An Alternative Representation of The Visualization of
Pedometer Sensor Data
!
!
SUBMITTED IN PARTIAL FULLFILLMENT FOR THE DEGREE OF MASTER
OF SCIENCE
N
IKKI
W
INANDS
6088651
M
ASTER
I
NFORMATION
S
TUDIES
H
UMAN-
C
ENTERED
M
ULTIMEDIA
F
ACULTY OF
S
CIENCE
U
NIVERSITY OF
A
MSTERDAM
July 6, 2015
Supervisor: Dr. Daniel Buzzo
FNWI, University of Amsterdam
Second Reader: Dr. Frank Nack
An Alternative Representation of The Visualization of
Pedometer Sensor Data
Nikki Winands
MSc Information Studies: Human Centered Multimedia
University of Amsterdam Amsterdam, The Netherlands
nikki.winands@student.uva.nl
ABSTRACT
With the availability of and increased attention to smart-phone connected wearables, people are able to collect infor-mation about their physical activities. This study explored alternative representations of step count data to improve the efficiency of understanding the physical activity. More specifically, it aims to provide a first iteration on improving the process of understanding pedometer sensor data. We sketched out alternatives for current visualizations of step count data: spatial representation, representation in body size, and representation in food. An interview-based ques-tionnaire has been conducted to explore participants’ un-derstanding of and experience with these alternatives. The results show that a representation in food is best under-stood, followed by a spatial representation. We discovered that relatedness to tangible objects plays an important role in representing step count data.
Categories and Subject Descriptors
H.5.3 [Information interfaces and presentation (e.g., HCI)]: Miscellaneous—Design
General Terms
Design, Human Factors
Keywords
Information visualization, user interface design, exploratory design, pedometer sensor data, meaning-making
1.
INTRODUCTION
Recently, quantified self-tracking applications receive in-creased attention with the availability of wearables connected to mobile applications, such as the Apple Watch1
, Jawbone UP2
, and the Fitbit3
. According to PwC[1] there is a wear-able future ahead, that can alter the landscape of society and business. By using wearables, open data, and visual-izations, based on sensor technologies, one is able to collect quantified-self data. These personal informatics help people to collect information about themselves, such as behaviour, health state, and performance.
1 http://www.apple.com/watch/ 2 http://jawbone.com/up 3 http://www.fitbit.com/
1.1
Pedometer Sensor Data
Pedometer sensor data is part of the quantified-self data. The sensor data is translated into step counts of the user. The measured data comes from the sensor within the mobile device, not specifically from the user itself. Pedometer sen-sor data can be retrieved from physical activity wearables in the form of bracelets, clips, and smartwatches. Smart-phones that include sensors such as GPS, gyroscope and an accelerometer can be used for pedometer sensor data [2].
Physical activity wearables and mobile devices display the measured sensor data to the user, by using a mobile appli-cation. The visualization of measured data can be used to translate data to knowledge, which may be relevant and im-portant to the user. Preliminary study shows that current visualizations of physical activity data fall short in describ-ing the meandescrib-ing of the data, in terms of understanddescrib-ing the data by the user. Despite this, according to Li et al. [3] there is no guidance for making these systems, dashboards, and visualization, more effective. Users want to receive in-formation about themselves to reflect on, and systems that support this activity need to be efficient and easy to under-stand. Simply displaying data is insufficient for gaining this knowledge.
However, because of poor design implementations, most dashboards fail to communicate effectively and efficiently, and lack transparency and immediacy [4]. To gain knowl-edge one needs clear and instructive communication that is focused on the user its needs, interests, and the decisions they have to make [5].
To date, recent searches in the the Apple’s App Store4 show that there are over more than 500 pedometer related applications, and an equivalent number of applications in the Google Play Store5
. These applications display among other the step count of the user. Current applications show different types of visualizations with a variety of features, such as graph visualization, goals and progress, and smart coaches. Nonetheless, according to the preliminary study, the understanding of step count data is low, and there is a demand for explanation.
1.2
Domain and Research Question
The domain of this study concerns physical activity mon-itoring of the quantified self by using wearables and mobile application. As a first iteration of design work and evalua-tion, the focus lies on the visualization and representation 4
http://www.apple.com/itunes/ 5
of pedometer step count data, and the understanding of this data by the user. This study contributes to the human-computer interaction (HCI) and health domains by identi-fying a new facet of information visualization of step counts to impel the understanding of pedometer data. The present paper’s objective is to explore the design space of alterna-tive representations of pedometer sensor data, to improve the efficiency of understanding the physical activity of the user. The central research question is:
“What alternative techniques can be applied to the visualiza-tion of pedometer sensor data to improve the efficiency of understanding activity ?”
To answer this research question, we designed early proto-types of alternative techniques to represent step count data, followed by an interview-based questionnaire.
First, we will define more precisely how current applica-tions visualize pedometer sensor data, results of the prelimi-nary study, and related work that can be applied to this type of data. Then, we will concentrate on alternative techniques that are being designed and used in the user test. Next, we will discuss the results of the user test. The concluding section contains an evaluation and recommendation for the next iteration of design work.
2.
RELATED WORK
This work builds on top of other research that has looked into investigating health and physical activity applications, the visualization and consumption of information. This pa-per focuses on the users’ understanding of pedometer sensor data.
