century classrooms, the

more relevant it becomes to implement such study material in schools. So, one way to

address the important educational tasks highlighted in these post-truth times, the development

of students’ ability to work with data, may be to employ visual analytics tools (VA) in

education (Stenliden & Jern, 2012). For students this will involve processes which includes

the action of thinking about something in a logical, sensible way when using interactive

information visualizations. This study explore emerging patterns when students and VA

interact and how these patterns can help to explain students’ reasoning processes.

VA is a research area developing technology that aims to support the management of

large amounts of data and information by using interactive visual interfaces, including

diagrams and graphs. It is a ‘key technology’ for processing ‘raw data’, often in the form of

visualizations of official statistics (Andrienko, Andrienko, Keim, MacEachren, & Wrobel,

2011). The technology builds on the basic idea of integrating human capabilities in terms of

visual information exploration that engages analytical reasoning processes and the processing

power of computers to form a powerful knowledge discovery environment (Andrienko et al.,

2011). VA gives access to a ‘universe’ of data about the ‘real world’ and through this

technology students may be offered access to a range of official databases – there are official

statistics on every neighborhood and nation. Official statistics are an indispensable element in

the information system of a democratic society, providing the government, the economy and

3

the public with data about the economic, demographic, social and environmental situation.

Methods and procedures for the collection, processing, storage and presentation of official

statistical data should consider scientific principles and professional ethics (United Nations

Statistics, 2013). This kind of information may help teachers to build students’ knowledge

and opinions on facts and avoid opinions built on emotional affect.

However, it is important to remember that the increasing use of visual presentations of

data (in the media and everyday life, including VA tools) is based on the assumption that

graphs and diagrams are transparent to the viewer, meaning that they will gain an immediate

understanding of the visual message (Glazer, 2011). But, as Dreyfus and Eisenberg (1990)

pointed out three decades ago, comprehension of (static) graphs may not be transparent since

reading a graph is a complex task. In addition, visual media such as VA offer interactive

diagrams and graphs displayed on an interactive digital screen. These graphs provide

immediate information, but as the data changes after interaction with the visual interface, the

graphs also change. Hence, reading interactive graphs is different from reading printed ones.

Although these graphs help in processing information quickly and efficiently, they often

increase the complexity of translating the visual information (Stenliden, 2015). Even though

the process is highly complex, previous studies have shown that students’ knowledge

construction, in grades 4–9 in the social sciences, does benefit from working with VA

(gaining fruitful insights by working with the interactive visualizations), but that students are

often left alone to interpret the interactive graphs (Stenliden, 2014). In line with these results,

this study argues that it is important to consider the problems of complexity and to develop

knowledge about how best to support students’ ability to read interactive graphs. It is

important for students to understand how each visual property is related to the physical world

(Treagust, Duit, & Fischer, 2017). In addition, students may benefit from understanding how

visual properties might direct attention towards phenomena that are not the focus of their

investigation and that they might end up with information in relation to the physical world of

which they had been previously unaware. It is also important to stress that different

representations of the same data can highlight different aspects of information (Bodén &

Stenliden, 2019). Yet, as these understandings of how vision and the visual properties of

graphs connect and how attention might be distracted are central, it is also important to be

aware that this is only part of the multifaceted process of being literate in graph reading. It is

a process that for example includes visual as well as analytical reasoning (Andrienko et al.,

4

2010). Visual reasoning characterizes axiomatically what one can see in a picture, a diagram

or graph and how an interpretation determines the meaning of such an image when it is used

in reasoning (E. S. Ferguson, 1992; Wang & Lee, 1993). Thus, visual reasoning becomes the

process of analysing visual information and being able to reach a certain conclusion and solve

problems based upon it. This is a component of a nonverbal skill, meaning that a person does

not have to use verbal language in order to solve visual problems (Zazkis et al., 1996).

Analytical reasoning refers to the ability to look at information, qualitative or quantitative,

and discern patterns within it. It is the ability to analyse a situation, think in a step-by-step

method to evaluate alternatives, eliminating those that do not fit the case, and find a solution

that best matches the problem at hand. It is based on propositions that are axiomatic in that

their ‘truth’ is self-evident, commonly expressed reasoning based on facts (Palmquist, 1993).

So, if students are to create meaning through reading graphs by visual interpretation, this

probably means including processes of reasoning.

However, the main diagram research community has not yet, to any great extent,

explored connections between interactive graphs and their interpretation, in relation to

understand reasoning processes, especially not in schools (Purchase, 2014). Therefore, in this

paper we address the problem of graph interpretation and emerging reasoning processes, with

a specific interest in secondary-school classrooms. For our investigation, we view reasoning

as a concept that might help to consider how knowledge constructs heuristics and biases, and

to discuss the potential for uncertainty in a problem-solving and knowledge-building process

(Zuk & Carpendale, 2007). The aim is to map and clarify the reasoning processes in

secondary-school students’ interactive map and graph reading when visual analytics is

employed in social-science education for students aged 14–15 years. The empirical attention

is directed towards the interactions: the connections between students and the visual entities

at a dynalinked interface, emerging reasoning processes and verbal discussions, building

networks in the classrooms. Based on this, our research questions are:

1. What patterns emerge when the students and the visualized official statistics interact?

2. How can these patterns help to elucidate students’ reasoning processes?

The study is situated in social-science education with the aim of developing students’

skills and awareness of circumstances in society. Students therefore seek to solve problems

concerning societal circumstances with the help of graph parsing as they try to extract

5

knowledge from the concurrent data mining. In school classrooms, we see the interactions

between both social and material actors as they work not only with but also upon each other,

coordinating the actions as networks (Latour, 1987). Therefore, what can be studied is how

social and material actors interact and constitute particular patterns, networks, etc., not the

reasoning processes per se. The term translation is used to describe how knowledge appears

and develops, emphasizing that when actors, e.g. social viewers, and the material visual

entities connect, they change one another through their interactions. When this happens, as in

translation, knowledge progresses. Therefore, the paper discusses reasoning abilities and

knowledge-building processes that in the long run might support the development of

students’ opinions and views of the world, their political subjectivity and values as citizens

(Reimers & Martinsson, 2017).

