Abstract
Among runners, there is a high drop-out rate due to injuries and loss of motivation. These runners often lack personalized guidance and support. While there is much potential for sports apps to act as (e-)coaches to help these runners to avoid injuries, set goals, and maintain good intentions, most available running apps primarily focus on persuasive design features like monitoring, they o er few or no features that support personalized guidance (e.g., personalized training schemes). Therefore, we give a detailed description of the working mechanism of Inspirun e-Coach app and on how this app uses a personalized coaching approach with automatic adaptation of training schemes based on biofeedback and GPS-data. We also share insights into how end-users experience this working mechanism. The primary conclusion of this study is that the working mechanism (if provided with accurate data) automatically adapts training sessions to the runners' physical workload and stimulates runners' goal perception, motivation, and experienced personalization. With this mechanism, we attempted to make optimal use of the potential of wearable technology to support the large group of novice or less experienced runners and that by providing insight in our working mechanisms, it can be applied in other technologies, wearables, and types of sports. Keywords: personalization; app; e-coaching; tailoring; running; training; workload; experience; heart rate; RPE 1. Introduction In recent years there has been an exponential increase in the availability and use of sports and physical
group of novice or less experienced runners and that by providing insight in our working mechanisms, it can be applied in other technologies, wearables, and types of sports. Keywords: personalization; app; e-coaching; tailoring; running; training; workload; experience; heart rate; RPE 1. Introduction In recent years there has been an exponential increase in the availability and use of sports and physical activity-related monitoring devices such as smartphone applications (apps), activity trackers, and sports watches [1,2]. This increased use of these monitoring devices is consistent with trends like Quanti ed Self [3] and mHealth [4], which emphasize the potential of these monitoring devices to contribute to a healthy and active lifestyle by supporting behavior change [5]. In particular, smartphones have several advantages. They are widely used, are embedded in everyday life [6,7], and allow people to collect data anywhere, anytime [8]. Since most people already have a smartphone (up to 76% of adults) [9] and apps are relatively cheap, often even free of charge, apps are accessible for almost everyone. The use of sports apps in Western Europe is mainly re ected in individual, recreational sports such as running, cycling, walking, and tness. Mainly among runners, apps are widely used. Research shows that approximately 5075% of (event) runners use a running app, Sensors2020,20, 4590; doi:10.3390 /s20164590 /journal/sensors
Sensors2020,20, 4590 2 of 19 especially novice or less experienced runners [1]. Among these runners, there is a high drop-out rate due to injuries and loss of motivation. Because these runners often lack personalized guidance and support. While there is much potential for sports apps to act as (e-)coach to help these runners to avoid injuries, set goals, and maintain good intentions [6,10,11]. While most available running apps primarily focus on persuasive design features like monitoring, they o er few or no features that support personalized guidance (e.g., personalized training schemes) [2,12]. Vos et al. [6] designed Inspirun, an e-coach for runners. Inspirun is a running app that o ers a personalized coaching approach with the automatic adaptation of training schemes based on biofeedback and GPS-data. In the study of Vos et al. [6] a full description is given on how they designed this app, and the steps they took to develop the app. Whereas, the present paper gives a detailed description of the working mechanism; the personalized coaching approach with the automatic adaptation of training schemes based on biofeedback and GPS-data. But also aims to give insight into how end-users experience this working mechanism. The paper is organized as followed. First, we provide a short overview of related work. Second, we shortly describe the development and working mechanism of Inspirun. Third, we introduce the study protocol used to test Inspirun. Fourth, we present the results of the end-user test. Finally, we discuss the results of the study and present suggestions for future work. 2. Related Work Persuasive technology is studied extensively in the literature, also in relation to apps. A framework called Persuasive Systems Design (PSD) model is widely used for designing and evaluating systems that in uence the attitudes or behaviors of users [13]. A review of Matthews et al. [12] on persuasive technologies used in apps, concludes that the most commonly used persuasive feature was self-monitoring. Thereby, apps often use data that is collected by the app to motivate the user to stay engaged (i.e., rewards, reminders, and suggestions). Matthews and colleagues [12] also conclude that many
attitudes or behaviors of users [13]. A review of Matthews et al. [12] on persuasive technologies used in apps, concludes that the most commonly used persuasive feature was self-monitoring. Thereby, apps often use data that is collected by the app to motivate the user to stay engaged (i.e., rewards, reminders, and suggestions). Matthews and colleagues [12] also conclude that many proven persuasive features are not utilized. Personalization is an example of one of these features that is (often) not implemented. Personalization is described by PSD framework as o ering personalized content and services to the user. So, a tailored or personalized feature is one that is adapted to the characteristics of the end-user [14]. Speci cally for running,Van Hooren et al.[11] propose a framework to optimize real-time feedback for reducing injury risk and improving performance and motivation. They argue that personalized real-time feedback on workload and running technique can be provided based on the individual preferences, experiences, and motives. For example, personalizing the type of feedback to suit preferences of the runner or personalizing the runners' training session to suit the runners' workload capacity. Research on the development of running (and other sports) related apps include systems that use several persuasive features as described in the PSD model. These technological systems can be divided into three di erent groups of studies, based on their objectives. The rst group of studies focuses on improving running technique to minimize injury. For example, Aranki et al. [15] developed RunningCoach, a mobile health system that monitors and gives feedback on running cadence to optimize it, and (possibly) minimize running injury. They use self-monitoring and a way of tailoring (feedback can be changed by the user) as a persuasive feature in their system. Another example is Runmerge, an app developed by Kiss et al. [16], which enhances body awareness using visualization of their steps to help runners towards a better running experience. Authors found that enhanced proprioception (i.e., `knowing your body') can be bene cial for everyday running training. Nylander and Tholander [17] developed Runright, which provides real-time visual and audio feedback about the current
is Runmerge, an app developed by Kiss et al. [16], which enhances body awareness using visualization of their steps to help runners towards a better running experience. Authors found that enhanced proprioception (i.e., `knowing your body') can be bene cial for everyday running training. Nylander and Tholander [17] developed Runright, which provides real-time visual and audio feedback about the current running rhythm. Their non-interpretive visualization led users to their interpretation of the feedback. Valsted et al. [18] developed Strive, a wearable that aims to assist runners in achieving rhythmic breathing; a breathing technique that potentially leads to improved running results and lower injury risk. The user's understanding of feedback patterns based on self-monitoring was assessed.