2.1
Health and Physical Activity Applications
Thousands of healthcare and physical activity applica-tions apps are now available for download to mobile devices, which can be used as part of the users’ wellness, prevention or treatment regimens. The IMS Institute for Healthcare Informatics [6] studied consumer health care apps that were available in 2013. Their study includes a comprehensive analysis of 43.689 healthcare applications, assessed on po-tential value. Aitken [6] found that the most common func-tionality of healthcare applications is providing information, approximately sixty percent of all applications, only half of them has a multi-functionality of informing and instruct-ing, and twenty percent informs and records data. Fewer applications have other features. Only five percent of the applications guide the user, and even less include reminders or alerts. Guiding the user is relevant in order to give the user a better understanding about and an explanation of measured sensor data.
It is important to take into consideration the wants and needs of the user, in order to support the effectiveness of health applications including pedometer related applications. Users of health applications are looking for information that helps them to take action and to give them insights about their health state and physical activity. Users do not want to conduct an analysis aside from the application, which takes a lot of time, to look for information or an explana-tion about their health related data [7]. The tasks and the understanding of personal informatics is the primary prob-lem according to Li [8], health and physical activity related applications should provide support. A survey of 250
pa-tients [9] specified five main requirements, from users, of health applications based on priority:
• “Give people more control over their condition, or keep them healthy”
• “Be easy to use”
• “Be able to be used regularly”
• “Allow networking with other people like them, or with people who understand them”
• “Be trustworthy”
It seems that many apps have a lack of approval or clinical testing. However, this approval ought to be made transpar-ent, to help guiding consumers [7]. Furthermore, it is not enough to have a health application that merely helps to manage or keep track of one’s health, users also desire feed-back. Additionally, a health application should be accurate, follow clinical and behavior modification guidelines [10], and be personalized [11]. This relates as well to pedometer sen-sor data.
Heffernan et al.[12] investigate the process for determin-ing the delivery of personalized health messages to target users. They highlight guidelines for message construction, along with recommendations to deliver targeted messages. They recommend among other that personal health messag-ing should be “tailored per user, based on user input, in-teractions with the application”, “calibrated and cleaned with target users to ensure understanding of the intended mes-sage”, and be “tied to gamification to reinforce learning by leveraging a heuristic to prompt for behavior when not using the app”.
Consolvo et al.[13] presented four design requirements for technologies that encourage physical activity, including the pedometer. Most important requirement, and relevant to this study, concerns giving the user proper and sufficient information about their physical activity. The pedometer application often provides deceptive measurements of phys-ical activity, however the measurement on its own does not provide sufficient information to the user. This requirement is in line with the results of the preliminary study.
2.2
Consumption of Information
Humans have known cognitive and perceptual limitations. Therefore, the amount of information one can perceive and analyze at a certain moment is limited [14]. Information vi-sualizations transform information into a visual form that makes use of the human its visual capabilities [14, 15]. It aims to reduce the complexity of understanding information for the user, to let the user engage with data, to give the user a deeper understanding of data, and encourage users to dis-cover relations and details in a visual setting [14]. This study presents alternative representations of step count data, these representations need to take into account the perceptual ca-pabilities and limitations of the user.
It is important that visualization of data should give the overall picture of the data, and it should match the common human mental model [16, 17]. The visualizations should help the user to consume information, not just be appealing. To consume information the human brain should be used as a contextual filter based on prior knowledge [18]. Madsen [18] listed guiding principles for the visualization of data. A
few of the principles are: don’t use jargon, avoid acronyms, over-label content, keep it simple, and provide benchmarks as much as possible. This also applies to current step count data overviews, which do not follow all of these principles, such as providing benchmarks.
People behave in fundamentally social ways to computers, media, as well as mobile applications. Therefore, it is im-portant to understand how “features of computer interfaces relate to the features of interaction between human users or between a human user and computer agent and then match the interface design to the desired outcomes” [19]. Addition-ally, as stated by Maracas et al. [20], if computer use enables the methods of interaction to require less cognitive effort of the user; the user is able to pay more attention to the task at hand. If step count data overviews do not provide in-teraction, it takes more cognitive effort of the user to get a (better) understanding.
3.
METHOD
This study focuses on the user its understanding of pe-dometer sensor data. Its point is to design alternative rep-resentations to understand the activity data more efficiently, and describe in detail what aspects of representation give a better understanding, compared to step count visualized by using numbers. In the first phase a preliminary study has been conducted. Its focus is on gaining insights of the user experience with mobile applications that visualize physical activity data. In the second phase, three alternative rep-resentations are designed based on the preliminary results and literature study, see section 3.3. In the third phase, a user test has been conducted to define what alternative representation can be applied to improve the efficiency of understanding activity.
3.1
Generative research
To uncover potential opportunities for novel applications and services that meet unmet needs, related to physical ac-tivity trackers, a preliminary study has been conducted. The focus of this study is on gaining more insights in the user experience with activity trackers and activity monitoring. We interviewed four participants (two male, two female) in an informal fashion. The participants are being asked to wear the Jawbone UP24 wearable for five days, and use the corresponding mobile application (see figure 1). This wear-able is a bracelet being wear-able to track automatically steps and sleep, and manually food. The corresponding mobile application includes a smart coach that provides the user textual information based on the collected data. The in-terviews are semi-structured; we made use of an interview guide that served as a checklist of topics to be covered, with default wording and order of questions. Each interview took roughly twenty minutes, is being recorded, transcribed and coded.
