Creating realistic map-like visualisations: Results from user studies

ARTICLE IN PRESS

JID: YJVLC

[m5G;October 9, 2017;8:28]

Journal of Visual Languages and Computing 0 0 0 (2017) 1–11

Contents lists available at ScienceDirect

Journal of Visual Languages and Computing journal homepage: www.elsevier.com/locate/jvlc

Creating realistic map-like visualisations: Results from user studies Patrick Cheong-Iao Pang a,∗, Robert P. Biuk-Aghai b, Muye Yang b, Bin Pang b a b

School of Computing and Information Systems, The University of Melbourne, Victoria, Australia Department of Computer and Information Science, Faculty of Science and Technology, University of Macau, Macau SAR, China

a r t i c l e

i n f o

Article history: Received 15 February 2017 Revised 22 September 2017 Accepted 25 September 2017 Available online xxx Keywords: Maps Information visualisation Map-like visualisation Hexagon tiling algorithm

a b s t r a c t Maps have traditionally been used for displaying geographical information. However, apart from this obvious purpose, the metaphor of maps has been applied to other uses, such as information visualisation and novel user interfaces, since the map metaphor is easy-to-understand and allows users to explore data intuitively. There are several methods for creating these map-like visualisations and user interfaces, but there is little understanding on how people perceive these non-geographical maps, and how to make the visualisation output more realistic. As such, we aim to find preliminary answers on these issues by conducting user studies with a series of map-like visualisations. In this paper, we report on the results of the studies and reveal the factors that have an impact on the human perception of visualisations that are designed to resemble geographic maps. Based on these results, we propose design suggestions for building realistic map-like visualisations. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction In modern society, nearly everyone is familiar with maps, defined as “drawings or other representations of the earth’s surface or a part of it made on a flat surface, showing the distribution of physical or geographical features...” [29]. Since their emergence millennia ago to represent what we know about our surroundings, many kinds of maps have been created to present various information, e.g. showing places with geographic maps, illustrating terrain with topographic maps, demonstrating demographic attributes with thematic maps, etc. Despite their different appearance, almost all people, from childhood onwards, have at least some sense of how to read a map. Research has confirmed that, thanks to the early exposure of maps in education, our ability to read maps starts developing in pre-school [7,16]. In addition, maps allow us to freely explore and navigate places around the world [8]. Being understandable and navigable by most people are advantages that researchers have used to employ the map metaphor for other purposes by generating map-like visualisations, in which the underlying data is depicted using the figures, shapes and notations used in geographic maps but do not represent geographic coordinates [3,31]. In terms of their applications, map-like visualisations have been used to represent knowledge domains (e.g. a corpus of scientific papers) or hierarchical data (e.g. a file system). In these cases, large and complicated datasets can be represented through ∗

Corresponding author. E-mail addresses: [email protected] (P.C.-I. Pang), [email protected] (R.P. Biuk-Aghai), [email protected] (B. Pang).

the map metaphor, so that users can perceive the information as if reading a geographic map, without the need for prior training [4]. In such cases, data can be more easily searched and is more discoverable for novice users [31]. Whereas in recent years there have been several research efforts devising methodologies for creating map-like visualisations, we have little understanding of the factors that make end-users utilise the map metaphor for understanding the content conveyed through these maps. More particularly, whether or not the visualisation is being viewed as a map affects how people perceive the information [24]. To our knowledge, most of the research work focuses on technical details of implementation, rather than on the design process for making the generated results appear similar to geographic maps. This leaves a gap in the literature on how to create map-like visualisations and how to make them more realistic. In this paper, we aim to answer this question: “What makes people regard a map-like visualisation as a geographic map?” From the answers to this question we will derive design suggestions for those who wish to create more realistic visualisations. As such, we conducted two experiments involving reading both geographic maps and map-like visualisations, and collected feedback from human participants. The first experiment, as presented in an earlier VINCI conference [32], obtained a basic overview of the factors that affect the realism of map-like visualisations, and the collected data were synthesised into themes for explaining why (and why not) certain images were perceived as maps. Based on these findings, we conducted a follow-up experiment and asked participants to compare visualisation images with different parameters of the best visualisation reported in the first experiment, so that we can fur-

https://doi.org/10.1016/j.jvlc.2017.09.002 1045-926X/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC 2

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

ther identify the parameters that lead to a better correspondence to the map metaphor. Finally, we discuss several factors and design recommendations for generating realistic map-like visualisations. This paper is structured as follows: we first review the applications of the map metaphor and the recent development of map-like visualisations, followed by our research design and an overview of participation, and then we discuss the results and qualitative feedback collected in our studies. At the end of this article, we propose a framework and design recommendations for creating realistic map-like visualisations. 2. Related work 2.1. The map metaphor The general population is taught to read and use geographic maps at an early stage of education [7]. Maps are typically created using formal methods and metaphors, employing unified visual languages regardless of the type of data [13,15]. A map can be seen as a communication vehicle to convey abstract information to readers [24]. Therefore, people are easily able to understand the information across different kinds of maps, without additional prior training. Exploiting this advantage, scientists take a step further to incorporate the map metaphor to produce “science maps”, so that a vast amount of scientific information can be conveyed to average readers [9,43], and even to children [8]. Although not representing any geographical or topological data, the map metaphor relies on the correspondence between map elements (e.g. regions and borders) and non-spatial data (e.g. data points) to convey information [13,14,32]. In the cartography domain, the formal process of translating non-spatial data into geographical representations is called spatialisation [40]. Skupin and Fabrikant highlight the interdisciplinary nature of spatialisation which requires the collaboration of cartographers, computer scientists, human-computer interaction experts, and data engineers to enhance the outcomes of the visualisations [39]. Another important feature of geographic maps is that they enable exploration and navigation [8]. From the days we rolled a globe to find interesting places, evolving to using Google Maps nowadays, we actually use maps to freely explore spatial information. Studies have investigated the possibility of exploring nongeographic data with this type of intuitive interaction, by spatialising arbitrary data into the map metaphor. Examples include exploring teaching materials in high school education [31] and assisting in the discovery of topics in conference proceedings in a knowledge domain [37]. The literature about visual cognition helps us to understand how maps are interpreted. Marr proposes that people create sketches from the retinal image, and then progressively enhance the sketches to establish the perception of an object [26]. Another model from Pinker suggests that the cognition of a graph is an iterative activity [33], in which the message in the graph will be obtained after a series of visual encoding processes. Additionally, the figure-ground theory shows a number of factors that affect people to recognise shapes and regions in maps [24]. Particularly, the concepts about the contour, surroundedness, relative size and convexity of a figure are relevant in the map-like visualisation context. In the next section, we review some algorithms of producing such visualisations and their recent development. 2.2. Map-like visualisations Information visualisation helps people interact with large amounts of data by transforming data into an effective visual form [34,41]. Whereas many types of information visualisation require