Literature review

To contextualize this study, we will address research on visual and graphical entities and their

use in instructional settings. First, we try to classify the graphical entities, charts, diagrams

and graphs. Then, a short chronological overview of the literature is given, along with a

topical overview of perspectives commonly used in this area of research in relation to

education and schools.

Charts constitute a large group of methods for presenting information in the form of

graphs, diagrams, maps or tables. Graphs provide information by presenting the data in a

visual format. They are composite graphics that consist of a graphic space, a set of

constituents and a set of relationships involving these constituents. The purpose is to show

mathematical relationships between sets of data. In other words, graphs make pictures of

numerical information. Bar graphs, line graphs and pie charts may be the most commonly

used forms of graphs, but many other kinds exist. Maps comprise another category of charts,

with information shown in relation to geography. Charts as diagrams, graphs and maps are

ubiquitous, and, as a means of communicating information, can be found in all areas of

research and practice. Leveraging the power of the visual processing system, they provide a

flexible means for presenting information in an engaging and direct manner (Pettersson,

2019). In this study, a VA that includes interactive graphs, maps and textual information was

introduced into schools (Figure 1). The specific VA, the Statistics eXplorer, can be seen as

offering ways to support both teachers’ ‘transposition’ of ‘raw material/information’

(Ongstad, 2006) and pupils’ “visual and analytical reasoning” (Andrienko et al., 2011).

6

Simply stated, Statistics eXplorer is a storytelling tool that allows teachers, based on their

educational plans, to: (1) access statistical data from, for example, the World DataBank (a

database of official statistics) through a direct API interface, (2) make discoveries and derive

insights by exploring trends and patterns, (3) create a story in which visual discoveries are

captured by snapshots together with descriptive metadata and hyperlinks in relation to

analytical reasoning and (4) share the story with colleagues or students by embedding it in

educational blogs or HTML pages. When the visual story is published, it is called a ‘Vislet’.

Figure 1. The VA introduced is the Statistic eXplorer platform, which makes it possible to: import statistical data from official databases; explore and gain insight into the visualized data; create a visualized interactive story and publish it as a blog or web page.

However, even though different possibilities to collect, organize and manipulate digital

data, as in the Statistics eXplorer, have been discovered and open up a perspective on abstract

facts, new complexities and realities, these graphs and maps are useless if not effectively

communicated and understood. Knowledge acquisition from graphs is a highly complex

process that often presents a multifaceted cognitive challenge to students wherein multiple

representations of information must be decoded and interpreted within a specific learning

context (Lowrie, Diezmann, & Logan, 2011). A thorough understanding of this process of

comprehension requires knowledge from a wide range of disciplines (Purchase, 2014). The

problem of understanding diagrams and graphs received a fair amount of interest during the

1980s and ‘90s (R. W. Ferguson & Forbus, 1998; Srihari, 1995). At the time, much of the

existing literature on graphical representations covered specific types of graphics or specific

aspects of their syntactic structure (Card, Mackinlay, & Shneiderman, 1999). The strategies

built, for example, on rules assuming that the visual primitives worked on a specific set of

diagrams (Horn, 1998). Then, methods were proposed so that graphs might be analyzed as

7

working with vector representations of these diagrams (Kembhavi et al., 2016 ). So far, the

21st century has been mixed in its approach to studying and understanding visuals (Boling,

Gray, Modell, Altuwaijri, & Jung, 2014). Engelhardt (2002) proposed an understanding that

builds upon a representational language of graphic images that works with a diverse set of

diagrams. Scholars studying images and charts as part of multimodal texts emphasize

interpretation on the part of readers as a central feature of how images are used (Engelhardt,

2002; Kress & Van Leeuwen, 2006; Van Leeuwen, 2005). Watkins, Miller and Brubaker

(2004) studied 60 elementary students using visuals in the context of science learning and

observed that these students “demonstrated a propensity for constructing their own

interpretations to describe visual representations”. Only two of the students chose to read any

part of the accompanying text that was intended to clarify the meaning of the image (p. 23).

Eraso (2007) tells us that students explore visual tasks by utilizing both visual and analytical

reasoning. He argues that, through connecting visual and analytical reasoning, students can

narrow down the options to find correct solutions. Research on improving spatial ability has

usually concentrated on students’ visual reasoning, thus ignoring the contribution of their

analytical reasoning to the development of their graph reading (Eraso, 2007). More recently,

Seo, Hajishirzi, Farhadi, and Etzioni (2014) introduced a method for understanding graphs by

the identification of visual elements in a diagram while maximizing agreement between

textual and visual data.

Boling et al. (2014) claim that learners use a fluid, active, and inter-related mix of

strategies to decide what a visual entity is supposed to mean to them in the context of a

learning activity. However, they highlight the often-inaccurate idea that specific visual forms

ensure specific outcomes when used to support learning, and that simple images or visuals

are simple in the context of a learning activity. Instead, as Glazer (2011) notes, there are

several major challenges related to reading visual entities and their relations within graphs

and diagrams: confusing the slope for the height, confusing an interval for a point, conceiving

a graph as constructed of discrete points, focusing on x-y trends, unclear formatting or

inappropriate choice of visual features, and teachers’ expertise (or lack of it). Also, some

graphs are difficult to understand because they provide too much information: “Instead of

using the right graph it is also possible to use the wrong graph” (Glazer, 2011, p. 197). Boling

et al. (2014) explain that there is a rich interplay between meaning-making and decision-

making on the part of learners, who can be ascribed interpretive agency when working with

8

visuals (such as graphs). Bodén and Stenliden (2019) show how, when reading interactive

graphs and maps, secondary students in social-science classrooms become ‘interlinked’ with

the visual properties displayed at the interactive interface. The visual entities in an interactive

graph, for example, such as highlighting, movement, and color, influence the reader’s gaze,

producing multiple possibilities to connect the visuals and the vision – together with the

interactivity, this multiplicity creates an increased complexity. These learners use distinct

types of strategies that are discernible and worthy of further exploration.