3.1.1
Findings
In the interviews, participants were easily able to under-stand the features and capabilities of this device, which led to in-depth conversations regarding physical activity track-ers and the measurements.
Setting and achieving goals regarding steps and sleep, is the main feature of the Jawbone UP24 application. Par-ticipants mentioned that this is a motivational and useful aspect of using this application. However, two participants
Figure 1: Two screen captures of the Jawbone UP24 mobile application showing sleep and step informa-tion. With an illustration of the Jawbone device be-ing used by the participants durbe-ing the preliminary study.
mentioned that it did not motivate them because it was not personal. Participant 1 (P1) said: “The data was not personal enough so it did not encourage me to move. Addi-tionally, when achieving a goal it just said ‘congratulations’ when I achieved my goal. And sometimes that it was above or below average”. The smart coach was found interesting, despite not personal. Participants found it interesting for a short period of time because of the health facts the smart coach provided. P3 mentioned: “[...] you have to drink eight glasses of water a day. It had nothing to do with my steps, if I had walked 10km a day or 1km”.
We asked the participants about their experience and un-derstanding regarding the dashboard, the step and sleep measurement overviews. The dashboard has been found clear and well organized by all participants, they mentioned “sleek design” and “beautiful colours”. Concerning the spe-cific step and sleep information, participants do not know what to do with the data. P3 said: “I am not a sleeping analyst or movement coach. I would like to know what is normal for me, in terms of deep and light sleep, and how I can improve it if necessary”.
The participants were asked what they would like to im-prove based on their five-day experience with this specific ac-tivity wearable. All participants agreed on that they would like to have more explanations about the meaning of their activities.
Conclusion
We found a tendency that current physical activity track-ers and their corresponding applications do not completely meet the needs of the user, and that there is a need for ex-planation about their measurement. The sample size of this preliminary study is significantly small, but helped to define early indications, to form a direction, and to structure the questionnaire and interview in the next phase. This study was excessively useful to focus and delimit the research ques-tion.
Figure 2: An illustration of high end and low end design sketches in the iterative design process of al-ternative representations of step count data.
3.2
Participants
Eighteen participants have been recruited from the Uni-versity of Amsterdam, Amsterdam UniUni-versity of Applied Sciences, friends, and relatives. All participants make use of a smartphone, and are aware of the existence of physi-cal activity trackers. Participants (nine male, nine female) are aged between 17 and 54 (28.9 average with a standard deviation of 9.403). Five participants currently use a physi-cal activity tracker. The reasoning for using activity track-ers are related to losing weight, analyzing sport activities, and to gain knowledge about their physical activity. Thir-teen participants do not use a activity tracker currently, of which three participants used an activity tracker in the past. The reasoning why the participants did not use the activ-ity tracker any longer is because it is not entertaining over a longer period of time, there is no new knowledge, and becomes uninteresting. The participants who never used a physical activity tracker state that they are not familiar with the possibilities, believe it is too much effort, are not inter-ested, or believe that going to the gym gives them sufficient knowledge.
3.3
Procedure
The study consisted of the design of early prototypes, and a face-to-face interview-based questionnaire. We used early prototypes to focus on answering the research question con-cerning an alternative representation of step count data, in order to explore the needs of the participants. The proto-types are not used in the way commercial offering would do; the prototypes are build to visualize the experience, not the technology. We designed alternative representations using an iterative, user-centered design process. In the first phase of designing representations, we took a closer look at the available applications to date, relating to pedometer data. We analysed and listed the design elements that are being used to represent the measurements, such as graphs, lists of numbers, and goal progress. Based on the results of the preliminary study, literature, and the analysis of current ap-plication, we started mind mapping and sketching out ideas that represent step count data. We ended up with roughly twenty useful exploratory design pieces (see figure 2). The design pieces resulted in three domains of representations: spatial representation (1), representation in body size (2),
Figure 3: Four screen captures of step count data overviews of mobile applications, which was shown to the participants as an example of current visual-izations, displaying factual data in graphs and lists.
and a representation in food (3). The alternative represen-tations are built by using the software tools Balsamiq6
and Adobe Photoshop CC7
.
In the following subsections, each alternative representa-tion of step count data will be further explained.
3.3.1
Interview-based questionnaire
The purpose of the interview-based questionnaire is to gain deep insights in the preference and best understanding of the user, relating to step count data. We used this method to gather quantitative and qualitative data.
The questions are structured in seven groups: demograph-ics, current use of physical activity trackers, current appli-cations, spatial representation, representation in body size, representation in food, and at last preferences. Each group contains open-ended and closed-ended questions, with a fo-cus on the opinions and attitude of the participants. The interviews are being held in face-to-face conversations, the questionnaires are printed out on paper, and images being used during the interview are displayed on a tablet to enable the participant to take a closer look at the images.
The questions about the current applications are related to the understanding and experience of the use regarding these visualizations. Additionally, we asked for a overall re-action towards the visualization, containing difficulty level, satisfaction, and interest. At the end of this group of ques-tions, we asked the participant to name positive and negative aspects. We asked similar questions for each representation. The participants are being asked about their understand-ing of and experience with the representation, includunderstand-ing a comparison with the current visualizations as shown before. Furthermore, we asked if this would meet their needs. We also asked about the overall reaction towards this represen-tation. At the end of each representation we asked them again to name positive and negative aspects.