Sensors2020,20, 4590 3 of 19 A second topic that is often studied is the social aspects of running. In this category, dialogue and social support features are common. For example, Timmerman [19], investigated how technology can support a group of runners. In line with that, Mueller et al. [20] investigated how technology can support the social aspects of running. By introducing `Jogging over a Distance', a system that allowed runners all over the world to run together using an audio-based social comparison feature. Further, HeartLink [21], a system that broadcasts live biometric data to social networks, and RUFUS [22], a system that enabled runners to communicate with supporters using `praise' during races, are examples that also focus on the social aspects of running. The third group of studies aims to enhance the motivation of runners. For example, the e-coaching ecosystem [23] o ers interactions between end-users and human trainers to enhance motivation and stimulate a healthy and active lifestyle. Again, social support features are used as part of persuasive technology. A human trainer was essential in this design. Runners were more engaged when professionals o ered or supervised the training sessions, compared to a group of runners with self-made training sessions and no supervision. Whereas most work related to motivation is focused on novice runners, Knaving et al. [24] determined a framework and guidelines to design technology for experienced runners. In conclusion, most of the discussed studies do use persuasive features, mostly social support or dialogue support features (as classi ed in the PSD-model). None of them implement personalization features in general, and none of them aim to personalize training sessions based on the runners' workload capacity. 3. Development and Working Mechanisms of Inspirun 3.1. Development from a Multi-Disciplinary Perspective Inspirun is an Android-based system built with Ionic Framework using JavaScript and AngularJS. The app is connected (reverse engineering of the API) to the heart rate monitor (Wahoo TICKR X, Atlanta, GA, USA) to collect the heart rate during running. Inspirun was designed by a multidisciplinary team. This team was composed of experts in behavioral sciences (n=1), human movement sciences
Inspirun is an Android-based system built with Ionic Framework using JavaScript and AngularJS. The app is connected (reverse engineering of the API) to the heart rate monitor (Wahoo TICKR X, Atlanta, GA, USA) to collect the heart rate during running. Inspirun was designed by a multidisciplinary team. This team was composed of experts in behavioral sciences (n=1), human movement sciences (n=2), electronic engineering (n=2), and industrial design (n=1). All experts were selected based on their educational background and having at least 5 years of experience in this domain. To understand the runner and to serve their interest and needs, crossovers between these di erent elds of work were necessary. The multidisciplinary design approach and the steps taken are reported in previously published work [6]. In this paper, we extend and elaborate on this work, through an in-depth analysis of the personalized coaching approach, including the automatic adaptation of training schemes based on biofeedback and GPS data. 3.2. Algorithm That Automatically Adjusts the Training Schemes To create an algorithm that automatically adjusts the training scheme, we analyzed the approach that experienced coaches and trainers use when creating training schedules for runners. Interviews with experienced coaches and trainers (those who had at least 5 years of experience in coaching and training) revealed that in general, most coaches and trainers take the following steps (see also Figure). 1. 2. Select running goals for the upcoming training period and analyze the data of the current running level. 3. 4. Monitor during the running session and coach during/after the running session (comparing the executed data to the prescribed data). 5. 6.
