TYDR – Track Your Daily Routine. Android App for Tracking Smartphone Sensor and Usage Data [original]

TYDR – Track Your Daily Routine.

Android App for Tracking Smartphone Sensor and Usage Data

Felix Beierle, Vinh Thuy Tran

Service-centric Networking

Technische Universität Berlin

Telekom Innovation Laboratories

Berlin, Germany

beierle@tu-berlin.de,vinh.t.tran@

campus.tu-berlin.de

Mathias Allemand

Department of Psychology

University of Zurich

Zurich, Switzerland

m.allemand@psychologie.uzh.ch

Patrick Neff, Winfried Schlee

Clinic and Policlinic for Psychiatry

and Psychotherapy

University of Regensburg

Regensburg, Germany

[email protected],winfried.

[email protected]

Thomas Probst

Department for Psychotherapy and

Biopsychosocial Health

Danube University Krems

Krems, Austria

thomas.probst@donau-uni.ac.at

Rüdiger Pryss

Institute of Databases and

Information Systems

Ulm University

Ulm, Germany

ruediger.pryss@uni-ulm.de

Johannes Zimmermann

Psychologische Hochschule Berlin

Berlin, Germany

j.zimmermann@

psychologische-hochschule.de

ABSTRACT

We present the Android app TYDR (Track Your Daily Routine)

which tracks smartphone sensor and usage data and utilizes stan-

dardized psychometric personality questionnaires. With the app,

we aim at collecting data for researching correlations between the

tracked smartphone data and the user’s personality in order to pre-

dict personality from smartphone data. In this paper, we highlight

our approaches in addressing the challenges in developing such

an app. We optimize the tracking of sensor data by assessing the

trade-off of size of data and battery consumption and granularity of

the stored information. Our user interface is designed to incentivize

users to install the app and fill out questionnaires. TYDR processes

and visualizes the tracked sensor and usage data as well as the

results of the personality questionnaires. When developing an app

that will be used in psychological studies, requirements posed by

ethics commissions / institutional review boards and data protec-

tion officials have to be met. We detail our approaches concerning

those requirements regarding the anonymized storing of user data,

informing the users about the data collection, and enabling an opt-

out option. We present our process for anonymized data storing

while still being able to identify individual users who successfully

completed a psychological study with the app.

CCS CONCEPTS

•Human-centered computing →Ubiquitous and mobile com-

puting;•Applied computing →Psychology;

Permission to make digital or hard copies of part or all of this work for personal or

classroom use is granted without fee provided that copies are not made or distributed

for profit or commercial advantage and that copies bear this notice and the full citation

on the first page. Copyrights for third-party components of this work must be honored.

For all other uses, contact the owner/author(s).

MOBILESoft ’18, May 27–28, 2018, Gothenburg, Sweden

ACM ISBN 978-1-4503-5712-8/18/05.

https://doi.org/10.1145/3197231.3197235

KEYWORDS

context-aware computing; psychometrics; sensor data; Android

ACM Reference Format:

Felix Beierle, Vinh Thuy Tran, Mathias Allemand, Patrick Neff, Winfried

Schlee, Thomas Probst, Rüdiger Pryss, and Johannes Zimmermann. 2018.

TYDR – Track Your Daily Routine. Android App for Tracking Smartphone

Sensor and Usage Data. In MOBILESoft ’18: 5th IEEE/ACM International

Conference on Mobile Software Engineering and Systems, May 27–28, 2018,

Gothenburg, Sweden. ACM, New York, NY, USA, 4 pages. https://doi.org/10.

1145/3197231.3197235

1 INTRODUCTION

Context-aware applications consider the context the user is cur-

rently in. This usually includes factors like the location, weather,

or time. As previous research suggests, higher level information

like the user’s personality could be beneficial to improve context-

awareness in mobile applications. This has been shown for areas

like mobile health [

] or mobile social networking [

]. Assessing

the user’s personality is a tedious task that is usually done by filling

out questionnaires. As researchers in computer science and psy-

chology, we aim at predicting the personality of a user by their

smartphone usage behavior and sensor data. For the purpose of col-

lecting data for training a prediction model, we developed the app

TYDR (Track Your Daily Routine). It tracks smartphone sensor and

usage data and queries the user with standardized psychometric

questionnaires that yield measurements of personality traits.