Along with (Behnke, 2017), we believe that further research is required, especially

around knowledge-acquisition strategies through graphics. There is little research related to

how students solve tasks in social-science education and the reasoning/analysis processes

present in problem-solving that include interactive diagrams or graphs. As argued, such

knowledge-building processes might affect students’ political subjectivity and values

(Reimers & Martinsson, 2017). Hence, it becomes especially important to know more about

learning activities related to social-science education and the subjects’ problem-solving tasks,

which may include VA.

Theoretical stance

By drawing on Actor-Network Theory (ANT), the activities in classrooms can be

investigated and analyzed using the approach of relational materiality, which means that both

social actors (teachers and pupils) and material actors (technology and other objects and

matters) in interaction form activities (Fenwick & Edwards, 2010). Using this approach, it is

possible to study how the social and material actions together perform particular enactments,

rather than focusing on the humans’ interactions with artefacts, or their use of the tools

(Latour, 2005). Thus, when students and VA technology are studied, working with relational

materiality helps to examine not only how humans interact in new ways through technology

but also how technology (digital and other) plays an active part in shaping those interactions

(in the classroom practice). Hence, by employing the approach of relational materiality, the

expectation is to clarify how actions are shaped by all, both social and material, actors’ (the

VA application, the students, the teachers and other actors) interactions.

In this study, interactions between the social and material actors are understood as

actors who/which create networks. These networks can be seen as assemblages of actors

forming links through interaction (Law, 2007). In other words, the interactions occurring

between different actors in the classrooms together constitute a network, which emerges due

9

to their ability to align in pursuit of their interests. This means that the activities in the

classroom are not seen as being performed under established conditions in a social context;

instead, it is the possible actions of all the actants that are seen to construct the learning

activities within the environment: in the network (Pinch & Bijker, 1987). Here, we will

illustrate how the interactions affect, perform and translate the networks. That is to say, the

interactions between the different actors (the students, the visual entities at a dynalinked

interface, the teachers, the school assignments, emerging reasoning processes and verbal

discussions), become significant to study (Sørensen, 2011).

Within this theoretical approach, manifestations of the actors’ interactions are called

translations. The concept of translation defines what happens when actors interact, changing

each other and forming links. This means that actors do not simply work with other actors;

they also transform and work upon each other. The term emphasizes that when actors – for

example social viewers and the material visual entities – connect, they change one another

through their interactions. This dual/multiple relationship affects and coordinates things and

actions, transforming the actors (Callon, 1986). When this happens knowledge progresses,

and the term translation is used in this paper to describe how knowledge appears and

develops. The translations in a network can vary and may at times weaken and/or strengthen

the ties among the actors (Callon, 1986). Accordingly, exploring the interactions in the

social-science classroom by employing the concepts of translation, we can examine what

enables the network of actors to interact. By following the actors’ interactions, we can track

socio-material relations between students and the different information entities in the

dynalinked view and the emerging reasoning processes, as well as the discussions that

materialize. Altogether, interaction, actor, network and translation will be used as analytical

concepts in this study.

Method: producing data

One issue we encountered when attempting to study interactions between a VA and students

is that, so far, this has only occurred in a few classrooms. In the present case, we must open

lines of communication for introducing our thoughts about VA to teachers. Therefore, the

data was acquired by working with a design-based research perspective (DBR) (Anderson &

Shattuck, 2012). A DBR study can be defined as one situated in a real educational context,

focusing on the design and testing of a significant intervention (Anderson & Shattuck, 2012,

p.16).

10

A teacher–researcher team

In this case, DBR means working closely in a teacher–researcher team (TRT), here a group of

15 teachers and three researchers. In the TRT, the researchers provided theory and previous

research findings related to developing educational models aligned with theories of

implementing VA in an educational setting. The teachers provided their classroom

experience, well-tried methods and didactic skills. The team members shared their views and

knowledge in several meetings, through testing the application and in-depth methodological

discussions. In this way, the teachers discussed their views and knowledge on four occasions

(half or full day). Specific ways of using the VA, instructions, assignments and assessment

methods were produced in the form of lesson plans. Using the VA application Statistics

eXplorer (Lundblad, 2013), two vislets (interactive data visualizations, Figure 2) were also

produced.

Figure 2. Visual properties of the Statistics eXplorer platform: a map, a scatterplot, a chart and a text box with explanations and student assignments (image source: Mikael Jern, Linköping University).

The vislets demonstrate official statistics about the world in an interactive and visual-

analytic manner that makes it possible for students to interact with the visual information,

analyze it and draw conclusions. Statistics eXplorer and other VA applications are intended

to facilitate analytical reasoning and provide a deeper understanding that will enhance

knowledge-building for both students and teachers (Jern & Stenliden, 2011; Lundblad & Jern,

2012; Rosling et al., 2007). The vislets encompass world statistics from the World Data Bank

11

of demographic development and relations between indicators, such as expenditure on

education, health, general net income (GNI), trade and life expectancy. Together, the vislets

consist of seven chapters with instructions and questions. The questions are of two types, one

where the answer can be found by interaction with the graphs and the map, the other is by the

teacher termed ‘reflection questions’. The chapters are: Introduction, GNP and GNP/capita,

HDI (Human Development Index), Population, Import and Export (trade), Export and

Sweden. There are eight questions of the first type and eleven of the second. The developed

lesson plans included plans for how to implement the vislets in the classroom.

Classroom practices

The lesson plans were put to work in two social-science classes in a secondary school, for

K12 students aged 14–15 (grade 8). The teachers introduced the vislet and, together with the

class, reviewed how it works. Then the students, for two lessons, worked in pairs with

different parts of the Vislet. Every part of the lesson involved questions that were formulated

to prompt the students to engage with the statistics and the different forms of visualization.