Sensors2020,20, 4590 4 of 191. Collect data current running level 2. Select running goals & analyzing data current running level 3. Select training sessions 4. Monitoring training session (during and after) 5. Adjust next training session Figure 1. Steps taken by coaches and trainers when creating training schedules for runners. Final step is to continue iteratively with step 4 and step 5. 3.2.1. Steps 1 and 2: Data and Analysis of the Current Running Level and Select Running Goals To collect the runners' current running level and select running goals for the upcoming training period (Step 1), we rst designed a questionnaire to give insight into their current running level and running goals. Runners had to choose one of four possible answers: `I am completely new to running', `I can run for 15 min without walking', `I can run for 30 min without walking', or `I can run for 60 min without walking'. Next, the running goal selection in our system was based on this self-declared current running level. If they were completely new to running, the easiest running goal (running 15 min without walking) was automatically assigned. If they were already able to run for 15 min without walking, again a running goal was automatically assigned, namely running 5 km without walking. If a runner was able to run 30 or 60 min without walking, he/she was allowed to choose their own goal. Available options were (i) running 5 km faster, (ii) running 10 km without walking, or (iii) running 10 km faster (see Figure detailed owchart). In order to collect objective data about runners' current running level, we designed di erent test programs that consist of three running sessions. For each test session, we collected the heart rate (body feedback), GPS-data, and the perception of the training intensity (Rating of Perceived Exertion (RPE-score), a subjective parameter). In the test sessions, Inspirun only gave instructions to the runner but no feedback. For example, in the rst session one of the instructions was: `start running at a comfortable pace' and `you are doing well if you breathe heavily but are
(body feedback), GPS-data, and the perception of the training intensity (Rating of Perceived Exertion (RPE-score), a subjective parameter). In the test sessions, Inspirun only gave instructions to the runner but no feedback. For example, in the rst session one of the instructions was: `start running at a comfortable pace' and `you are doing well if you breathe heavily but are able to have a conversation as well'. With this instruction, runners started running at their own comfortable pace. Meanwhile, the app registered heart rate data, running speed, and RPE-score. During the test sessions, information was collected on the heart rate values and RPE scores for di erent running speeds (jogging, easy running, comfortable speed, hard running, and very hard running) We labeled this relation between speed, heart rate, and RPE as the Personal Running Pro le (PRP). See Table after the three test sessions.
Sensors2020,20, 4590 5 of 19Sensors 2020, 20, x FOR PEER REVIEW 5 of 20 Figure 2. Simplified flowchart of Inspirun. Starting from the running experience, to running goal, and which training scheme fits best. For full details see Supplementary File S1 Detailed Flowchart. Figure 2. Simpli ed owchart of Inspirun. Starting from the running experience, to running goal, and which training scheme ts best. For full details see Supplementary File S1 Detailed Flowchart.
Sensors2020,20, 4590 6 of 19 Table 1. Example of a current Personal Running Pro le (PRP) after the three test sessions, heart rate values (beats per minute), and Rating of Perceived Exertion (RPE) scores for di erent running speeds (kilometers per hour) (jogging, easy running, comfortable speed, hard running, and very hard running). Test 1 Test 2 Test 3 Current PRP Pace RPE SpeedHeart rate RPE SpeedHeart rate RPE SpeedHeart rate RPE SpeedHeart rate Jogging 4 10.0 121 4 10.2 124 4 10.1 123 Easy 5 11.6 139 5 11.6 139 Comfortable 6 12.3 145 6 12.2 146 6 12.1 143 6 12.2 145 Hard 7 13.8 156 7 13.4 150 7 13.6 153 Very Hard 8 14.8 173 8 14.8 173 3.2.2. Step 3: Select Training Sessions that Match their Level and Contribute to the Running Goals The second step in the process was the selection of training sessions that match the runner's current running level and that contribute to their goal. We used ve generally accepted training principles (i.e., individualization, progression, overload, variation, and objective and subjective monitoring of the performance [2527]) in combination with the expertise of the human movement scientists, involved in the development of the Inspirun e-coach, to create training schedules. For each goal (e.g., running 5 km without walking), a training schedule was constructed. In total, a schedule consists of 20 sessions divided over several weeks dependent on the number of sessions per week (between 1 and 3). To make the training schedules and sessions applicable to all runners, we chose to personalize training sessions based on workload and intensity. This means that the training schedule (i.e., distance, total time, and type of training) was the same for a runner with the same running level and goal, while the intensity varies per runner. We chose RPE as the parameter for intensity because of its validity, reliability, and internal consistency [28,29]. In the app, there is an explanation how runners should use the RPE score. For each score the runner can read an explanation how this RPE should feel in terms of breathing, the ability
level and goal, while the intensity varies per runner. We chose RPE as the parameter for intensity because of its validity, reliability, and internal consistency [28,29]. In the app, there is an explanation how runners should use the RPE score. For each score the runner can read an explanation how this RPE should feel in terms of breathing, the ability to talk, and the RPE is expressed in words like, hard/very hard/comfortable [29]. See Table 3.2.3. Step 4: Monitoring during the Running Session and Coaching during/after the Running Session (Comparing the Executed data to the Prescribed Data) The third step is to monitor each running session and coach the runner (by comparing the performance data to the prescribed data). To prescribe a training session, the most current PRP (current PRP is an average of the last six sessions) is used for each session (see Table This means that for each RPE present in that speci c session, the matching speed and heart rate are selected.
Description
The study investigates the personalization of training schemes for runners using the Inspirun app.