training to be able to interpret the displayed data, map-like visualisation inherits the benefits of the map metaphor which allows intuitive reasoning, better accuracy and ease of comprehension [4,6,30]. In addition, a study has found that lay people prefer visualisations with a standard appearance and abstracted information [35], which makes map-like visualisations potentially useful in many applications. Fig. 1 presents examples of images produced by several map-like visualisation algorithms, which we discuss next. Map-like visualisation, also called metaphoric maps [13], can be applied to different kinds of data. Skupin uses the visualisation to show the academic world of geography by mapping conference abstracts into a map [38] (Fig. 1a). Later, an improved version was created for better cognitive plausibility when visualising user-generated content such as Wikipedia [10]. Mashima et al. convert graphs into maps by clustering relevant nodes into large areas ¨ [27] (Fig. 1b). Gronemann and Junger visualise networked graphs as topological maps [19] (Fig. 1c). Knees and his colleagues create a virtual map of music which lets users explore and navigate in music repositories freely [21] (Fig. 1d). This series of works show the possibilities of mapping different geographic elements (such as topography and attitudes) to different categories of data. In addition, using Fig. 1e and 1f as examples, map-like visualisations can be used for showing hierarchical data such as document corpora, organisational charts and file systems [1,4]. By representing data hidden in levels of the hierarchy as map elements (e.g. nations, provinces, cities), readers can perceive the relationship between data entities in the hierarchy with familiar notations of maps. At the same time the structure of the hierarchy and the data within are visually revealed without the need for traversing all the levels. A recent study by three of the authors shows that map-like visualisations can perform better than treemaps in some scenarios when they are used to present the same set of hierarchical information [5]. As seen above, map-like visualisations use the mapping between map elements and the actual data to convey information. This is particularly useful for novice users without specific training of reading the visualisation [3]. However, if a visualisation image does not look like a map, this correspondence cannot be established. The judgement of the degree of realism of a map-like visualisation is subjective and varies from individual to individual. In the next subsection we review the recent work about interpreting information visualisation, which sheds light on understanding how people regard an information visualisation as a geographic map. 2.3. The interpretation of information visualisation There are several frameworks in the human-computer interaction (HCI) discipline which can help to understand the process of interpreting the output of information visualisation algorithms. Card et al. have proposed a now classic human information processing model called Model Human Processor (MHP) [12], which compares the cognitive process with an information processor, suggesting that the human as a system follows a linear process of stages to make sense of inputs and produce outputs. Although many researchers argue that the MHP model has limitations in accounting for how people interact with computers in the presence of other factors [36], many other cognitive models have been derived from MHP, such as Goal-Operators-Methods-Selection (GOMS) model [12], Model Human Processor with Real Time Constraints (MHP/RT) [20], and Queueing Network – Model Human Processor (QN-MHP) [23]. These frameworks suggest that there is a mental process flow in the human cognitive process. Other research investigates more specifically issues involving novice users who do not have proper training or relevant expertise to read visualisations. Designing visualisations for non-expert audiences is a known challenge [11,18], but a proper and effec-

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

3

Fig. 1. Images produced by different map-like visualisation algorithms.

tive design can address these issues [17]. These findings show the need for understanding how people interpret map-like visualisations and making them realistic. In addition, Lee et al. have proposed the Novice’s Information Visualisation Sensemaking (NOIVS) model [22], which illustrates five activities involved in the sensemaking of information visualisation. Such work helps to explain the feedback of our participants and to conceptualise our findings. These theoretical frameworks and models wrap up this literature review section. In the next section we continue to describe the purpose and the design of our experiments.

3. Research design The aim of our research is to discover factors that contribute to users perceiving visualisations as geographic maps, and to derive design recommendations for map-like visualisation algorithms from these factors. As such, we designed and conducted two experiments with human participants viewing and providing feedback on a number of map-like visualisation images. These experiments were carried out through a web-based survey, one done remotely and one with participants in the lab. The two experiments were set up with different foci and used different question sets. The first experiment was an exploratory study which reflected the broad range of factors that affects the perception of the map metaphor. In this experiment, we asked our participants to review a number of images and to provide feedback on each image. After the data analysis of the first experiment, we selected one of the surveyed map-like visualisation algorithms to conduct a second experiment whose focus was to determine the impact of different visualisation parameters on the readability of the resulting map-like visualisation images.

In the following subsections, we briefly introduce the map-like visualisations used in these experiments, as well as the detailed design of each experiment. 3.1. Visualisation used In order to test the perception of different map-like visualisations, we employed four of our own algorithms that produce maplike visualisation images, as follows: • • • •

Polygon Expansion Algorithm - PEA [2] Enhanced Polygon Expansion Algorithm - EPEA [6] Hexagon Tiling Algorithm - HTA [4] Enhanced Hexagon Tiling Algorithm - EHTA [42]

The selection of these algorithms was mainly based on the availability of source code or runnable code, as well as the technical feasibility, such as the types of input data and system requirements. To briefly explain, all these four algorithms transform hierarchical data into a map-like visualisation with nested areas corresponding to nesting relationships in the data. Areas in these visualisations are sized to correspond to a size attribute of the underlying data, which could represent the number of files in a directory of a file system, or the number of sub-categories in the category hierarchy of a library catalogue, to name just two potential applications. The size of areas is assigned in two principally different ways: the PEA and EPEA algorithms emulate the pouring of a liquid onto a surface, such that the liquid expands outwards from the point where it touches the plane until it reaches a predefined minimum height; the HTA and EHTA algorithms tile hexagons in a lattice of hexagons, beginning at a starting point and then expanding outwards in a random fashion. The EPEA algorithm differs from PEA in

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

ARTICLE IN PRESS

JID: YJVLC 4

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

Fig. 2. Images presented to participants (4 pre-processed geographic maps and 4 map-like visualisations).

that it better handles anomalies in area expansion, and thus produces smoother areas. The EHTA algorithm differs from HTA in that it guarantees correct area sizing and produces areas that look more irregular, similar to areas in a geographic map. More details of each algorithm can be found in the above cited papers.

