1 Introduction
The Internet of Things (IoT) has arrived in the home. More and more everyday objects are sensor equipped and connected to the Internet at large. These smart devices are meant to offer novel interactions and possibilities in the home, e.g. more comfort, more security, more safety, or more efficiency. For example, the market for the »smart home« proposes a plethora of seemingly smart thermostats, door locks, or other remote assistants. Still, sensors are getting smaller and less expensive and are increasingly used to make the most mundane objects such as trash cans »smart« . Energy-efficient, networked sensors with a size of a few cm3 are already in development since some time [1], making the vision of microscopic small, cheap, and ubiquitous sensors that resemble »smart dust« [2] not a distant utopia. It comes to no surprise that HCI and design research on IoT for the home has seen an upturn in recent years. There is, for example a flourishing discourse around aspects of user integration and improved usability of smart home products [3]. With the wave of mass market IoT products ahead, HCI and design scholars are seeking to understand the social implications of IoT in the home and to develop frameworks to value privacy, data security, trust, and agency in such socio-technical systems [4].
While participatory design has a proven record of contributing to understanding use and context of future interactive systems together with potential users, comparably little is collectively known about how to involve people into actively designing IoT for the smart home. Still, HCI and design research explore a variety of novel ways to involve potential users in designing and understanding IoT in the context of the home. A number of analogue design card games exist to foster ideation and scenario building for IoT in the smart home [5, 6]. Also, digital tools to quickly prototype smart things and services for the home certainly exist: Yet, they either require smart home products to begin with. For example, the popular IFTTT platform allows to automate sequences of triggers and connected functions [7]. Yet, most of these tools only allow to prototype technological breadboard constructions without much possibility to actually test them in the home together with users.
This comes as a surprise, because IoT technology bears the potential to manufacture small, cheap, and ubiquitous sensor and actuator tools with relatively ease. Research on IoT for the smart city has shown that IoT toolkits are valuable research tools, not least because they are technically functioning and thus can be used to investigate use and context together with participants. Such research devices usually consist of simple sensors that are wirelessly connected, some form of local or cloud computing and displays or actuators as output device [8, 9]. We will review related work in Sect. 2.
As we have been dissatisfied with the scope and functionality of toolkits for participatory research in the smart home, we developed a toolkit to involve people into designing and understanding use and context of IoT in the home. We call this combination of an explorative device and methods of user involvement »Sensing Home« . The main goal of the toolkit is to work »out of the box« . We wanted it to run on simple sensors that resemble the principle of the IoT paradigm of »small, cheap, everywhere« at its core. We rely on the TI SensorTag, which we heavily modified both in firmware and exterior. The toolkit is also ready to use, as it contains a tablet computer for providing visualizations of sensor data, internet connectivity out-of-the-box and computing on our own hardware. The design rationale and technology of this toolkit will be described in Sect. 3. The toolkit allowed us to engage people in a range of participatory workshops and studies. The several modes of participatory exploration enabled by »Sensing Home« will be illustrated by three use cases, which we present in Sect. 4. The first use case reports how people used our toolkit to explore usage scenarios and to develop custom sensor applications within their homes. The second use case describes an interdisciplinary class, where students appropriated Sensing Home to develop and conduct empirical in-the-wild studies of smart sensing in the home. In the third use case our Toolkit was deployed in households to explore and to make sense of collected sensor data together with inhabitants. With a subsequent reflection in Sect. 5 on using Sensing Home we draw out themes for improving such toolkits as research artifacts and methods for future studies on the IoT in the smart home.
2 Related Work
The IoT for the home is a particularly challenging topic for design. It combines the tangible materiality of the home with the intangible materiality of data, services and networks. Also, the home is a particular sensible private area, which is loaded with personal meanings and values. As we see new forms of interaction emerging between these (im-)material configurations, questions around configurations of future use and context arise. Because of this, the design space of IoT in the home may be best explored in close participation with those affected by it [10]. HCI and design research have thus proposed plentiful design research artifacts and methods to involve people into exploring the design space of IoT in the home. These research artifacts and methods include analogue design card games, digital prototyping platforms and working prototypes deployed and evaluated in context.