The students were also instructed to write their answers in a document (using Google Docs).

To do this, they had to use their ability to interpret the visualization, transform the

information and reconstruct their insights. Apart from giving the initial instructions, the

teachers only observed and assisted the students, when requested, during the lessons. During

a third and final lesson, the students were expected to present their findings through a verbal

presentation to another pair of students.

Video captures

Methods of documentation and analysis that advance knowledge about interactions and

relationships during activities are of particular interest when studying media technology in

classrooms. Therefore, video observations were made that facilitated thorough documentation

of the classroom practices. The students’ activities were recorded using their laptops. These

recordings were captured by Camtasia and included the faces, voices and gestures of the

students and their actions and articulations on the screens. In addition to field notes, the

empirical material includes 42 hours of webcam recordings. At the beginning of each lesson,

when the recording started, some students were very aware of the camera, and even talked to

it, but after a while this faded. All the students and their parents were informed about the

study in advance and gave consent in writing.

12

Analytical attention

When trying to detect, comprehend and amplify patterns of interaction (socio-material

relations) in the data the starting-point for this study is relational materiality, as mentioned

earlier. The field notes, the video recordings, and the 42 hours of webcam recordings where

viewed by one researcher. Field notes and video recordings gave an overview of lessons

studied and Camtasia recordings gave a detailed view. Parts of the recordings identified as

crucial to the aim were viewed several times by both researchers and transcribed (Heath,

Hindmarsh & Luff, 2010). In this work it is not enough to observe the dominant work

practices; instead, one must also grasp the complexities as well as the unexpected in the

attempt to understand what is and is not happening in the classrooms. Accordingly, in this

process the idea of moment analysis was used to capture what appeared to be spur-of-the-

moment actions that were highly significant to the actors and the subsequent interactions,

what prompted such actions and the consequences of such moments including the reactions

by other actors (Wei, 2011). This approach is a multisensory enterprise in which perception,

cognition and affect come together in an open-ended practice of sense-making in relation to

the research questions (Massumi, 2002). The purpose was to find ways of engaging that were

attuned to the shifting connectivity of persons and objects (McLure, 2013). Our attention was

focused on moments of interaction within the networks, the translations between different

actors in the classroom, such as, for example, the students and the visual entities in the

dynalinked view (the VA Statistics eXplorer). During this analytical process, the concepts of

interaction, actor, network and translation were used. Altogether, this approach enabled us to

recognize crucial patterns in the students reasoning processes.

Findings

In the empirical data we have discerned how reasoning processes emerge in the networks.

Within these processes, we have identified three patterns that are significant in students’

reasoning processes. The patterns, labelled according to their character, are ‘decoding’,

‘maneuvering’ and ‘incorporation of prior knowledge’. Decoding is defined in this paper as

the ability to understand, interpret and read the diagrams. In its simplest form, it is a question

of understanding what the two axes and a bar together represent on a chart. This is

comparable to reading ‘ordinary’ static charts, but only to a certain extent. In our study there

is also a scatterplot and a map with statistical information on the screen. A crucial difference

is that all the charts and the map in the VA are interactive. The second pattern emerges as

13

maneuvering, meaning manipulating the interactive charts in order to receive/create a

different view/diagram containing the requested data. As will be shown, there are traces of

both decoding and maneuvering abilities when the students and the VA interface interact. The

third pattern appears when the students are working with the given reflection questions and

they incorporate prior knowledge into the translation process. These three patterns will be

illustrated with significant examples consisting of six episodes from four different networks

trying to solve their assignments. Episodes 2, 4 and 5 are from the same network, the other

episodes are from three other networks.

Episode 1: Zooming, decoding and maneuvering

In this network, two boys and the vislet’s visual entities of GNP and GNP/capita interact. The

problem is to discover which three countries have the lowest GNP/capita in the world. Just

before this episode, the network has been occupied with the question of which countries have

the highest GNP and now continues to work:

In the scatterplot, the cursor establishes a square marked by thin lines (Figure 3). This

is an area that can be zoomed (to make the countries inside the square become

enlarged and discernible).

Figure 3. Maneuvering by using the space for zooming.

14

Seven steps of zooming occur. Although some mistakes take place (and once the

students choose to restart from the beginning), after a while part of the screen shows

blue bubbles to the left and some light yellow bubbles to the right (Figure 4).

Figure 4. Decoding the country with the lowest GNP/capita.

The cursor has rapidly moved to the visual entity representing the country with the

lowest GNP/capita: Burundi. The two countries with the next lowest values are

thereafter identified when the curser moves over them and highlights the Central

African Republic and Malawi.

This episode exemplifies how interactions within the network support the students to

find the visual entities in the scatterplot and how the network manages to handle the

interactive features in order to reach a view of the screen where the relevant information can

be detected. The boys have decoded what values the two axes represent and have figured out

where and how to zoom. Thus, without including verbal language, the actors in the network

have together been able to reach a certain conclusion and solve the problem based upon it

(Zazkis et al., 1996). The network has translated and identified the countries with the lowest

GNP/capita. By our identification of patterns of decoding and maneuvering, we can

15

illuminate and thereby understand the visual reasoning process that occurs within this

network.

At the outset of this episode, the parameters in the scatterplot were such that zooming

could be achieved without other prior changes. As the graphs are interactive, the task was

more challenging in the next episode because the parameters must be changed in order to

answer the given question.

Episode 2: Maneuvering through changing parameters

The assignment in this episode is the same as above but the interactions between the students

and the visual interface start by changing the parameter for one axis of the scatterplot.

The cursor moves over the screen and a list of parameters appears. The cursor marks

“GDP per capita (current US$), (Figure 5).

Figure 5. Changing a parameter in the scatterplot.

The scatterplot changes and after some sequences of zooming it shows a couple of

countries gathered in the lower left-hand corner (Figure 6).

16

Figure 6. Ongoing zooming.