3.2. Design of the first experiment In our first experiment we showed participants a range of different map-like visualisations produced by different algorithms and asked them to provide both qualitative and quantitative feedback on these visualisation images. Thereby we were able to determine which visualisation algorithm performs better in terms of generating realistically looking map images, as well as the reasons why our users regard the images as maps. The feedback we obtained helped give our effort to improve our algorithms a preliminary direction. In this experiment, we showed a mix of eight images (Fig. 2), including four pre-processed geographic maps (R1-R4) and four map-like visualisation images (V1-V4). As to the data used for generating these images, for the geographic maps the compositions of the jurisdictions of the corresponding countries were retained but the colouring was introduced in order to give the impression of separate larger regions; whereas for the map-like visualisations each image illustrated a randomly generated two-level hierarchy. Each hierarchy contains five regions. This made our images more readable and comparable when used by our participants. As the focus of this experiment was not on understanding the data but on determining which images looked like a map, we did not create deeper or wider hierarchies. We created this particular design with both real geographic maps and generated visualisation images for two reasons: firstly, to verify that making slight modifications to the map images for the purposes of our experiment will have no significant effect on the responses, i.e. we expect participants will still consider them to be geographic maps after our modifications; secondly, to provide a contrast to the map-like visualisation images, which can help elicit responses on the differences between geographic maps and maplike visualisations. The next sub-section further explains the methods used for generating these images.

3.2.1. Image preparation To make the comparison with map-like visualisation images consistent, the geographic map images were manipulated for the purposes of our experiment. Firstly, since not all map-like visualisation algorithms generate text labels in the output, we removed the text labels in all visualisation images (V1-V4 in Fig. 2). Similarly, we removed the labels from the geographic maps (R1-R4 in Fig. 2) for a fair comparison. Additionally, as many people are familiar with maps of the whole world, the five continents and some well-known countries, this may introduce biases when conducting the experiment. Therefore, we selected maps of countries that are further away from our own location and thus less likely to be recognisable by our participants, and we rotated the images to reduce the effect of prior knowledge. The colouring and the composition of regions were also unified in both groups of images for consistency. Whereas the geographic maps had different colours in their original version (to show administrative divisions), we adopted the Qualitative Colour Scheme [25] and repainted the images to eliminate these differences in colouring. As demonstrated in Fig. 2, we also constrained the details of display to three levels of hierarchy on both categories of images, i.e. country, province, district in the geographic maps, and three tiers of data for the map-like visualisations. 3.2.2. Survey design We conducted the first experiment as an online survey consisting of four parts: entry questionnaire, training, evaluation, explanation. The time required to complete this survey was less than ten minutes. The different parts of the survey are explained below. The first part, the entry questionnaire, included demographic questions about the participant: • Age and gender • Degree currently being studied for (bachelor, master, Ph.D. or non-student) • Faculty/unit in the university • IT skills with three options: basic (knowing how to use office software and Internet); advanced (knowing how to install software); programming (knowing how to write computer programs) The second part, training, showed two images, one at a time, to the participant and asked the single question “Does this look

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

like a map?”, with possible answers for each question being “yes”, “no” and “unsure” . One of these images was a geographic map, the other was a map-like visualisation image. Responses to these two questions were discarded. The purpose of these questions was to let the participant learn what questions are to come, and thereby be prepared for the “real” questions. Participants did not know that these two questions were for training as there was no such information, and the flow from the training part to the evaluation part was seamless. In the third part, evaluation, eight images (as shown in Fig. 2) were presented to the participant, one after another in a randomised order, again with the single question: “Does this look like a map?” and the same three possible answers as during training. These images included four images generated from real geographic maps, as well as four map-like visualisation images, so that we have a balanced number in each category of images. Participants were asked to respond to questions quickly without thinking too deeply, thereby trying to capture their first impression, and were specifically instructed not to search for any hints on the Internet. The fourth part, explanation, showed the same eight images from the evaluation part, again one at a time, together with the question “Please explain why you answered (yes, no or unsure) for this image.”, with the answer shown being that which the participant gave during evaluation. A multi-line text box was provided for capturing qualitative feedback. 3.2.3. Recruitment We sent out invitation emails to all students at two of our university’s residential colleges, and they were allowed to forward the invitation to other people. Residential colleges have students from across all our university’s faculties, thus providing a representative sample of students from different disciplinary backgrounds. Combined these two colleges have about 900 students. Additionally, we posted the information about the survey on Facebook. The messages that we sent and posted asked recipients to help in our research by clicking on a link that would take them to an online survey, and that the survey would only take a few minutes of their time. Whereas this is an experiment related to reading maps, none of the students recruited for our experiment were geography majors (and our university does not even have a geography major). Until now we have briefly introduced the structure of the first experiment. In the next subsection we continue by introducing the design of the second experiment, which is a follow-up study of the first one. 3.3. Design of the second experiment The second experiment focused exclusively on one visualisation algorithm, namely the one that received the highest evaluation score in the first experiment, meaning its images were considered to look most like maps. Participants were asked to evaluate several visualisation images which were generated using different parameters. This experiment was scheduled after the data analysis stage of the first experiment, and we used the results from the first one to inform our research design. The goal of the second experiment was to determine the suitable value for each visualisation parameter such that the resulting image would contribute to the greatest extent to the perception of the map metaphor. We displayed three sets of images to the participants, with each set consisting of images generated with three different levels of the parameter evaluated. In the next subsection we continue to explain these images. 3.3.1. Images used In the second experiment, we used three sets of map-like visualisation images generated from the EHTA algorithm, which the