6
https://balsamiq.com/ 7
Figure 4: Three alternative representations of step count data that are being used in the interview-based questionnaire. Left the spatial representation, middle representation in body size, and right illustrates the representation in food.
3.3.2
Current applications: Numbers
Recent searches in the Apple’s App Store resulted in over more than 500 pedometer related applications. These appli-cations display among other the step count of the user. We made a selection of four applications based on user reviews and different visualizations of step count data, see figure 3. We used screen captures of the applications as displayed in the App Store. The screen captures are shown to the par-ticipants, to introduce them with the current visualization in case they are not familiar with these visualization, and to have presentation of step count data fresh in mind. The se-lected applications show graphs, progress bars, and factual data lists.
3.3.3
Representation: Spatial
The sketches that are enclosed by the spatial representa-tion domain, resulted in one final spatial representarepresenta-tion. We used a representation of step count data that is equivalent to spatial distance. The representation is build up from two elements that play an important role of consuming informa-tion: textual and narrative explanation[5], and the use of images to match the mental model [16, 17].
We used the spatial distance of Central Park located in New York (United States of America), and the Great Wall of China (see figure 4). We assumed that participants can relate to these historic landmarks as familiar presence. If they could not relate to these places, we gave an example of an equivalent distance nearby one’s home. Our intention is to represent the step count data to the equivalent distance where one can recognize and relate to , as well as having an understanding of this distance, instead of showing a specific step count such as 13,480 steps.
3.3.4
Representation: Body size
Visuals make people pay attention, it takes half of the human brain’s resources [18]. People rely on what they see, and what the brains tell them. We designed a representation in body size to represent step count, because this is a visual only, and might match the human mental model. The in-tention is to represent a large number of steps as slim body size, and a small number of steps as overweight body size. We have no intention to present an ideal body size, only the
intention to give an understanding about step count data. Therefore, we made use of cartoon images, and not actual body images of people (see figure 4).
3.3.5
Representation: Food
Current applications combine step count measurements frequently with burned calories. However, step count data and the number of calories might have an unknown meaning to a user. To indicate if a user has a better understanding of step count data, we designed a representation in food of the equivalent number of calories. We made use of easy to recognize products: a hamburger and a bottle of Coca Cola, see figure 4. We assumed that the participants have an indication about the amount of calories of these products. We explained to the participants that these products are examples, and can also be replaced by products that are more familiar or related to the user.
4.
RESULTS
Each interview has been recorded and transcribed. We analyzed the interviews by coding the responses line by line. We identified themes from the lines of data during the anal-ysis of the interviews. Responses on the open questions re-lated to understanding of representation, meeting the needs of the participant, and naming positive and negative aspects of representations are listed per participant. The listed re-sponses of the open question related to meeting the needs of the participant, are evaluated into a Likert-scale score in order to execute a statistical analysis of this data. Closed questions are evaluated by using descriptive statistics and a Pearson chi-square test. Descriptive statistics are displayed in table one. The answers to the closed questions are based on a five point Likert scale.
The results of the user experience with current visualiza-tions and alternative representavisualiza-tions is shown in Table 1, including the mean and standard deviation. Meeting needs has been rated on a scale from 1 - to a great extent, 4 - not at all. Difficulty level has been rated on a scale from 1 - very easy, 5 - very hard. Satisfaction level has been rated from 1 - very satisfied, 5 - very dissatisfied. The perceived level of interest has been rated from 1 - extremely interesting, 5 - not at all interesting. And the perceived usefulness and
Table 1: The means and standard deviations for user experience of current applications and the alternative representations. These statistics clearly show significant differences in a positive (1) and negative (5) user experience among alterna-tive representations and the current applications (Numbers).a
Representations
Numbers Spatial Body Food
Meet needs 2.33 (1.02) 1.5 (0.51) 3.56 (0.85) 1.33 (0.49) Difficulty 2.44 (1.04) 1.78 (1.11) 2.67 (1.45) 1.56 (0.62) Satisfaction (0.92)2.56 (0.51)1.83 (0.96)3.89 (0.61)1.61 Interest 3.44 (0.98) 2.44 (0.71) 4.00 (1.24) 2.00 (0.48) Usefulness (0.92)2.83 (0.69)2.33 (0.98)3.61 (0.58)1.89 Pleasant 2.33 (1.09) 2.00 (0.91) 3.56 (1.04) 1.67 (0.485) a
Standard deviations are in parentheses.
pleasant has been rated on 1 - strongly agree, 5 - strongly disagree.
Except for two interviews, all interviews are spoken in Dutch. Quotations being used in this paper are translated to English by this author.
4.1
Current applications: Numbers
The visualization of step count data in current applica-tions meet the needs somewhat for seven out of eighteen (7/18) participants, four participants state that this visu-alization meet their needs to a great extent. Participants (4/18) mentioned that the meaning of such data is unknown. P1 said: “it does not mean that much to me, it is too profes-sional, and too much use of jargon. What is the meaning? ”. However, participants (9/18) mentioned positive statements about the current visualizations as clear and factual overview, accurate data, interesting, and useful. We specifically asked what their understanding is if one walks 13,480 steps on a day. The majority of the participants (11/18) directly in-dicated that they did not understand what this number of steps means, if it is a lot or not. Others describe this num-ber as just the amount of steps. Their understanding of the number is low, P15 indicates that “I need the average to ex-plain the exact meaning, to say if it is a lot or not”. P2 and P4 also indicate that there is no reference nor a realistic link to have an understanding about the meaning of the number of steps. Despite this, the clarity of words being used is perceived as good (15/18).