The cursor moves and tooltips one of the bubbles and a textbox with information

highlights the bubble ‘Iran, Asia’ (Figure 6). But instead of continuing the zooming,

as in the first episode, the cursor moves down to the bar chart, as can be seen in

Figure 7.

Figure 7. A quick change from scatterplot to bar chart.

17

The cursor then clicks on the control below the graph (see Figure 7, the red arrow

marks this control) and moves it to the right-hand end position. The bar chart now

shows the countries with the lowest GDP/capita. One student quickly both says and

notes: ‘Burundi, Central Africa and Malawi’.

In this sequence, the actors interact and one parameter in the scatterplot is changed. We

interpret this as the network having decoded the diagram (because a conclusion must have

been reached that something was not appropriate to find the answer) followed by

maneuvering through zooming efforts.

The change from scatterplot to bar chart is rather black boxed in this episode. The rapid

shift, after Figure 6, to the bar chart in Figure 7 can be interpreted as being guided by visual

reasoning based on earlier experiences that the bar chart is easier to handle. There are no

utterances or other interactions to explain why the cursor moves from one graph to another.

We understand this as an example of how the process of visual reasoning is hidden most of

the time. However, in the next episode, the visual reasoning is more traceable.

Episode 3: Maneuvering to test whether the decoding is correct

In this episode, the patterns of decoding and maneuvering are easier to discern. The actors,

here two girls and the visual entities, interact while trying to figure out which three countries

import the least.

The curser clicks on the link for a snapshot (‘import the most’). The bar chart

changes, which the girls notice.

A very brief conversation between the two girls demonstrates that they decode the bar

chart as showing the countries with the highest imports:

They test this by sliding (maneuvering) the diagram somewhat to the right. As they

do, the bars for the countries with highest imports disappear from the screen. At the

same time, new countries, with lower imports, turn up on the left-hand side of the

screen. To test if they are correct, they click on the snapshot again and when the bar

chart again shows the countries with the highest imports they conclude: ‘then we’re

right’.

By means of this maneuver, the girls confirm for themselves that they understand how

to read (decode) the graph. The clicking on the snapshot makes the graph change as

predicted.

18

Almost immediately, the chart changes again. Through the interactions between the

girls and the chart, it slides to the extreme left. The girls are then able to conclude that

the final three countries displayed are the ones with lowest imports.

Besides decoding and maneuvering in order to determine whether their interpretation of

the diagrams is correct, as in this episode, the next episode exemplifies how the interactive

graphs are used to test an idea about two related parameters (imports and exports).

Episode 4: Maneuvering to check a causal relationship

The questions to be answered in this episode are: which three countries have the highest

imports and why is it these countries? The network contrives to find the top three countries

and to use the statistics to test an idea about a nexus related to the second question:

The screen includes a bar chart showing that China, the USA and Germany have the

highest imports.

Boy 1: Why is it Germany, China and the USA that have the highest imports?

Boy 2: Because there are very many people living in all this countries. And Germany

consumes a lot of resources, like Sweden. Because of that they must… a lot is

consumed.

Boy 1: Germany needs to buy… they do have a lot of money, don’t they?

Boy 2: Yes.

Boy 1: They’re on an even level, they consume a lot, they buy a lot.

We can check that!

His arm reaches out for the computer, the cursor moves up to ‘Imports of goods and

services’ and modifies it to ‘Exports of goods and services’. The diagram changes and

displays Germany as the country with the third highest exports.

Boy 2: Yes, there we have Germany again, on exports.

Boy 1: Yes, they do have a lot of extremely large industrial cities.

The network verifies the assumption that, if a country has large imports, this is financed

through large exports. The remark about large industrial cities is an example of how prior

knowledge is incorporated into the network, although in a rather weak sense. Visual

reasoning (maneuvering the graph) is here combined with analytical reasoning (the student’s

prior knowledge of Germany being an industrial nation as a confirmation of large exports).

19

The next excerpt is an example of how prior knowledge helps the network to

understand that something is wrong with the parameters in an interactive graph.

Episode 5: Prior knowledge supports decoding

In another phase the same network is searching for the countries with the highest imports. At

the outset, the bar chart is as shown in Figure 8 (except for the visible list of parameters).

Figure 8. Mixing up the parameters between two graphs and a successful decoding.

A click on the first snapshot (the colored words below the chart), which would have

changed the diagrams into showing correct facts, is ignored. Instead, one of the two

boys in the network sees the text at the top left and says: ‘Imports of goods and

services’ (see the thin red line, inserted by us).

This is a mistake because that text/parameter corresponds to the scatterplot above the

text.

Boy 1: Is this imports? How can Bangladesh be so high? Where do they get the

money to import?

But wait!

The cursor moves to the list of parameters for the bar chart (see above) and changes it

to ‘Import of goods and services’. The bar chart changes.

Boy 1: Now it’s correct.

This is an example of decoding, a correct interpretation of the diagram. However, prior

knowledge about Bangladesh (see the doubts about the country having high imports) helped

20

the network to recognize that something is wrong. By thereafter maneuvering the parameters,

a correct diagram is located.

In the next and final episode, the network is concentrating on a question to reflect upon,

and the need to incorporate prior knowledge becomes more obvious.

Episode 6: Prior knowledge as a necessity to solve the task

When the network tries to tackle the question of which three countries have the lowest

GNP/capita, and most importantly why it is those countries, it quickly, through the bar chart,

finds that it these countries are Burundi, Malawi and the Central African Republic. The two

boys recognize that all of them are in Africa and decide not to produce country-specific

answers but instead to concentrate on why it is three African countries that have the lowest

GNP/capita. During a conversation, the boys mention factors such as a lack of natural

resources to export, issues of drought, that they are not so developed countries, having few

modern industries and a lack of well-educated citizens. Whereupon the following dialogue

takes place:

Boy 2: Yes, we can use that /referring to the presentation they are expected to make

the next day/. Few well-educated.