participants of our first experiment considered to look most map alike. The first two sets of images (Set A and B) differed in their border styles, while the last set (Set C) adopted different ways of displaying text labels. Fig. 3 shows the first two sets of images used for testing. We applied two different border styles in these two sets of images. Images A1 to A3 used curved border lines as boundaries of regions, whereas straight border lines were used in images B1 to B3. In each of these two sets of images, we varied a smoothing parameter which decreased from left to right. Thus the leftmost images (A1 and B1) have relatively smooth border lines, then the middle images (A2 and B2) have somewhat rougher border lines, and in the rightmost images (A3 and B3) the border lines are most rough. Fig. 4 displays another set (Set C) of images used in the second experiment, which used different methods in displaying text labels. In this image set, the first image (C1) contained no text labels at all, the second image (C2) showed a single label representing the entire large region, and in the last image (C3) individual subregions contained text labels. This setup allows us to assess the effect of text labels in perceiving the map metaphor.

3.3.2. Survey design The second experiment was also conducted using an online survey, similar to the implementation of the first experiment. However, this second experiment was conducted in a controlled setting. Our participants were invited to a computer room for attending a brief introductory section before the actual experiment. Researchers explained some basic concepts of map-like visualisations and briefly demonstrated the techniques of reading the visualisations. After the introductory section, participants were requested to use the computers provided to finish the survey. Regarding the content of the survey, we first presented participants with the demographic questions that were explained in Section 3.2.2, which covered age, gender, degree, and faculty studied at. Additionally, we asked for their self-reported IT skill and knowledge levels about information visualisation with a 7-point Likert scale for answers to these two questions, ranging from 1 (Not at All) to 7 (Expert). After the demographic section, the survey displayed three sets of images (Figs. 3 and 4) in sequence. With each image set, the three images were shown next to each other, and we asked a single question “How similar do the images look like a geographic map?”. The participants were allowed to answer on a 7-point Likert scale, ranging from 1 (Very Different) to 7 (Very Similar). Finally, we requested participants to explain why they chose a particular answer. They were also allowed to enter comments and feedback about these map-like visualisation images to the online survey system.

3.3.3. Recruitment Participants were recruited online in December 2016 from among our university’s students through various undergraduate and postgraduate student organisations. This constituted a wide cross-section of students across faculties, majors, degrees and age. As mentioned above, no students majoring in geography were involved in this experiment. Participants received a supermarket coupon as a reward for their time and effort spent in joining our survey. Following this description of our experiment design we next report on the results of both the experiments and the qualitative feedback collected from the participants.

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

5

ARTICLE IN PRESS

JID: YJVLC 6

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

Fig. 3. Two different border styles (curved, above vs. straight, below) of map-like visualisation images used in the second experiment. The amount of smoothing decreases from left to right.

Fig. 4. Map-like visualisation images with different text label styles used in the second experiment.

4. Results 4.1. Quantitative results of the first experiment Totally 75 participants (with 36 male and 39 female) completed the first experiment. As the residential colleges of our university mostly house undergraduate students, the majority of participants were studying for a bachelor’s degree. Thanks to additional participant recruitment on social media platforms, we received feedback from participants with other backgrounds. Their age ranged from 18 to 26 with an average of 20. A majority of them reported to have basic or advanced computer skills. Detailed demographic results are listed in Tables 1, 2 and 3. For the main survey question “does it look like a map?”, we recorded different patterns of responses from both categories (actual geographic maps vs. map-like visualisations) of images (Fig. 2). Table 4 lists the numbers and the ratios of participants’ choices. For geographic maps, on average a high percentage of our participants agreed that they looked like maps, ranging from 80% to 89%. Conversely fewer people agreed with the questions for maplike visualisation images, particularly for V1-V3. At most only 15%

Table 1 Degrees for which the participants of the first experiment were studying (N = 75). Degree

Count

Percentage

Bachelor Master Non-student

67 7 1

89.3% 9.3% 1.3%

Table 2 Faculties/units from which the participants of the first experiment originated (N = 75). Faculty/Unit

Count

Percentage

Arts and Humanities Business Administration Education Social Sciences Science and Technology Others

14 21 4 14 20 2

18.7% 28.0% 5.3% 18.7% 26.7% 2.6%

of participants thought V1-V3 were maps, and 68%−76% of them even explicitly declined the map metaphor. More than half of them

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

ARTICLE IN PRESS

JID: YJVLC

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

7

Table 3 Self-reported computer skill levels of the participants in the first experiment (N = 75). IT skill level

Count

Percentage

Basic (knowing how to use office software and Internet) Advanced (knowing how to install software) Programming (knowing how to write computer programs)

20 48 7

26.7% 64.0% 9.3%

Table 4 Responses of question “does it look like a map?” In the first experiment (N = 75).

Table 7 Similarity to a geographic map for images with a curved border style.

ID

Description

Yes (%)

No (%)

Unsure (%)

Image

Mean

SD

Median

R1 R2 R3 R4 V1 V2 V3 V4

Latvia Kazakhstan Ecuador Bosnia & Herzegovina EPEA HTA PEA EHTA

60 (80%) 67 (89%) 65 (87%) 66 (88%) 9 (12%) 10 (13%) 11 (15%) 45 (60%)

7 (9%) 6 (8%) 5 (7%) 6 (8%) 57 (76%) 54 (72%) 51 (68%) 20 (27%)

8 (11%) 2 (3%) 5 (7%) 3 (4%) 9 (12%) 11 (15%) 13 (17%) 10 (13%)

A1 – Lowest roughness level A2 – Moderate roughness level A3 – Highest roughness level

3.7 4.4 5.2

1.5 1.4 1.6

4 4 6

Table 5 Degrees for which the participants of the second experiment were studying (N = 40). Degree

Count

Percentage

Bachelor Master PhD

14 21 5

35.0% 52.5% 12.5%

Table 6 Faculties/units from which the participants of the second experiment originated (N = 40). Faculty/Unit