The overall reaction of the current applications is mainly perceived as easy in terms of difficulty level (10/18). It has been perceived as somewhat satisfying by the majority of the participants (10/18). Participants are slightly to mod-erately interested in this type of visualization and data rep-resentation (7/18). But it is found to be pleasant to use (12/18). Table 1 shows the means and standard deviations for the overall reaction questions. Figure 5 shows that par-ticipants stated that the factual data only meet their needs only somewhat to little.
Participants named negative aspect about the visualiza-tion of step count data as it is now in current applicavisualiza-tions. Participants (10/18) describe that there is no reference, no information to be gained, and an explanation is missing. P13 said: “It is just random data, and you need to perform some research by yourself to understand it”. It is being called “meaningless”. P2 mentioned “[...] and then what, what can you do with it, what is the goal? ”. Despite, it is positively perceived as a clear overview of numbers by eight partic-ipants. P1 and P15 found the simplicity of the overview pleasant, the use of colours give a good indication if a num-ber is positive or negative. Three participants found it useful to illustrate data with icons.
4.2
Representation: Spatial
At first glance, the participants (17/18) have a clear un-derstanding of the spatial representation of step count data. Nine participants noticed directly that the use of physical and real life objects gives a better understanding of the number of steps. P1 agreed that this makes more sense to a user. Compared to the overview of numbers in cur-rent applications, participants (9/18) stated that the spatial representation gives a better understanding. P13: “You un-derstand the measurement, if it is a lot or not, if it is far and so on. You start thinking in distances instead of num-bers”. All participants state that this representation meets their needs, somewhat (9/18) and to a great extent (9/18), see figure 5. Nine participants found this representation a good supplement to the step count measurement as being shown in current applications. P4:“It is a good addition to the numbers, in terms of comparing and it is much nicer to share”.
The overall reaction of spatial representation is perceived as very easy (9/18) and easy (7/18) in terms of difficulty level. Two participants found it somewhat hard because the information is not specific enough, and the distance numbers are missing, for example Kilometers or Miles. Participants are somewhat (13/18) to very satisfied (4/18) with this rep-resentation. The satisfaction level corresponds to the level of interest which is considered as moderately (7/18) to very interesting (9/18). The interest is reflected in the perceived usefulness, participants (12/18) found this a useful represen-tation. Table 1 shows the means and standard deviations for the overall reaction questions.
Participants found the representation as it is shown to them, too abstract (5/18). It was found as a negative as-pect that the representation is not precise enough, and dis-tance measurements are missing. The use of physical, real life objects was perceived as a positive aspect of the repre-sentation (6/18). It was found easy to understand and easy to use (8/18). P13 mentioned “This is understandable by ev-eryone, it also feels more personal instead of random data”. Others found it motivational in terms of physical activity, and entertaining. P18 said “You can see it as achievements, and gives a better understanding than just a number, it is less boring”.
4.3
Representation: Body size
The representation of body size was directly understood by the participants as an illustration of fat versus skinny (9/18), and related to losing weight (3/18). Even though participants found it a visually better to understand com-pared to the numbers, particularly at first glance it is easy
Meet needs food Meet needs body Meet needs spatial Meet needs numbers
Figure 5: The results of meeting the needs of partic-ipants towards the factual data representation and alternative representation. It shows that the spatial and food representations meet the needs of the par-ticipants significantly better than the representation in body size.
to quickly understand if one has been walking a lot of steps or not. However, it was found to be not specific enough compared to the numbers (7/18), and perceived as hard to measure and analyze. P7 said “It is a very visual way of showing steps, so yes it is easier to understand but very con-fronting”. P7 and five other participants found it likewise confronting, offensive, and negative. P16 mentioned that it was not realistic nor personal, “ [...] it does not show you that you walked more or less, people can be fat and still walk a lot. ”.
The majority of the participants (14/18) state that this representation of step count data does not meet their needs (see table 1 and figure 5). Three participants mentioned that it would meet their needs in a situation that one would like to lose weight, and that this representation is motivational and stimulating. The negative views on this representation are reflected in the overall reaction. The level of difficulty is divided as shown in table 1 among the user experience. Par-ticipants are somewhat (11/18) to very dissatisfied (4/18) with this representation, half of the participant found it not at all interesting and not useful. In line with the meaning of the participants who stated that this representation is con-fronting and offensive, they considered this as unpleasant to use.
P13 mentioned that the most negative aspect of this rep-resentation is that it is “ [...] an aggressive way of showing data”, and P14 amplified this by saying that it “ [...] can make you insecure”. P2 stated that this type of representa-tion is not measurable, there are no gradarepresenta-tions. Despite this, participants consider this representation as fun (7/18) and motivational for losing weight (7/18). P15: “It does show you like this is how I do not want to look like”. Addition-ally, P6 mentioned that it is fun, but the use of activities in
combination with body size would be better, such as “sitting in front of the TV, and running through the park ”.