Boy 1: They lack training for qualified jobs

/…/

Boy 1: Then they’ve had, what is it called, they have…. like this …. someone has

colonized them, so they’ve been behind all the time

Boy 2: Yes! Colonization could be the reason. That other countries have taken over

Africa, taken the natural resources that existed. They did that. Took it for themselves.

/…/

Boy 1: Yes

Boy 2: Then there was not so much left. Then Africa didn’t mean so much anymore,

then they had no use for Africa anymore /…/

Boy 1: Yes

Boy 2 So then it just got to be…the poor got to stay, those who couldn’t afford to

move on

Boy 1: Then there was the slave trade too.

Boy 2: Yes, colonies and the trafficking of slaves. This we can develop in more detail

tomorrow.

21

The reference to ‘tomorrow’ concerns a presentation of their conclusions to another

pair of students.

Our interpretation is that, besides the process of visual reasoning (based on patterns of

decoding and maneuvering) used to identify the three countries with the lowest GNP/capita,

analytical reasoning (based on prior knowledge) is employed when addressing the question of

why (it is these countries). The network produced a hypothesis about still-active traces of

colonization and the trafficking of slaves. These are important factors but are not handled in

the VA, although some of the other aspects mentioned, for example the trading of natural

resources and levels of education, certainly can be searched for within a VA.

The pattern incorporation of prior knowledge is integrated into the network when

dealing with the question of why these three African countries have the lowest GNP/capita.

To reach an explanation, the network incorporates factors that are not found in the VA but

build upon earlier experiences. Sometimes sufficient facts can be found in the statistics, but in

this episode, the choice is to rely on the students’ ideas (colonization and the slave trade)

when preparing their presentation for another group. The circumstances are somewhat

different in episode 5, where prior knowledge (about Bangladesh) leads to the insight that the

parameters in the graph are incorrect and that accurate facts will not be readable without a

change. In episode 5, prior knowledge helps with handling the graphs and in episode 6 it is

needed because an answer to the ‘why question’ is not to be found solely within the VA.

Emerging reasoning processes elucidated by the patterns

In relation to our first research question, we have been able to distinguish the three patterns of

decoding, maneuvering and incorporation of prior knowledge in the translations. These were

illustrated above using empirical episodes. The recognition and description of these patterns

is in turn supportive in elucidating the students’ reasoning processes.

Visual and analytical reasoning are earlier attempts to interpret how graph reading is

done. However, these are analytical/theoretical concepts that cannot be easily observed. But,

supported by the patterns discerned here, it is possible to get a glimpse of the reasoning

processes. We see decoding as part of visual reasoning. This is mainly a way of reading the

information hidden in the graphs. In our interpretation, maneuvering is also mainly related to

visual reasoning since it concerns how to handle the graphs’ interactive features – in order to

find an adequate configuration – and then read/interpret it in relation to the given task. In our

recordings, decoding and maneuvering are mostly achieved without much talk. The

22

incorporation of prior knowledge, on the other hand, is frequently combined with

conversations between the students. This is not unexpected since it is not only a matter of

reading graphs for fact-finding but is more related to situations within the networks where

doubts have arisen, or the given task cannot be solved entirely by interacting with figures

within the VA. The network needs verbalized information to successfully continue with the

ongoing work. A conclusion is therefore that the higher the degree of integration of prior

knowledge and analytical reasoning, the more oral conversation will be used.

Discussion

The aim of this study is to map out and clarify the reasoning processes used in secondary-

school students’ interactive map and graph reading when visual analytics is employed in

social-science education. Using almost a basic research approach, the analysis shows that the

interactions, the connections between students and the visual entities at the dynalinked

interface, emerged mainly as three patterns. It is argued that these patterns help in elucidating

the reasoning processes in which students engage when graph reading in the classroom. This

knowledge is important because earlier research has highlighted the centrality of considering

the problems of the complexity of interactive graph reading and thus dealing with issues

concerning students’ abilities to read and interpret such graphs, when presented as part of

interactive information visualization technology. Such abilities are central if students are to

be able to navigate in a society pervaded by data and false facts.

However, this study is limited in the sense that the patterns found can only be seen as

parts of a larger and even more complex case – the multifaceted process of reading graphs,

which includes many more abilities that can be linked, for example, to visual, data,

information, and media literacy.

Nevertheless, teachers are probably better able to support their students if they know

more about the translations and the patterns that emerge when students try to engage in

interactive graph reading. This is in line with Treagust et al. (2017), who argue that teachers

need to consider the use of both visual and non-visual processing strategies in diagram and

graph reading. We argue that the strategies outlined in this paper can serve as such important

understandings for teachers supporting students’ knowledge building. That kind of

comprehension may also assist them in employing VA in the teaching, as it is a key

technology when it comes to handle large volumes of data about the ‘real world’ and the

processing of ‘relevant facts’. Such learning activities – analysis from interactive graph

23

reading – might in the long run contribute to the development of students’ opinions and views

of the world. At the same time teachers need to engage the student’s in thinking critically

about information and knowledge from different perspectives. The spread of misinformation

online has been ranked as one of the 10 most significant issues currently facing the world

(WEF, 2013) and is threatening democracy. Facts, such as official statistics, need to be more

influential in shaping public opinion than appeals to emotion or personal belief (Del Vicario

et al., 2016; Ehrenberg, 2012; Rosling, Rönnlund, & Rosling, 2007) in order to arrive at the

well-known position that “everyone is entitled to his own opinion, but not to his own facts”

(Moynihan, 2010). In order to cope with the demands of contemporary society in terms of

information overflow combined with fake news, schools, teachers and students should do

more than just focus on the core subjects. It is vital to develop the students’ ability to

participate in society as democratic citizens.