Count

Percentage

Chinese Medical Sciences Arts and Humanities Business Administration Education Law Social Sciences Science and Technology

2 6 6 1 4 7 14

5.0% 15.0% 15.0% 2.5% 10.0% 17.5% 35.0%

accepted V4 as a map, which could lead to a comparison of these results between different visualisation methods. 4.2. Quantitative results of the second experiment For the second experiment, 40 participants (with 18 male and 22 female) completed the evaluation of the map-like visualisation images. The ages of these participants ranged from 17 to 28 and the average was 22.6 years. Regarding the composition of participants, we recruited a diversity of students across different degrees and disciplines. The demographic figures are listed in Tables 5 and 6. In terms of IT skills and knowledge levels about information visualisation, we observe that our participants represent a group of moderately knowledgeable computer users and a cohort of relatively novice information visualisation users. For the self-reported IT skills levels, the average value of the responses was 5.2 (SD = 1.4, min = 3, max = 7); whereas the average value of the information visualisation knowledge levels was 3.6 (SD = 1.3, min = 1, max = 6). 4.2.1. Comparisons among border styles Table 7 shows the descriptive statistics of the responses to the question “how similar do the images look like a geographic map?” for the curved border style (Images A1-A3 in Fig. 3). The results

Table 8 Similarity to a geographic map for images with a straight border style. Image

Mean

SD

Median

B1 – Lowest roughness level B2 – Moderate roughness level B3 – Highest roughness level

4.0 4.4 2.7

1.8 1.8 1.5

4 5 2

Table 9 Similarity to a geographic map for images with different text labels. Image

Mean

SD

Median

C1 – No labels C2 – Single label C3 – Individual region labels

3.6 4.3 5.8

1.6 1.7 1.5

4 5 6

suggest that the image with highest roughness level (i.e. with the least amount of smoothing) was considered to look most like a geographic map, with an average of 5.2. Wilcoxon signed rank test (Wilcoxon, 1945) demonstrates that the differences of A1 & A2 (Z = 23.0, p < .001) and A2 & A3 (Z = 113.5, p < .005) are statistically significant. Table 8 presents the descriptive statistics of the results about images with a straight-line border style. As illustrated in Table 8, we find that the image with a moderate roughness level, i.e. a moderate amount of smoothing (B2 in Fig. 3), was considered the most realistic one when comparing to a geographic map, with an average of 4.4. After sorting by their mean values, Wilcoxon signed rank test shows that the differences of B3 and B1 (Z = 38.0, p < .001) and the differences of B1 & B2 (Z = 89.0, p < .05) are statistically significant. By comparing the best visualisation images from both groups with different border styles, i.e. A3 and B2, we observe that A3 outperformed B2 with a statistical significance (Z = 47.5, p < .005). Therefore, we conclude that the curved border style is more realistic than the straight one as reported by our sample. 4.2.2. The effect of text labels With the image set C (Images C1–C3 in Fig. 3), we are able to study the effect of the text labels in map-like visualisations, which we could not test in the first experiment due to the limitations of the different map-like visualisation implementations. Table 9 presents the comparison of similarity to a geographic map with different rendering methods of text labels. The results reveal that the images with individual region labels were found to look most similar to a geographic map. Statistical tests validate that there are substantial differences among C1 & C2 (Z = 37.5, p < .001) and C2 & C3 (Z = 9.5, p < .001).

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC 8

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

4.3. Qualitative feedback This subsection presents the qualitative feedback collected in both experiments. Qualitative feedback from our participants was processed and analysed in two stages. First, all non-English responses were translated to English for further analysis (English responses were left unchanged, including spelling and grammar errors made by our non-native English speaking participants). Then, all responses were reviewed and iteratively reduced to a number of codes [44]. Due to the web-based surveying environment in the first experiment, it was not possible to guarantee all responses were complete and meaningful. A small number of responses were excluded from analysis for this reason. After the analysis of the comments of participants, we synthesised four categories of codes indicating the reasons of perceiving or declining the map metaphor. The following subsections report on these findings respectively. 4.3.1. Outline and border This category of responses suggested that the outlines and the borders in the images had a strong effect in the cognitive process. Too round and smooth lines neither look natural nor similar to geographic maps, and thus made the images distinct from maps. On the other hand, clear and rough borders were more similar to ones seen in actual geographic maps. Some of the representative quotes include: • • • • •

“The edges are irregular, similar to how maps are (looked like).” “Border has a clear distinction and different colouring.” “The outline is too round, not so natural.” “The boundary of districts is too smooth.” “The outline is too simple which doesn’t seem to be realistic enough for affordance to people to think that it is a map.”

4.3.2. Shape and appearance Apart from the outlines, the shape and appearance of regions were criteria used to judge whether the images looked like maps. Participants proposed that regular shapes (e.g. circles and rectangles) contributed to the dissimilarity between visualisations and maps. For example, some participants said: • “The shape is more alike the real map.” • “The countries will not be round in maps.” • “Natural landscape should not have regular shapes.” Based on the feedback of participants, we identified a process of matching when they answered questions in the survey. They compared the images displayed with things they had seen in their real lives. Once they found a match in their minds to physical objects other than maps, the perception of maps was disregarded. For instance: • “Looks like a part of the brain, not like a map.” • “It looks like cells or human organs.” • “It feels like a colour palette.” Participants expected administrative regions shown in a map to have different sizes as found in some countries in the world; therefore they considered this aspect while reading the map-like visualisations. Too large regions and regions of too similar size were regarded as unrealistic for the map-like visualisation images. Some relevant quotes are listed below: • “Each coloured region has different shape and size, similar to a map of countries.” • “The sizes of regions are too even. Real maps will not have this.” • “Each of these blocks are too large. They should have more variety in size.”

Fig. 5. Elements of a map-like visualisation.