4.4
Representation: Food
The understanding of the representation of step count data in food, is perceived as clear and familiar (15/18). The use of tangible objects is found positive and gives a bet-ter understanding compared to the numbers (6/18). It is found “’visually attractive, recognizable” (P8), and motiva-tional (4/18). All participants stated that this meets their needs somewhat (6/18) and to a great extent (12/18), see table 1 and figure 5. However, it is preferred to supplement the numbers such as step and calorie count to bring the food more in perspective (7/18), as P3 indicates: “[a] good combi-nation with the numbers, a nice addition, but it should give also the precise calorie count”.
The overall reaction towards this representations reflects the needs of the participants. The perceived level of diffi-culty is easy (8/18) to very easy (9/18). Participants are very satisfied (8/18) with this representation, and are very interested (13/18). Additionally, it has been found useful (14/18) and pleasant to use by all participants. See also table 1.
Participants were asked to name any negative aspects re-garding this representation. It was found confronting for users “who are overweight” (P2), and it “might be hard to process” (P5). As mentioned before, participants (4/18) state that they miss the exact numbers next to the object. Interestingly, P17 noted that “it can be the case that you only eat when it “allows” you to eat, only when you burned it”. Seven participants could not name any negative aspects. Yet, the use of tangible object was found the most positive aspect of the representation (7/18). P13 argued that food and movement are directly associated with each other, and therefore well understood. It was found motivating (7/18), and it gives a better understanding according to the partici-pants (8/18). P10 states: “It is motivating, you get triggered to think about what you eat. There is a real life connection to make it more understandable”. Others (7/18) support this statement by saying that the representation makes the user (more) aware of food intake.
4.5
User Preference
At the end of the interview the participants were asked about their preferences. The combination of the represen-tation in food and distance, together with the current visu-alizations of step counts (numbers), is mostly preferred re-garding to best meeting the needs of the participants (5/18). Table 2 lists the representations that meet the needs of the participants best. Overall, the representation of step count data in food has been chosen by participants that meet their needs best. However, the difference is small with the spatial representation that has been chosen by 13 participants. The representation in body size has not been chosen regarding to meet the needs best of the user. See also figure 5, which shows significantly that the spatial representation and rep-resentation of food meet the needs best, with little difference between them.
Out of all three representations, there is a clear trend of best understanding the step count data. The representa-tion in food is best understood by the participants (10/18). Other stated the spatial representation as best understand-ing of step count data (8/10). None of the participants had
Table 2: The results of participants’ preference of representation(s) that meet their needs best. The combination of representations spatial, food and numbers has the highest number of participants.
Representation # of participants Spatial 0 Food 2 Body Size 0 Spatial + Numbers 4 Food + Numbers 3
Body Size + Numbers 0
Spatial + Food 2
Food + Body Size 1
Spatial + Food + Numbers 5
the best understanding with the representation in body size. A Pearson chi-square test was conducted on the total num-ber of participants (n = 18) to determine if gender is re-lated to the best understanding of representation of step counts. The differences were not statistically significant, χ2 = 0.90, df = 1, p > 0.05(p.343). Another chi-square test was conducted to determine if there is a significant asso-ciation between participants who currently use an activity tracker and the preference of best understanding of step count data. However, there is no significant association, χ2= 0.055, df = 1, p > 0.05(p.814).
5.
DISCUSSION
Our study explored three alternative representations of step count data, in order to improve the efficiency of under-standing physical activity. We think that this exploratory study is the first iteration of design work that provides a relatively clearer picture of how current step count data can be better understood by alternative representations.
Food
This study showed that step count data is best understood by representing this data in food visualizations. Current visualizations of step count data include burned calories to the factual overview, based on the counted steps. Calories are a measurement for energy in food. Energy and step count are in line; by knowing how many calories there are in food, an understanding and balance is given, concerning the energy that is put into the human body and the energy that is being used [21]. Some participants indicated that the representation in food is positive since food and movement are associated with each other, and therefore relevant and well understood.
The purpose of the information visualization is to under-stand the factual step count data, and give insights related to this type of data. The use of information visualization amplifies the cognition, and can reduce the search for in-formation [22]. This is in line with the statements of par-ticipants. They found it easier to understand the factual numbers because one does not have to look up additional information to explain the step count data nor the calorie counter. Participants experienced that using tangible ob-jects gives them a better understanding, and brings perspec-tive to factual data. It seems that visualizing factual data in tangible objects gives a better understanding considering one can relate to the visualized objects. This is also
sup-ported by Gotsis et al. [23], according to them, relatedness is one of the main innate developmental needs of a human being. Participants indicated that by showing objects that are part of one’s daily intake, makes the retrieved informa-tion more interesting, useful, and pleasant. Addiinforma-tionally, others mentioned that showing images of food related to the physical activity makes one also aware of food intake, and is therefore also motivational.
Distance and Location
The spatial representation of step count data has a small difference with understanding step count data best, com-pared to understanding the representation in food. This can be explained by the use of physical objects. Participants mentioned that the use of equivalent distances give a good understanding of one’s physical activity because one can re-late to the located distance, for example Central Park, New York. This representation supports the step count data, in such way that it explains a specific number by using a spe-cific location that has an equivalent walking distance. This is as well supported by Gershon et al. [24] (p.39): “[...] be-cause we live and perceive in a physical world, it is easier to convey the information to the observer if the information is represented by being mapped to the familiar physical space”. However, some perceived this representation as too ab-stract, and that the use of concrete distances in measure-ment numbers should be included. This might be caused, considering we only made use of sketches that illustrate a background image of the location of the equivalent distance, and textually described the equivalent distance. We did not include specific details about the number of distance nor did we include a map of the location. Even though we mentioned to the participants clearly that these images that are being shown to them are sketches, and that it is not a prototype soon to be live; they already perceived it as something that was already in use or soon to be in use.