Analog IoT cards are a common research device to quickly ideate and prototype design scenarios. Clustered in categories and equipped with simple rules, such cards have a proven record in nourishing creativity in participants [11]. KnowCards are a prime example of such design cards for the IoT. They present the technological elements of IoT products in four categories (power, connection, sensors and actuators) which then can be expanded with sets of actors and interactions. Thus, knowCards can support ideation for multisensory interactions and environments [5]. Another example, Tiles Cards consist of »primitive cards« for the basic concepts of inputs and outputs, and »game cards« that define the dynamic and rules how to use the »primitive cards« [6]. The advantage of these analog design cards is that they do not bind the imagination of designers, because they abstract complicated technological components like sensors or actuators and other properties like places, things, situations or dynamics to simple cards. However, these cards do not enforce rules nor point out invalid combinations. Thus, physically impossible combinations may go unnoticed. This calls for high abstraction abilities and background knowledge from involved participants. Lastly, they are by their nature non-functional and require subsequent prototyping. Digital IoT tools provide actual functionality and allow users to experience how sensors and actuators behave. By this, they require much less abstraction abilities. The »littleBits« tool is a good example of such a kit. It consists of electronic functional components for power, sensors, outputs and additional connectors [12]. These components can be easily combined by a magnetic connection system and allow a relatively easy way to design working IoT prototypes. A plethora of similar tools exist. ConnectUs [13], WoTKit [14], or Cube-in [15] are designed to teach connecting and programing sensor and actuator combinations. While these tools teach creative and functional combination of input and output they either need some technical or programming skills or come—in the case of littleBits—with predefined and thus limited functionality.
The abundance of ideation tools for the IoT in the home may be one of the reasons, why even the most mundane thing in the home—like a trashcan—has been smartified [16]. In contrast to this, the lack of toolkits for participatory design work together with people is rather surprising. Especially, because a lot of such sensor toolkits exist for participatory exploration, appropriation, and evaluation of sensor data in the smart city. Sensor toolkits for the smart city provide the means for citizens to explore their urban surroundings with the help of IoT sensors. For example, Smart Citizen Kit [8] enables users to collect data and to measure, understand, and compare sensory qualities of their city. It consists of sensors to measure air composition, temperature, humidity, light, and sound, as well as data-processing, data-transfer, and battery. Users can place it within the city to explore issues like sound pollution or air quality. Data from all users is accrued and displayed together with its anonymized on a website. As such, users can connect and reflect on such issues. Another toolkit for the smart city is the Air Quality Egg. It also contains sensors to measure the air quality and to display accrued data on a website [9]. These sensor toolkits for the smart city tend to focus on critical issues by combining the data from a large number of sensors to a given context like air quality. Yet, they also provide the freedom to explore several more issues in combining both data from the included sensors and from those sensors employed by other users.
3 Conceptualizing »Sensing Home«
Inspired by this gap in research on sensor toolkits for the home and the availability of sensor toolkits for the smart city, we wanted to create a similar toolkit for use in the home. This toolkit should allow participants to independently collect sensor data in different locations in their home, to observe the collected data, and to further process and annotate it. We wanted to focus especially on the data of simple sensors that are typically small and cheap, e.g. for light, temperature or humidity. Complex and a priori critical sensors, e.g. cameras or microphones, should not be used. Our goal was involving people in the examination of sensor data from their homes. Our system should work without prior experience with smart technology in the home. Therefore, we designed our toolkit as a self-contained system without the need for additional infrastructure, easy to set up and easy to use. Our toolkit does not rely on participants’ internet connectivity nor on 3rd party cloud storage or computing services.
We chose a sensor platform that corresponds to the IoT paradigm of being small, cheap and versatile. Our devices represent the most common functions of IoT products for the home: (simple) sensors/actuators, power, computing and communication. We explicitly chose a device with several simple sensors on-board. That gives us a high number of possibilities how to use the devices. Also, in contrast to most commercial smart home products available, we chose a platform that allows us complete control over data flows. The toolkit contains several wireless sensors that are connected to the internet via a gateway. We explicitly designed our whole system to handle all data only on our own hardware and software. Thus, the sensor data is transmitted to a server of our research group. The toolkit is also equipped with a pre-configured WiFi hotspot for instant use without configuration. We modified the outer shell of the sensor platform in order for it to be un-specific and non-descript, in order for participants to question the inside. As such we iterated with a variety of shells before settling on a final enclosure. In order to gather feedback on the initial iteration of Sensing Home, we used various outer shells for the sensors that do not reveal what they entail. Further, we presented this first prototype and a somewhat fuzzy notion that it contains sensors for the home to various people, from peers at CHI workshops to potential participants. This open ended narrative helped to engage with people to inform us on potential use cases: What would people like to sense in their homes? Where would they hide, attach, connect different sensors? Based on this feedback we settled on a 3D-printed color coded shells that reveal the front of the sensor platform and that simultaneously allows for easy attachment to a variety of things. The development of our technology was done in a process of field trials and learning. With each field use we evolved our probe pack from a working minimal version up to the preliminary final setup. We optimized in this process the technical functionality as well as the visualizations that the participants use to analyze their data.
