Page 288 - Special Topic Session (STS) - Volume 2
P. 288
STS493 Sofie d.B. et al.
cannot. Examples include detailed sleeping patterns, calorie usage, more
detailed activity tracking or electrocardiography. Typical sensors that are
included in wearables are: motion sensors, Bluetooth, GPS, heart rate, screen
and vibration. Less common sensors are: Blood oxygen levels, camera, light
sensor, magnetic field, NFC (near field communication), speaker, thermometer
and Wi-Fi.
3.2 Existing sensor data
Sensor data may already be collected and owned by various parties and
not necessarily for statistical purposes. We term these data secondary sensor
data. We distinguish data based on user-owned sensors and data based on
the public Internet-of-Things (IoT) sensors (i.e. not owned by private users).
Although a clear distinction is hard, we view privately owned IoT type sensors,
such as weather stations or burglar protection systems, as user-owned
sensors. So the main distinction is between individual and public use. In this
project, we consider linkage of secondary sensor data to samples of a target
population after respondent consent.
Within user-owned sensors, a distinction can be made between
commercial and non-commercial data collectors. A range of companies
produces apps for mobile devices, in particular wearable devices, to provide
paid services to individual customers. Other companies, such as Google, also
provide unpaid services in exchange for the right to use the resulting data for
commercial purposes. These services are usually continuous and,
consequently, create a dynamic, vast stream of sensor data. The resulting data
are mostly based on self-selection, i.e. the initiative is with the user, and not
based on invitation. Non-commercial parties may collect mobile device sensor
data for research or policy making motives. Such data collection often has a
finite time horizon and is invitation-only. In all these cases, however, the sensor
data are self-initiated by users, but the data is stored, handled and owned by
others. Apart from privacy and legal constraints, such user-owned sensor data
can potentially be linked to individual respondents in surveys. After linkage, a
hybrid data collection follows where part of the data may be asked through
questions, part of the data may be measured on respondent mobile devices
and part of the data may be linked. An example is activity tracker data owned
by a third party linked to persons in the sample supplemented by survey data
on health perceptions and health determinants. Another example is budget
expenditure diary data linked to supermarket scanner data and supplemented
by scanned receipts for other types of stores. IoT sensors are usually owned
by government authorities and implemented for very specific purposes, such
as monitoring of weather, pollution or traffic. The sensors are attached to
objects, often have a fixed location, operate continuously and measure
activities of multiple persons. Data is stored, handled and owned by the same
277 | I S I W S C 2 0 1 9