Body Size
The representation in body size is directly understood by the majority of the participants; a high number of steps is represented as an average weight person, and a few steps is represented as an overweight person. To date, this sup-ports the mental model; taking more steps, means being more physically active, and might result in a slim body size, and vice versa. Matching the human mental model is im-portant to understand the overall picture of the data [16]. As Schneiderman [25] (p.336) mentioned: “A picture is often cited to be worth a thousand words and, for some (but not all) tasks, it is clear that a visual presentation - such as a map or photograph - is dramatically easier to use than is a textual description or a spoken report”.
Even though the representation in body size is consider-ably understood, it does not meet the needs of the partici-pants nor they are satisfied with this representation. Show-ing body images make us pay attention since the human vision exceeds all other human senses. The human brain tells what the human sees, and thereby it is open for in-terpretation [18]. Body images as a representation of step count data, is interpreted as unrealistic, unpleasant, con-fronting, and offensive. A healthy body is not defined by its shape and size, it comes in different shapes and sizes [26]. Therefore, participants which indicated this representation as unrealistic, is fully appreciated.
The negative interpretation of the participants towards a representation in body size can be explained by the associa-tion with overweight. Overweight is seen, by both men and women, as being physically unattractive and associated with negative characteristics such as laziness, lack of willpower, and being out of control [27]. In order to improve the neg-ative interpretation of a representation in body size, which can be felt as being personally approached, might be per-ceived more positively if human activities are being shown to the user, instead of human body sizes. As P6 indicated, visualized activities can explain to the user that one’s step count is a lot or not.
Conclusion
Most of the participants indicated that a combination of one or more alternative representations and the current visual-izations of step count data (factual numbers), meet their needs best. Thus, they still prefer factual data in addition to an alternative representation, as a supplement to their measurement. This is supported by Consolvo et al. [13], who found that the measurement only does not provide suf-ficient information to the user. The overview of the numbers is perceived as being clear, and terms being used are well understood. However, the individual numbers are not well understood, and are therefore perceived as not providing enough available information concerning step count. Some participants indicated that one needs to be a professional to understand if a step count number is good or bad, or needs to look up information to understand this number. Raw sen-sor data, as what step count data actually is, requires more cognitive effort to understand [20], and therefore an alterna-tive representation in food and/or a spatial representation in addition to the numbers can decrease the cognitive effort, and improve the efficiency of understanding activity.
The research question “What alternative techniques can be applied to the visualization of pedometer sensor data to improve the efficiency of understanding activity ?” is being answered. The use of spatial representations and represen-tations of food can be applied to understand the pedometer sensor data more efficiently. The main reason for selecting the spatial and food representations as best understood, and meeting the needs of the participants best, is their related-ness to tangible objects. Using real-live objects ,where one can relate and refer to, give people a better understanding. According to Conroy et al. [28], people are seeking for multiple mobile applications to initiate and maintain be-haviour change regarding to physical activities. However, we believe that the user first need to understand the data correctly and efficiently.
6.
CONCLUSION
In this paper we described alternative representations of the step count data, received from mobile sensors. We sketched out three alternatives for current visualizations of step count data: spatial representation, representation in body size, and a representation in food. We conducted a interview-based questionnaire (n=18), focusing on findings related to understanding of and experience with current visualizations of step count data, and presented the alternative represen-tations.
Current visualizations of step count data was perceived as being a clear overview of factual data. However, the meaning
of the presented data is not well understood and an expla-nation is missing. The majority of the participants found that the representation in food is best understood. With a small difference, the spatial representation is well under-stood to represent step count data. The representation of step count data in body size has not been perceived posi-tively. Visuals of body size is perceived as being confronting and offensive to the user. Body size should not be related to the user, since body size is not related to health state. This study discovered that relatedness to tangible objects plays an important role in representing step count data, in order to understand raw sensor data better; as being used with the spatial representation and the representation of food.
This exploratory research through design is the first it-eration of design work to improve the efficiency of under-standing step count data. We believe that this study con-tributes to the field of information visualization for different health monitoring activities that support the process of un-derstanding sensor data.
7.
FUTURE WORK
Future research should be the next iteration of design work. We acknowledge that there are challenges in further formalizing this type of research. The results of this study show that participants prefer personalization of health re-lated applications. This subject should be taken into ac-count, in order to measure if this affects the understanding of sensor data. Additionally, according to other researchers [8] context information is a factor that affects individuals‘ behaviour. We did not take into account behaviour change nor to maintain behaviour. However, this is an important factor for the reasoning of using health and physical activity related applications.
We hope that this research encourages other fellow re-searchers to further investigate into this topic.
8.
ACKNOWLEDGMENTS
The author would like to thank the supervisor Daniel Buzzo for the invaluable guidance and feedback throughout the master thesis. I would also like to thank all participants for taking part in this study.
9.