Schools, as Martinsson and Reimers (2017) argue, can support knowledge construction

that might, over time, support the development of students’ opinions and views on the world,

their political subjectivity. As future citizens, students need to know how to use their

knowledge and skills, applying them to new situations, analyzing information,

comprehending new ideas, communicating, collaborating, solving problems and making

decisions (Ananiadou & Claro, 2009; Glazer, 2011; Hepworth & Walton, 2011; Schleicher,

2011). Students’ knowledge and opinions should be built on facts rather than on emotional

affect. This article is an effort to contribute to such a mission. The three patterns we have

identified and the possibility that they can help us to become acquainted with both visual and

analytical reasoning processes emerging in graph reading will hopefully be useful for further

research and development.

References

Ananiadou, K., & Claro, M. (2009). 21st Century Skills and Competences for New Millennium Learners in OECD Countries. OECD Education Working Papers, No. 41. Retrieved from http://search.ebscohost.com/login.aspx?direct=true&db=eric&AN=ED529649&site=eds-live

Anderson, T., & Shattuck, J. (2012). Design-Based Research. Educational Researcher, 41(1), 16-25. doi:doi:10.3102/0013189X11428813

Andrienko, Andrienko, N., Demsar, U., Dransch, J., Dykes, J., Fabrikant, S. R., Tominski, C. (2010). Space, time and visual analytics. International Journal of Geographical Information Science, Special Issue: Geospatial Visual Analytics: Focus on Time

24

Special Issue of the ICA Commission on GeoVisualization, Volume 24(10). doi:10.1080/13658816.2010.508043

Andrienko, Andrienko, N., Keim, D., MacEachren, A. M., & Wrobel, S. (2011). Challenging problems of geospatial visual analytics. Part Special Issue on Challenging Problems in Geovisual Analytics, 22(4), 251-256. doi:10.1016/j.jvlc.2011.04.001

Behnke, Y. (2017). Mapping the Reading of Graphic Visualisations in Geography Textbooks In N. Pyyry, L. Tainio, K. Juuti, R. Vasquez, & M. Paananen (Eds.), Changing Subjects, Changing Pedagogies: Diversities in School and Education. Helsinki Oy Nord Print Ab, Helsinki.

Bodén, U., & Stenliden, L. (2019). Emerging Visual Literacy through Enactments by Visual Analytics and Students Designs for Learning, 11(1), 40-51. doi:http://doi.org/10.16993/dfl.108

Boling, E., Gray, C. M., Modell, M. G., Altuwaijri, A., & Jung, J. (2014). Learners Interpreting Instructional Images: Meaning-Making and Decision-Making Strategies. Journal of Visual Literacy, 33(2), 27-52. doi:10.1080/23796529.2014.11674716

Bu, Z., Xia, Z., & Wang, J. (2013). A sock puppet detection algorithm on virtual spaces. Knowledge-Based Systems, 37, 366-377. doi:https://doi.org/10.1016/j.knosys.2012.08.016

Card, S. K., Mackinlay, J. D., & Shneiderman, B. (1999). Readings in information visualization: using vision to think. San Francisco: Kaufmann.

Del Vicario, M., Bessi, A., Zollo, F., Petroni, F., Scala, A., Caldarelli, G., & Quattrociocchi, W. (2016). The spreading of misinformation online. Proceedings of the National Academy of Sciences, 113(3), 554-559. doi:10.1073/pnas.1517441113

Dreyfus, T., & Eisenberg, T. (1990). Paper presented at the Proceedings of the 14th PME conference, Mexico City.

Ehrenberg, R. (2012). Social media sway: Worries over political misinformation on Twitter attract scientists' attention. Science News, 182(8), 22-25. doi:10.1002/scin.5591820826

Engelhardt, Y. (2002). The language of graphics : a framework for the analysis of syntax and meaning in maps, charts and diagrams. The Netherlands: Universiteit van Amsterdam, Amsterdam.

Eraso, M. (2007). Connecting visual and analytic reasoning to improve students' spatial visualization abilities: A constructivist approach (Dissertation ProQuest ETD Collection for FIU). Florida International University, Florida. Retrieved from https://digitalcommons.fiu.edu/dissertations/AAI3298580

Fenwick, T. J., & Edwards, R. (2010). Actor-network theory in education [Elektronisk resurs] (Vol. 1st ed.). London; New York: Routledge.

Ferguson, E. S. (1992). Engineering and the mind's eye. Cambridge, Mass.: MIT Press. Ferguson, R. W., & Forbus, K. D. (1998). Telling juxtapositions: Using repetition and

alignable difference in diagram understanding. In K. Holyoak, D. Gentner, & B. Kokinov (Eds.), Advances in Analogy Research (pp. 109-117). Sofia: New Bulgarian University.

Glazer, N. (2011). Challenges with graph interpretation: a review of the literature. Studies in Science Education, 47(2), 183-210. doi:10.1080/03057267.2011.605307

Hartshorne, R., Heafner, T. L., & Thripp, R. (2019). Handbook of Research on Emerging Practices and Methods for K-12 Online and Blended Learning [Elektronic resourse].

25

Heath, C., Hindmarsh, J., & Luff, P. (2010). Video in qualitative research: Analysing social interaction in everyday life. Los Angeles: Sage.

Hepworth, M., & Walton, G. (2011). A longitudinal study of changes in learners' cognitive states during and following an information literacy teaching intervention. Journal of Documentation, 67(3), 449-479. doi:10.1108/00220411111124541

Horn, R. E. (1998). Visual language: global communication for the 21st century. Bainbridge Island, Wash.: MacroVU, Inc.

Jern, M., & Stenliden. (2011). Visual Storytelling applied to educational world statistics. Paper presented at the Statistics for policymaking: Europe 2020, Charlemagne Building, Brussels.

Joordens, S., Kapoor, A., & Hofman, B. (2019). Let's riff off RIFS: Relevant, interesting, fun, and social - Best practices for engaging the online mind. In T. Heafner & R. Harsthorne (Eds.), Handbook of Research on Emerging Practices and Methods for K-12 Online and Blended Learning: IGI Global.