There are several things to consider when developing such an

app. The more users we can attract, the more reliable our results

will be. The app should have an appealing interface and should offer

some feature for the user. As we are developing for a mobile device,

we have to consider the restrictions these devices pose, like battery

and space limitations. The concrete study we plan includes a ques-

tionnaire that has to be filled out daily. Finding users for this might

have to be incentivized externally. When conducting psychological

studies, it is common that users get paid or are compensated with

university credit points. For such compensation, it is important to

2018 ACM/IEEE 5th International Conference on Mobile Software Engineering and Systems

MOBILESoft ’18, May 27–28, 2018, Gothenburg, Sweden F. Beierle et al.

know which users successfully completed the study. At the same

time however, linking the sensitive data that TYDR tracks to iden-

tifying information about the user is highly undesirable because

of privacy concerns. To summarize, the requirements for an app to

conduct our study, are: (1) be appealing to attract users, (2) consider

restrictions posed by developing for mobile devices, and (3) privacy,

especially the collection of anonymized data while still being able

to tell which individual users completed the study successfully.

In this paper, we present TYDR and highlight the challenges in

developing an Android app for the defined use case. In the follow-

ing, we present our design and optimizations for tracking sensor

data (Section 2). In Section 3, we detail how TYDR makes the data

tracking its core feature by processing and visualizing it for the

user. In Section 4, we describe the measures we implemented for

privacy protection and data anonymization.

2 SENSOR DATA TRACKING

Google released the Google Awareness API

that offers developers

to retrieve different context data through one API (time, location,

places, beacons, headphones, activity, weather). There are two ways

of retrieving data: Fence API and Snapshot API. A snapshot yields

current data from the seven sources. Through the Fence API, the

developer can register listeners and receives a callback when the

desired conditions are met. These two approaches are useful for

the development of context-aware applications that either instantly

need the current context of the user or want to be notified when the

user is in a specific context. Aiming at tracking the user’s context,

we have to go beyond what the Google Awareness API offers. In

TYDR, we track: location, weather, ambient light sensor, accelerom-

eter, activity, steps, phone un-/lock, headphone un-/plug, battery

and charging, Wifi, Bluetooth, calls metadata, music metadata, pho-

tos metadata, notifications metadata, app usage, and app traffic.

For some of the listed data sources we want to track, a passive,

listener-based approach is possible: the music that is played is

broadcast by (most) music player apps. We can just register a listener

and track the played back music whenever the user is listening to

it [

]. The same approach can be used for tracking the usage of

the phone: registering listeners for locking and unlocking events

enables us to track when the user was (most likely) interacting with

the phone.

Besides such a listener-based approach, in some cases we have to

do periodical tracking. Given the related permission, the Android

system offers information about which apps were used for what

number of seconds. Android also offers information about the traffic

each app caused. Such data about the user is especially meaningful

when we have it on a fine-granular level. For example: having

one data point per week that indicates for how long a user used

a specific app would probably yield less interesting insights than

knowing the app usage duration for each hour. As there are no

listeners available for querying such app statistics, it is necessary

to schedule periodic queries.

For periodically tracking context data, there is the trade-off be-

tween the

(1) amount of data we have to store,

(2) battery-consumption, and

1https://developers.google.com/awareness/

(3) level of detail of the information.

In the most extreme case, the frequency with which we query

context data is very high, in order to increase the level of detail for

quickly changing data. Then we – and the user – have to accept a

higher battery consumption. An example for this quickly changing

data is that of the light sensor or the accelerometer. We did some

further optimizations, for example, by only tracking accelerometer

data when we detect movement indicated by the step count we

track. Although the light sensor only yields data when the phone

is unlocked, the frequency with which it yields data is still so high

that it would use too much space. Furthermore, such detailed light

sensor data would not be that useful for our research purpose. What

could be interesting is to check whether the user is in a dark or in a

well-lit place. To reduce the amount of data we store from the light

sensor, we divided the possible range of light sensor values into

several segments and only store changes between the segments. To

counteract the effects of rapid changes between the limits of the

defined segments, we implemented a hysteresis. The introduced

inaccuracy is negligible for our research purpose.

3 USER INTERFACE

In this section, we describe TYDR’s user interface. In Section 3.1,

we show the main screen where the users can see the data that is

tracked about them. Section 3.2 is about the permanent notification

and how it can be configured. Section 3.3 introduces the mobile

questionnaire interface that will be used in psychological studies.

3.1 Main Screen

Figure 1 shows the main screen. The tile-based design gives the user

an immediate overview of the data for the current day. Each tile can

be touched to slide open a bigger tile with a weekly view of the data.

In the figure, the user views the weekly data for her/his phone usage

times. The red tile with the "Grant Permission" button shows the

number of locations the user stayed at. As the location permission

was not granted, the tile cannot display data and displays the button

instead. Not shown in the screenshot due to space limitations are

two additional tiles related to the number of notifications per app

and the number of photos taken.

3.2 Customizable Permanent Notification

In order to improve the chances of TYDR not being stopped by

task cleaner apps, we implemented a permanent notification. It

runs those services in the foreground that track data with a high

frequency. To make it appealing, we followed the same approach

as for the main screen: show the user meaningful and informative

figures based on processed tracking data.