4.3.3. Correspondence to the real world Some participants tried to find a correspondence between the images shown and physical locations (such as a country or a continent) in the real world. They agreed that the visualisation looked similar to a map if they recalled a place that had a comparable outline or appearance. On the other hand, this factor was also used to disagree with the map metaphor. For example: • “It makes me think of Europe.” • “It looks a bit like the map of Russia.” • The blue part is very similar as how Norway looks in the actual map.” • “The first two images have sharp angles and some lines that look like some North African Countries’ maps.” • “According to my geography knowledge, I haven’t seen maps with shapes like these.” 4.3.4. Map elements A number of participants expected to see elements and symbols that we often find in maps. Examples include blue lines for rivers, grey lines for roads, dots for cities, etc. They reported that the lack of these elements made the images look unusual, as suggested below: • “Each of these colour blocks have grey lines within, just like rivers in maps.” • “Maps should have symbols of rivers, roads, etc. Now it doesn’t.” • “The first (image) doesn’t have any words on map so (it does) not like geographic (maps).” This section reports on the results collected from both experiments. Based on these results, we realise that EHTA is the algorithm that produces images that look most similar to geographic maps, and different border styles and text labels have an effect on the perception of the map metaphor. In the next section, we elaborate on these results and further conceptualise these findings. 5. Discussion In this section, we further argue our recommendations for the design of realistic map-like visualisations as a part of the spatialisation process [39,40]. Before the actual discussion, we introduce our bottom-up framework to systematically describe design elements for an abstraction of map-like visualisations (Fig. 5). Regions are the smallest unit in this abstraction. One or more regions are combined to form a larger connected area, known as a continent.

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

9

Finally, a map consists of one or more continents, and provides a context to users for the perception of the map metaphor. We discuss our recommendations for each design element accordingly.

not displayed adjacent to each other. Readers will not be as likely to perceive the even sizes at first sight if they are laid out in this manner.

5.1. The micro-level: region

5.3. The top level: context

Regions (as shown in Fig. 5) are the lowest level of the design elements in map-like visualisations. As they are the major components in the visualisation image, their design, including their boundaries and shapes, is important because their appearance conveys an instant overview to the reader. This overview forms the intuition or feeling that causes the image to be interpreted as a map. The findings from our experiment suggest that boundaries and lines of regions should use round and smooth border lines. However, the overall pattern of the map-like visualisation output should introduce some roughness, instead of using simple lines or curves. By looking at a world map, we can observe that coastlines and boundaries are often in a rough form. This finding matches the cartography literature which highlights the importance of the contour and the convexity when making maps [24]. As such, designers can utilise these characteristics to construct map-like visualisation images that resemble geographic maps. Moreover, designers should avoid regular shapes such as circles and rectangles when constructing map-like visualisations given that the shapes of countries in the world have an irregular appearance, which our experiment subjects also mentioned. As such, we suggest employing some constraints to control the shapes of regions when generating maps, as in the algorithms of Mashima et al. [27] and Yang and Biuk-Aghai [42]. From a technical perspective, a hexagon-based layout can be used effectively to create a zigzag contour, as adopted by Skupin [37,38]. Yang and Biuk-Aghai take this further by introducing some constraints on the selection of hexagons when tiling areas in a hexagonal lattice [42]. Their use of an estimation function, together with some heuristics, results in compact shapes with few or no undesirable features such as holes.

We have found that providing a context is helpful to readers to understand their tasks of interpreting a visualisation image as a map. This is consistent with the HCI concept called “affordance”, which refers to providing cues to users so that they perceive how to interact with an object [28]. There are a number of ways to achieve better affordance in map-like visualisations. An explicit method is to inform readers that they are using a map-like visualisation, for example, by using titles or on-screen help. On the other hand, the user interface can make use of common visual hints in ordinary geographic software, such as the magnifier icon for zooming, crosshair cursor for panning, scale indicators, etc. to implicitly suggest the visualisation can be navigated and explored just like a map. Another approach for enhancing the map context is to include more map elements, i.e. symbols frequently used in geographic maps, into a visualisation. The advantage of this approach is twofold. Elements found in geographic maps, such as roads and highways, streams and rivers, cities and points of interest, etc., can create a correspondence to additional data attributes, and thus increase the dimensions of information that an image conveys. Additionally, as suggested by our survey participants, these elements can enrich the context of a map-like visualisation so that readers perceive the map metaphor more readily. Colours, while not specifically mentioned by our experiment subjects, have significant value in building a context of the map metaphor. Though we used the Qualitative Colour Scheme in our sample images for distinguishing regions clearly, many map-like visualisation methods have applied changes of hue and tone to represent an extra data dimension [19,30], which resembles the colouring schemes used in topographic maps. In this way, the colouring helps to create a linkage between both during cognition, and strengthens the sense of context to the reader. As reported by Skupin [38], the colour scheme effectively conveys the cartographical metaphor.

5.2. The macro-level: continent The layout strategy of map-like visualisations plays an important role in the overall appearance. It determines the degree of similarity when comparing with geographic maps. As illustrated in Fig. 5, as individual regions are combined to become a continent, the outline of the entire image is not only determined by the appearance of these smaller regions, but also by the layout strategy. An analogue to this process in cartography is to compose a world map using a bottom-up manner, i.e. from districts, provinces and countries, going up level by level. Designers should avoid layout strategies resulting in a continent with regular patterns. For example, some algorithms use a radial layout (e.g. [2]), which makes a region grow outwards from the centre and merge into other regions (such as image V3 in Fig. 2). Consequently the entire visualisation demonstrates a circlelike pattern. This is unfavourable for the map metaphor as reported by our participants. To address this issue, map-like visualisation algorithms can adopt other layout methods or limit the aspect ratio to avoid creating round areas. The sizes of regions are found to have an effect on the impression of map-like visualisation images. As reported by some of the participants, a continent became unrealistic when the sizes of regions within it were too even. This is often hard to control because the size often represents a data attribute, e.g. the size of a region may correspond to the number of files in a folder, when visualising a file system. In this case, we suggest an alternative by moving apart regions with similar size, so that these regions are