REFERENCES
[1] PwC, “Contents :The Wearable Future,” tech. rep., PwC, 2014.
[2] D. Lupton, “Self-tracking Modes: Reflexive Self-Monitoring and Data Practices,” no. August, pp. 1–18, 2014.
[3] I. Li, A. K. Dey, and J. Forlizzi, “Understanding my data, myself,” Proceedings of the 13th international conference on Ubiquitous computing - UbiComp ’11, p. 405, 2011.
[4] S. Few, “Information Dashboard Design,” in Information Dashboard Design, ch. 1, pp. 1 – 37, O’Reilly Media, 2006.
[5] NarrativeScience, “Narrative Analytics : A Narrative Science Whitepaper,” tech. rep., Narrative Science, 2014.
[6] M. Aitken, “Patient Apps for Improved Healthcare From Novelty to Mainstream,” Tech. Rep. October, IMS Institute for Healthcare Informatics, 2013.
[7] C. Kratzke and C. Cox, “Smartphone Technology and Apps: Rapidly Changing Health Promotion,”
International Electronic Journal of Health Education, vol. 15, no. 11, pp. 72–82, 2012.
[8] I. Li, “Personal informatics & context: Using context to reveal factors that affect behavior,” Journal of Ambient Intelligence and Smart Environments, vol. 4, pp. 71–72, 2012.
[9] A. Banner, C. Nead, A. Wyke, J. Case, and
T. Newbold, “What do people want from health apps ? A survey of 250 patient and consumer groups,” PatientView, pp. 5–9, 2013.
[10] H. J. West, C. P. Hall, L. C. Hanson, D. M. Barnes, C. Giraud-Carrier, and J. Barrett, “There’s an app for that: Content analysis of paid health and fitness apps,” Journal of Medical Internet Research, vol. 14, p. e72, May 2012.
[11] P. Leijdekkers, V. Gay, P. Leijdekkers, and V. Gay, “Improving User Engagement by Aggregating and
Analysing Health and Fitness Data on a Mobile App,” vol. 9102, pp. 325–330, 2015.
[12] K. J. Heffernan, “The potential of eHealth Apps to Support Targeted Complex Health Messages,” Journal of General Practice, vol. 02, no. 05, 2014.
[13] S. Consolvo, K. Everitt, I. Smith, and J. A. Landay, “Design requirements for technologies that encourage
physical activity,” in Proceedings of the SIGCHI conference on Human Factors in computing systems, pp. 457–466, ACM, 2006.
[14] L. Chittaro, “Information visualization and its application to medicine,” Artificial Intelligence in Medicine, vol. 22, pp. 81–88, 2001.
[15] N. Gershon, S. G. Eick, and S. Card, “Information visualization,” interactions, vol. 5, pp. 9–15, Mar. 1998.
[16] A. Cuttone, “SensibleJournal : A Mobile Personal Informatics System for Visualizing Mobility and Social Interactions,” SensibleJournal, pp. 1–83, 2013.
[17] J. S. Yi, Y.-a. Kang, J. T. Stasko, and J. a. Jacko, “Understanding and characterizing insights: How Do
People Gain Insights Using Information
Visualization?,” Proceedings of the 2008 conference on
BEyond time and errors novel evaLuation methods for Information Visualization - BELIV ’08, p. 1, 2008. [18] L. B. Madsen, Data-Driven Healthcare: How Analytics
and BI are Transforming the Industry. Wiley, 2014. [19] J. K. Burgoon, J. a. Bonito, B. Bengtsson,
C. Cederberg, M. Lundeberg, and L. Allspach, “Interactivity in human-computer interaction: A study of credibility, understanding, and influence,”
Computers in Human Behavior, vol. 16, no. 6, pp. 553–574, 2000.
[20] G. M. Marakas, R. D. Johnson, and J. W. Palmer, “A Theoretical Model of Differential Social Attributions Toward Computing Technology: When the Metaphor becomes the Model,” International Journal of
Human-Computer Studies, vol. 52, no. 4, pp. 719–750, 2000.
[21] E. Whitney and S. R. Rolfes, Understanding nutrition. Cengage Learning, 2007.
[22] S. K. Card, J. D. Mackinlay, and B. Shneiderman, Readings in information visualization: using vision to think. Morgan Kaufmann, 1999.
[23] B. Spring, M. Gotsis, A. Paiva, and D. Spruijt-Metz, “Healthy apps: mobile devices for continuous monitoring and intervention,” Pulse, IEEE, vol. 4, no. 6, pp. 34–40, 2013.
[24] N. Gershon and S. G. Eick, “Visualization’s new tack: making sense of information,” Spectrum, IEEE, vol. 32, no. 11, pp. 38–40, 1995.
[25] B. Shneiderman, “The eyes have it: A task by data type taxonomy for information visualizations,” in Visual Languages, 1996. Proceedings., IEEE Symposium on, pp. 336–343, IEEE, 1996.
[26] L. A. Smolin and M. B. Grosvenor, Nutrition and eating disorders. Infobase Publishing, 2009. [27] S. Grogan, Body image: Understanding body
dissatisfaction in men, women and children. Routledge, 2007.
[28] D. E. Conroy, C.-H. Yang, and J. P. Maher, “Behavior change techniques in top-ranked mobile apps for physical activity,” American journal of preventive medicine, vol. 46, no. 6, pp. 649–652, 2014.