The permanent notification will show up in the notification bar

and the lockscreen. The notification is designed to be adaptive to

the user’s interests by offering the possibility to configure what

information is displayed, see Figure 2. The Preview section in the

figure shows what the notification will look like.

3.3 Mobile Questionnaire

TYDR utilizes questionnaires for demographic data that is not possi-

ble to track automatically. Additionally, we use standardized psycho-

metric questionnaires to assess the user’s personality with which

TYDR – Track Your Daily Routine.

Android App for Tracking Smartphone Sensor and Usage Data

MOBILESoft ’18, May 27–28, 2018, Gothenburg, Sweden

Figure 1: The main screen of TYDR that visualizes daily and

weekly summaries of collected data.

Figure 2: The user can configure the notification. The Pre-

view section shows what the notification will look like.

we label the collected smartphone data. To be able to update the

questionnaires independently from updating the whole app, the

latest questionnaire version is fetched from the backend.

To the best of our knowledge, currently there is no official or

widely adopted library for mobile questionnaires on the Android

platform. Following general mobile survey design guidelines2,we

developed a questionnaire UI. Only one question is displayed at a

time, which avoids scrolling. The users can switch between apps or

turn off the screen and continue where they left off when resuming

TYDR. A progress bar indicates how much of the current question-

naire is already filled out. The incentive for the user to fill out the

personality questionnaires is to see their results in the related tile.

4 PRIVACY PROTECTION

In TYDR, we deal with highly sensitive data. In this section, we

detail what measures we took in order to ensure the user’s privacy.

In Section 4.1, we describe what decisions we made during the

design phase. In Section 4.2, we describe how we designed and

implemented a way to identify individual users without linking

to their data. With this method, we are able to contact study par-

ticipants that successfully completed the study without knowing

which smartphone data points are related to them.

4.1 Privacy by Design

The first important thing is that the app does not require any login.

This way, the user does not have to remember any login data. Addi-

tionally, the user’s data can only be stored anonymously. However,

in order to achieve our research goal to find correlations between

the user’s personality and the collected smartphone data, we need

to identify which data point belongs to which user. For this, we

utilized an ID provided by Google Play Services, that is unlikely to

change during the duration of the study. We used the identifier in

salted and hashed form, in order to further disallow any potential

linking to other databases.

Furthermore, wherever possible, we store only metadata of the

sensor or usage data, for example regarding notifications, photos,

music, or calls. Whenever we could potentially track data that

would make the user personally identifiable, we store a salted and

hashed form of that data point, for example Bluetooth device IDs

or Wifi SSIDs. Using the same salt and hash function, we still can

re-identify the same IDs without knowing the IDs themselves. The

hashing is already performed on the phone, before writing to the

local SQLite database and before uploading to our servers.

There are three main requirements the data protection official

from one of our universities stated:

(1)

Very clearly present to the user which data is being collected.

(2)

Inform the users about the data, its collection, and upload

before those processes of the app are started.

(3)

Inform the users how they can stop the data collection and

transfer.

The first requirement is fulfilled by our privacy policy in which we

list all the data we collect and explain how it is stored and what is

transferred to our servers. The second requirement is fulfilled by

the Google Play Store, where for each app, there can be a link to the

E.g., https://www.surveymonkey.com/mp/how-to-create-surveys/ or https://www.

uxmatters.com/mt/archives/2017/02/8-best-practices-for-mobile-form-design.php.

MOBILESoft ’18, May 27–28, 2018, Gothenburg, Sweden F. Beierle et al.

app’s or developer’s website and to the privacy policy. As most users

probably will not scroll down on the Google Play Store website

to click and read the privacy policy, we additionally implemented

a mandatory user confirmation of our policy in the app. Before

the app fully starts and starts collecting data, the user sees a terms

and privacy policy screen. Only after explicit confirmation, data is

being collected and transferred. The third requirement is fulfilled

by informing the user that uninstallation of the app will stop any

data collection and transfer. Additionally, we offer users the option

to contact us via a feedback form in the app to request the deletion

of their data.

4.2 Identifying Individual Users Without

Linking to Their Collected Data

In psychological studies, it is common that users are compensated

with university course credits, get paid to participate, or have the

chance to win money/vouchers in a raffle after study completion.

Furthermore, psychological journals – and some computer science

conferences as well – typically require approval codes from an

ethics commission / institutional review board when submitting

a paper utilizing data from a study. Data privacy is an important

factor for the approval of an ethics commission, especially when

the user data is as sensitive as the data TYDR collects. To recognize

how sensitive the collected data might be, consider, for example,

that when combining the data, we can see for a user which app

he/she used for how long at which location and what brightness

the location had. To alleviate the potential privacy concerns of

users and to receive required approvals of ethics commissions, we

developed a process that allows us to both

(1) check if users successfully participated in a study and

(2)

contact study participants without knowing which smart-

phone data belongs to them.