Kembhavi, A., Salvato, M., Kolve, E., Seo, M. J., Hajishirzi, H., & Farhadi, A. (2016 ). A Diagram Is Worth A Dozen Images. ECCV. Paper presented at the European Conference on Computer Vision.

Kress, G. R., & Van Leeuwen, T. (2006). Reading images: the grammar of visual design (Vol. 2. ed.). London: Routledge.

Latour, B. (1987). Science in action: how to follow scientists and engineers through society. Cambridge, Mass.: Harvard Univ. Press.

Latour, B. (2005). Reassembling the social : an introduction to actor-network-theory. Oxford: Oxford University Press.

Lewandowsky, S., Ecker, U. K. H., & Cook, J. (2017). Beyond Misinformation: Understanding and Coping with the “Post-Truth” Era. Journal of Applied Research in Memory and Cognition, 6(4), 353-369. doi:https://doi.org/10.1016/j.jarmac.2017.07.008

Lowrie, T., Diezmann, C., & Logan, T. (2011). Understanding graphicacy: students’ making sense of graphics in mathematics assessment tasks. International Journal for Mathematics Teaching and Learning.

Lundblad, P., & Jern, M. (2012). Visual Storytelling in Education Applied to Spatial-Temporal Multivariate Statistics Data. In J. Dill, R. Earnshaw, D. Kasik, J. Vince, & P. C. Wong (Eds.), Expanding the Frontiers of Visual Analytics and Visualization (pp. 175-198): Springer.

MacLure, M. (2013). The Wonder of Data. Cultural Studies Critical Methodologies 13(4), 228 - 232, https://doi.org/10.1177/1532708613487863

Massumi, Brian (2002). Parables for the Virtual: Movement, Affect, Sensation. Duke University Press.

Moynihan, D. P. (2010). Daniel Patrick Moynihan: A portrait in letters of an American visionary (S. R. Weisman, Ed., 1st ed.). New York, NY: PublicAffairs. Ongstad, S. (2006). Fag i endring. Om didaktisering av kunnskap. In S. Ongstad (Ed.), Fag

og didaktikk i laererutdanning. Kunnskap i grenseland (Vol. Oslo:, pp. 19-57): Universitetsforlaget.

Palmquist, S. R. (1993). Knowledge and Experience. In Kant's system of perspectives: an architectonic interpretation of the critical philosophy. Lanham: Univ. Press of America.

26

Pettersson, R. (2019). Information Design, Volume 3: Image Design [Electronisc resourse]. Institute for Infology.

Purchase, H. C. (2014). Twelve years of diagrams research. Journal of Visual Languages & Computing, 25(2), 57-75. doi:https://doi.org/10.1016/j.jvlc.2013.11.004

Reimers, E., & Martinsson, L. (2017). Education and political subjectivities in neoliberal times and places: emergences of norms and possibilities. London: Routledge.

Rosling, H., Rönnlund, A. R., & Rosling, O. (2007, 2004). New software brings statistics beyond the eye. Paper presented at the OECD World Forum on Key Indicators, Teatro Massimo, Palermo, Italia.

Schleicher, A. (2011). The case for 21st century learning. OECD Observer(282), 42-43. Retrieved from http://search.ebscohost.com/login.aspx?direct=true&db=buh&AN=62171507&site=eds-live

Seo, M., Hajishirzi, H., Farhadi, A., & Etzioni, O. (2014). Understanding in Geometry Questions. Paper presented at the AAAI Conference on Artificial Intelligence, Québec.

Srihari, R. K. (1995). Computational Models for Integrating Linguistic and Visual Information: A Survey. Artificial Intelligence Review, 8, 349-369.

Stenliden. (2014). Visual Storytelling Interacting in School Learning Conditions in the Social Science Classroom. (20). Linköping University, Linköping. Retrieved from http://liu.diva-portal.org/smash/record.jsf?parentRecord=diva2%3A410571&pid=diva2%3A719298&dswid=9340

Stenliden. (2015). Visual Analytics in K 12 Education – Emerging Dimensions of Complexity. International Journal of Social, Behavioral, Educational, Economic, Business and Industrial Engineering, 9(2), 663-671.

Stenliden, & Jern, M. (2012). Visual Storytelling – Knowledge and Understanding in Education. Journal of Systemics, Cybernetics and Informatics, 10(5), 7-13. Retrieved from http://www.iiisci.org/journal/CV$/sci/pdfs/HAB432WE.pdf

Stokols, D. (2017). Social Ecology in the Digital Age: Solving Complex Problems in a Globalized World. London: Academic Press.

Sørensen, E. (2011). The materiality of learning: technology and knowledge in educational practice. Cambridge; New York: Cambridge University Press.

Treagust, D. F., Duit, R., & Fischer, H. E. (2017). Multiple Representations in Physics Education [Elektronisk resurs]. Cham: Springer.

United Nations Statistics, D. (2013). Fundamental Principles of Official Statistics. Retrieved from http://www.unece.org/?id=3207

Van Leeuwen, T. (2005). Introducing social semiotics / Theo van Leeuwen: London: Routledge, 2005.

Wang, D., & Lee, J. R. (1993). Visual Reasoning: its Formal Semantics and Applications. Journal of Visual Languages & Computing, 4(4), 327-356. doi:https://doi.org/10.1006/jvlc.1993.1020

Wei, L. (2011). Moment Analysis and translanguaging space: Discursive construction of identities by multilingual Chinese youth in Britain, Journal of Pragmatics, 43(5), 1222-1235. doi.org/10.1016/j.pragma.2010.07.035.

27

Financial support

The research was supported by funding from the Swedish Research Council (grant No.

2015.01280) and from Marcus and Amalia Wallenberg memorial fund (grant No.

2014.0120).

Acknowledgement

The authors thanks in particular Ulrika Bodén, Linköping University, for valuable

collaboration, and the teachers and students who participated in the study. We like as well

express our warm thanks to Professor Mikael Jern posthumously. He contributed in so many

ways to our work and made visual storytelling possible for education.