Intelligent Messages: Study

Goals of the study

The goal of this study is to collect data about users’ mobile instant message behavior to learn models, which are capable of predicting opportune moments to notify users on received messages. This study is conducted as part of the project Intelligent Messages. Its aim is to develop an intelligent message management service, which delivers messages at non-interrupting moments to maximize user experience.

Research methods

If you are using the application MSGSTAZ, we will collect data about your mobile instant message behavior. Concretely, we record information about your context (location, time of day, etc.) as well as the sender of a message and its content. Your data will be used to train machine learning models. It is collected anonymous so that it is not possible for third parties to associate you with your data nor identify the sender of a message (see data protection for more details).

Conditions to be met for participation in the study

There are no explicit selection or exclusion criteria for participants other than country specific age requirements on Google Accounts.

Advantages and disadvantages for participants / risks

This study does not carry any risks or disadvantages for you, as data collection is anonymous and data values are depersonalized (see data protection for details). By participating in this study, you contribute to research and can use MSGSTAZ free of charge.

Source of funding

This work is supported by Swiss National Science Foundation, under grant UFO200021L_153644.

Compensation / reimbursement

You do not get compensation for participating in this study.

Right of withdrawal

As a participant, you have the right to withdraw from the study at any time without needing to specify any reasons nor facing negative consequences. You can withdraw by uninstalling MSGSTAZ. If you want to delete the data collected with your device, contact the experimenter (see bottom) to initiate the process.

Data protection

Communication between the application and the server is encrypted using the HTTPS-protocol. In the following table, we list the data we will collect from your device. We explain how we depersonalize privacy-critical data and how we anonymize data collection.

Data Description
Participant id To identify a user, we use the Instance ID that is unique for each Android device. From this id, it is not possible to infer your real identity nor your Google account.
Sender A sender is encrypted and stored as a unique non-invertible hash.
Message On your smartphone, a message is transformed into a word histogram containing each word and its number of occurrences in permuted word order. An example of a word histogram could look as follows: {amazing:1, an:1, house:1, it’s:1}. In addition, messages get depersonalized by deleting names from a message's text. Names are identified by using lists of common first and last names. Message logging can be deactivated.
Group A group is encrypted and stored as a unique non-invertible hash.
Messenger The name of the messenger is stored, e.g. “WhatsApp”.
Timestamps Various timestamps are stored. For instance, the time a message was received, the time the user reacted on it, the time the user last used the phone, etc.
GPS position If you allow MSGSTAZ to access your device’s location, we store latitude, longitude and the accuracy of the last GPS measurement. We shift latitude and longitude values by a unique random number generated on your smartphone so that it is not to infer real-world GPS locations. Position logging can be deactivated.
Activity Your current activity is stored. It is inferred via Google's API for activity recognition.
Ring mode and volume The current ring mode and volume of your smartphone is stored.
Calendar If you allow MSGSTAZ to access your calendar, we store a flag indicating whether you currently have an appointment and the time until your next appointment. Calendar logging can be deactivated.
Pending notifications The number of currently pending notifications is stored.
User reaction Your reaction towards an incoming message, reading or dismissing it, is stored.

Insurance coverage

Possible damages to your health, which are directly related to the study and are demonstrably the fault of ETH Zurich, are covered by the general liability insurance of ETH Zurich (insurance policy no. 30/4.078.362 of the Basler Versicherung insurance company). However, beyond the before mentioned, the health insurance and the accident insurance (e.g. for the way to or back from the study location) is in the responsibility of the participant.

Ethics approval

The ethics committee of ETH Zurich (EK 2017-N-15) approved this study.

Contact person

If you have any questions regarding the study please contact Christoph Gebhardt.