This way, we can directly draw and contact the winners of a raffle,

without having a link between collected data and meaningful user

identifiers, i.e., email addresses. Additionally generated participa-

tion codes can further be used for claiming university credit points,

if applicable. There are three steps to our design of this system:

User sign-up.

In order to participate in the study, additionally to

just installing the app, users have to sign up with an email address,

so we can contact the winners of the raffle. Entering an email

address has the potential to de-anonymize the data. We store the

data about the study participants in a separate table in the backend

that is not linked to the tables containing smartphone data.

Checking study participation success.

The planned study

includes the commitment of the participants to fill out daily psy-

chological questionnaires. Users who do not fill those out regularly

should not be eligible to receive compensation. As we have no link

between the identifiable study participants and their smartphone

data, we cannot check the rate of filling out the daily questionnaire

on the backend. However, we can check this rate in the app and

report to the backend whether it meets the required percentage.

Generating participation codes.

After successful study com-

pletion, the user can trigger the study completion by pressing the

related button in the sidebar menu. This triggers the app to let the

backend know about the successful completion of the study. The

backend then generates an individual participation code and replies

to the app’s query with it. This code can then be used to indicate

successful participation to, e.g., claim university credit points.

5 CONCLUSION AND FUTURE WORK

In this paper, we presented the design and implementation of TYDR,

an Android app which tracks the user’s smartphone sensor and

usage data and queries the user with personality questionnaires.

We will utilize this app for conducting a study with the goal to

predict the user’s personality from smartphone data. We detailed

the challenges in implementing such a system. Sensor data has to be

tracked efficiently by finding the right balance between the amount

of data to be stored, battery consumption, and level of detail of the

information. The user interface should be attractive and intuitive.

We turned the data collection into TYDR’s core feature of processing

and visualizing daily statistics for the user. Especially when dealing

with sensitive data like in TYDR, privacy and data anonymization

have to be a key concern. We showed how we designed our app

with privacy in mind, taking into account the concrete requirements

that are posed by data protection officials and ethics commissions

/ institutional review boards. Furthermore, we showed how we

can identify individual users that successfully completed a study

without linking their email addresses to the smartphone data we

collected from them. The process for this consists of storing contact

data separately from the collected data and letting the app check

the requirements for successful study completion.

Future work includes conducting the planned study. Both the

described implementation techniques for smartphone sensor and

usage data tracking as well as the results from the future study

can be incorporated in a variety of applications. This includes, for

example, applications related to mobile health [

] or mobile social

networking [1].

ACKNOWLEDGMENTS

This work was done in the context of project DYNAMIC

(grant

No 01IS12056), which is funded as part of the Software Campus ini-

tiative by the German Federal Ministry of Education and Research

(BMBF). We are grateful for the support provided by Soumya Sila-

ditya Mishra, Sarjo Das, Oksana Kalynchuk, Michael Segert, Parth

Singh, and Bernd Louis.

REFERENCES

[1]

Felix Beierle, Sebastian Göndör, and Axel Küpper. 2015. Towards a Three-tiered

Social Graph in Decentralized Online Social Networks. In Proc. 7th International

Workshop on Hot Topics in Planet-Scale mObile Computing and Online Social neT-

working (HotPOST). ACM, 1–6. https://doi.org/10.1145/2757513.2757517

[2]

Felix Beierle, Kai Grunert, Sebastian Göndör, and Axel Küpper. 2016. Privacy-aware

Social Music Playlist Generation. In Proc. 2016 IEEE International Conference on

Communications (ICC). IEEE, 5650–5656. https://doi.org/10.1109/ICC.2016.7511602

[3]

Felix Beierle, Kai Grunert, Sebastian Göndör, and Viktor Schlüter. 2017. Towards

Psychometrics-based Friend Recommendations in Social Networking Services. In

2017 IEEE International Conference on AI & Mobile Services (AIMS). IEEE, 105–108.

https://doi.org/10.1109/AIMS.2017.22

[4]

Sajanee Halko and Julie A. Kientz. 2010. Personality and Persuasive Technology:

An Exploratory Study on Health-Promoting Mobile Applications. In Persuasive

Technology. Vol. 6137. Springer, Berlin, Heidelberg, 150–161. https://doi.org/10.

1007/978-3-642-13226-1_16

[5]

Rüdiger Pryss, Manfred Reichert, Berthold Langguth, and Winfried Schlee. 2015.

Mobile Crowd Sensing Services for Tinnitus Assessment, Therapy, and Research.

In Mobile Services (MS), 2015 IEEE International Conference on. IEEE, 352–359.

https://doi.org/10.1109/MobServ.2015.55

3http://www.dynamic-project.de