6. Limitations We acknowledge that limitations exist in this study. For the first experiment, we included only four map-like visualisation methods in the evaluation. If we could provide a wide range of visualisation images, the participants may report more feedback in addition to our findings, since other visualisations may give them different impressions. Additionally, we did not assess the map reading proficiency of our participants, as their capabilities of reading maps may affect the outcome. In the second experiment, the method of conducting the online surveys may limit participants to freely express their thoughts, because only on-screen inputs are available and they cannot provide verbal and rich feedback. For an exploratory study like ours, qualitative interviews or focus groups may gain more insights from the user cohort. 7. Conclusions As the use of novel map-like visualisations is getting more popular, there is little discussion on their design, particularly in making them as realistic as possible. We attempt to fill this research gap by making a contribution in the area of visualisation design. With this background, we conducted two experiments to investigate how people perceive and interpret map-like visualisation im-

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC 10

ARTICLE IN PRESS

[m5G;October 9, 2017;8:28]

P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

ages, with the structures and the results of these experiments presented in this paper. Among the four map-like visualisation algorithms tested in our studies, EHTA is the algorithm that produces the most realistic map-like visualisations. In addition, based on our findings, we have proposed the three-level abstraction to describe the elements in map-like visualisations. Furthermore, we have proposed a threelevel framework to represent the components in a map-like visualisation and have suggested recommendations for achieving the resemblance to geographic maps at different levels. Our work will make map-like visualisations more intuitive and self-explanatory so that users can have a similar experience as if using a map, which facilitated the perception and exploration of information. However, the additional constraints on the appearance of the visualisation may bring difficulties in the implementation. This is an area for future research to address. Whereas our experiments were carried out using four visualisation algorithms, the outcomes are consistent with existing cartography and HCI literature. As such, we believe that our work has a great potential to be applied to a broader range of maplike visualisations. In conclusion, the contributions of this article have pointed out the path for using the map metaphor in a way that is favourable to lay-users, which will expand the use of the map metaphor to applications in other disciplines. Based on these findings, our future work will focus on implementing our recommendations in various map-like visualisation algorithms, and testing their effectiveness of conveying non-spatial data.

Acknowledgement We gratefully acknowledge the support for this research project from the University of Macau under grant number MYRG201400172-FST.

References [1] D. Auber, C. Huet, A. Lambert, B. Renoust, A. Sallaberry, A. Saulnier, GosperMap: using a gosper curve for laying out hierarchical data, IEEE Trans. Vis. Comput. Graph. (2013). https://doi.org/10.1109/TVCG.2013.91. [2] R.P. Biuk-Aghai, W.H. Ao, A novel map-based visualization method based on liquid modelling, in: Proceedings of the 6th International Symposium on Visual Information Communication and Interaction, New York, NY, USA, ACM, 2013, pp. 97–104. https://doi.org/10.1145/2493102.2493114. [3] R.P. Biuk-Aghai, C.-I. Pang, F.H.H. Cheang, Visualization of large category hierarchies, in: Proceedings of the 2011 Visual Information Communication - International Symposium, New York, NY, USA, ACM, 2011 (p. 2:1–2:10). https: //doi.org/10.1145/2016656.2016658. [4] R.P. Biuk-Aghai, C.-I. Pang, Y.-W. Si, Visualizing large-scale human collaboration in Wikipedia, Future Gener. Comput. Syst. 31 (2014) 120–133. https://doi.org/ http://dx.doi.org/10.1016/j.future.2013.04.001. [5] R.P. Biuk-Aghai, P.C.-I. Pang, B. Pang, Map-like visualisations vs. treemaps: an experimental comparison, in: Proceedings of the 10th International Symposium on Visual Information Communication and Interaction, New York, NY, USA, ACM, 2017, pp. 113–120. https://doi.org/10.1145/3105971.3105976. [6] R.P. Biuk-Aghai, M. Yang, P.C.-I. Pang, W.H. Ao, S. Fong, Y.-W. Si, A map-like visualisation method based on liquid modelling, J. Vis. Lang. Comput. 31 (Part A) (2015) 87–103. https://doi.org/http://dx.doi.org/10.1016/j.jvlc.2015.10.003. [7] M. Blades, J.M. Blaut, Z. Darvizeh, S. Elguea, S. Sowden, D. Soni, …, D. Uttal, A cross-cultural study of young children’s mapping abilities, Trans. Inst. Br. Geograph. 23 (2) (1998) 269–277. https://doi.org/10.1111/j.0 020-2754.1998.0 0269. x. [8] K. Börner, Atlas of Science, MIT Press, 2010. [9] K. Börner, C. Chen, K.W. Boyack, Visualizing knowledge domains, Ann. Rev. Inf. Sci. Technol. 37 (1) (2003) 179–255. https://doi.org/10.1002/aris.1440370106. [10] A. Bruggmann, M.M. Salvini, S.I. Fabrikant, Cartograms of self-organizing maps to explore user-generated content, in: Proceedings of the 26th International Cartographic Conference, Dresden, Germany, International Cartographic Association, 2013. [11] R. Buchanan, Wicked problems in design thinking, Des. Issues 8 (2) (1992) 5– 21. https://doi.org/10.2307/1511637.

[12] Card, S.K., Moran, T.P., & Newell, A. (1983). The Psychology of HumanComputer Interaction. [13] A. Celentano, F. Pittarello, From real to metaphoric maps: cartography as a visual language for organizing and sharing knowledge, J. Vis. Lang. Comput. 23 (2) (2012) 63–77. https://doi.org/http://dx.doi.org/10.1016/j.jvlc.2011.11.004. [14] H. Couclelis, Worlds of information: the geographic metaphor in the visualization of complex information, Cartography Geograph. Inf. Syst. 25 (4) (1998) 209–220. https://doi.org/10.1559/152304098782383034. [15] Encyclopædia Britannica, Map, In Encyclopædia Britannica, 2015. [16] D.M. Giannangelo, B.M. Frazee, Map reading proficiency of elementary educators, J. Geogr. 76 (2) (1977) 63–65. https://doi.org/10.1080/ 00221347708980884. [17] P. Gough, T. Bednarz, C. de Bérigny, J. Roberts, A process for non-expert user visualization design, in: Proceedings of the 28th Australian Conference on Computer-Human Interaction, New York, NY, USA, ACM, 2016, pp. 247–251. https://doi.org/10.1145/3010915.3010996. [18] P. Gough, C.D.B. Wall, T. Bednarz, Affective and effective visualisation: communicating science to non-expert users, in: IEEE Pacific Visualization Symposium, 2014, pp. 335–339. https://doi.org/10.1109/PacificVis.2014.39. [19] M. Gronemann, M. Jünger, Drawing Clustered Graphs as Topographic Maps, in: W. Didimo, M. Patrignani (Eds.), Lecture Notes in Computer Science, 7704, Springer Berlin Heidelberg, Berlin, Heidelberg, 2013, pp. 426–438. https://doi. org/10.1007/978- 3- 642- 36763- 2_38. [20] M. Kitajima, M. Toyota, Decision-making and action selection in two minds: an analysis based on Model Human Processor with Realtime Constraints (MHP/RT), Biol. Inspired Cognit. Archit. 5 (2013) 82–93. https://doi.org/http: //dx.doi.org/10.1016/j.bica.2013.05.003. [21] P. Knees, M. Schedl, T. Pohle, G. Widmer, Exploring music collections in virtual landscapes, IEEE. Multimedia (2007). https://doi.org/10.1109/MMUL.2007.48. [22] S. Lee, S.H. Kim, Y.H. Hung, H. Lam, Y.A. Kang, J.S. Yi, How do people make sense of unfamiliar visualizations? A grounded model of novice’s information visualization sensemaking, IEEE Trans. Vis. Comput. Graph. (2016). https://doi. org/10.1109/TVCG.2015.2467195. [23] Y. Liu, R. Feyen, O. Tsimhoni, Queueing Network-Model Human Processor (QNMHP): a computational architecture for multitask performance in humanmachine systems, ACM Trans. Comput. Hum. Interact. 13 (1) (2006) 37–70. https://doi.org/10.1145/1143518.1143520. [24] A. MacEachren, How Maps Work: Representation, Visualization, and Design, Guilford Press, New York, 1995. [25] Mapbox, Tips for Using Color in Maps, 2015 Retrieved January 30, 2015, from https://www.mapbox.com/tilemill/docs/guides/tips- for- color/. [26] D. Marr, Vision: A Computational Investigation into the Human Representation and Processing of Visual Information, 1, WH Freeman, San Francisco, 1982. [27] D. Mashima, S. Kobourov, Yifan Hu, Visualizing dynamic data with maps, IEEE Trans. Vis. Comput. Graph. 18 (9) (2012) 1424–1437. https://doi.org/10.1109/ TVCG.2011.288. [28] D.A. Norman, The Design of Everyday Things, Basic Books, Inc, New York, NY, USA, 2002. [29] Oxford English Dictionary, map, n.1, Oxford English Dictionary, Oxford University Press, 2015. [30] C.-I. Pang, R.P. Biuk-Aghai, Wikipedia world map: method and application of map-like wiki visualization, in: Proceedings of the 7th International Symposium on Wikis and Open Collaboration, New York, NY, USA, ACM, 2011, pp. 124–133. https://doi.org/10.1145/2038558.2038579. [31] C.-I. Pang, S.-S. Si, S.-K. Chio, A new way to use wikipedia in education: a pilot study of map-like Wikipedia visualization on iPad, Int. J. Future Comput. Commun. 3 (1) (2014) 45–49. [32] P.C.-I. Pang, R.P. Biuk-Aghai, M. Yang, What makes you think this is a map? Suggestions for creating map-like visualisations, in: Proceedings of the 9th International Symposium on Visual Information Communication and Interaction, New York, NY, USA, ACM, 2016, pp. 75–82. https://doi.org/10.1145/2968220. 2968239. [33] S. Pinker, A theory of graph comprehension, Artif. Intell. Future Test. (1990) 73–126. [34] P. Pirolli, S. Card, Information foraging, Psychol. Rev. 106 (4) (1999) 643. [35] A. Quispel, A. Maes, Would you prefer pie or cupcakes? Preferences for data visualization designs of professionals and laypeople in graphic design, J. Vis. Lang. Comput. 25 (2) (2014) 107–116. https://doi.org/http://dx.doi.org/10.1016/ j.jvlc.2013.11.007. [36] Y. Rogers, H. Sharp, J. Preece, Interaction Design: Beyond Human-Computer Interaction, third ed., John Wiley & Sons Ltd, West Sussex, UK, 2011. [37] A. Skupin, A picture from a thousand words [information visualization], Comput. Sci. Eng. 6 (2004) 84–88. [38] A. Skupin, The world of geography: visualizing a knowledge domain with cartographic means, Proc. Natl. Acad. Sci. 101 (Suppl. 1) (2004) 5274–5278. https://doi.org/10.1073/pnas.0307654100. [39] A. Skupin, S.I. Fabrikant, Spatialization methods: a cartographic research agenda for non-geographic information visualization, Cartography Geogr. Inf. Sci. 30 (2) (2003) 99–119. [40] A. Skupin, S.I. Fabrikant, Spatialization, in: The Handbook of Geographical Information Science, 2007, pp. 61–79. [41] R. Spence, Information Visualization: Design for Interaction, 2007 Retrieved from http://www.amazon.com/Information- Visualization- DesignInteraction-2nd/dp/0132065509 .

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002

JID: YJVLC

ARTICLE IN PRESS P.C.-I. Pang et al. / Journal of Visual Languages and Computing 000 (2017) 1–11

[42] M. Yang, R.P. Biuk-Aghai, Enhanced hexagon-tiling algorithm for map-like information visualisation, in: Proceedings of the 8th International Symposium on Visual Information Communication and Interaction, New York, NY, USA, ACM, 2015, pp. 137–142. https://doi.org/10.1145/2801040.2801056.

[m5G;October 9, 2017;8:28] 11

[43] F.B. Zhan, Atlas of science: visualizing what we know, J. Geogr. 112 (5) (2013) 229. [44] J.W. Creswell, Educational Research: Planning, Conducting, and Evaluating Quantitative and Qualitative Research (4th Ed), Merrill Prentice Hall, Upper Saddle River, NJ, 2002.

Please cite this article as: P.C.-I. Pang et al., Creating realistic map-like visualisations: Results from user studies, Journal of Visual Languages and Computing (2017), https://doi.org/10.1016/j.jvlc.2017.09.002