3.1 Technology
Sensing Home Toolkit consisting of components for collection, processing, transmission and viewing of sensor data (3 SensorTags, Raspberry Pi, WiFi hotspot, tablet pc, etc.) with material for documentation and data work
We advanced the SensorTag firmware and central host side software to make them more versatile and more usable for our field studies. This includes first of all improvements for higher reliability, availability, and battery runtime. With the initial stock setup we reached 1.5 days of runtime with all sensors enabled and reasonable sensor intervals. This might be enough for some usage formats like an ideation workshop but it is not enough for using the devices in a field study. Therefore, we had to implement some energy optimizations that allowed us to use devices in typical smart home scenarios with up to three weeks runtime. We realized more flexible sensor intervals and on device data-preprocessing (static and dynamic thresholding) to let the sensors react as fast as possible while still saving energy. Despite the SensorTag software being open source it is not as easy to use as e.g. Arduino. Even a simple change for a meaningful improvement in the software has a steep learning curve and requires deep understanding of the system internals.
To capture interesting domestic activities, it is necessary not only to use the right measurement metrics but also suitable measurement intervals and sensor position. Based on our own experimentation and other work [18] we preset meaningful sampling intervals. We set the sampling for typically slowly changing measures (ambient and object temperature, barometric pressure, and relative humidity) every 10 s, for faster transient measures (light) every 2 s, and externally triggered events (accelerometer, gyroscope, and magnetometer) every 0.1 s for 10 s once triggered by motion. Nevertheless, it is possible to tune all sensor parameters for certain specific usage scenarios beyond the named limits. For the sensor position we chose different approaches in the individual use cases described later. In general the limited range of the SensorTags must be taken into account. In our experience it is good enough for normal sized flats with a single edge gateway. With the help of additional edge gateways it is possible to cover even large homes/houses. We include a Raspberry Pi 3 (Raspi) as an on-site edge gateway as it offers connectivity via BLE and (W)LAN as default. Each Raspi can connect up to eight SensorTags (limitation of the BLE stack) but normally we provide only three as a compromise of flexibility and reliability. Node-Red serves as an easy to use and powerful IoT mashup software on the Raspi. We implemented advanced software modules to use the SensorTags with new functionality and better energy efficiency. Our Raspi-Portal allows for a quick headless setup. We also included pre-configured mobile WiFi hotspots for Internet connectivity. The Raspi forwards the data to a secure server in our department for storage (InfluxDB) and visualization (Grafana).
Screenshot of data visualization in graphs as seen by the participants on the tablet
Besides the storing and processing of the the generated sensor data an additional management on the server side is absolutely necessary when deploying more than just a few devices for a single user. This includes an account and device management. Every field use of the devices requires some kind of reinstanciation of the software on the Raspis as well as on the server side to keep the system safe and secure. This includes all account data and passwords as well as all cryptographic keys for transmitting and storing the generated data.
4 Sensing Home in Use
We conducted several participatory field studies for exploring sensors in the home together with users from various groups. With a focus on the diversity of goals and outcomes of these studies we will subsequently report on three such use cases. The first use case focuses on the exploration of potentially interesting applications of sensors in the home. Here, computer science students, staff of a computer science department as well as elderly volunteers critically explored possible scenarios of using our Toolkit within their homes. The second use case presents an interdisciplinary teaching project. Here, computer science and social science students collaborated in order to design and subsequently conduct user studies in different social settings. The third use case we report on is about sensemaking and empowering people in understanding the potential and pitfalls of potential sensors in their home. Here, we deployed Sensing Home within nine homes and explored the data together with the inhabitants in order to make sense of sensor data collected in their homes. While all three use cases employed the very same Sensing Home toolkit, they allow us to highlight various themes of participatory exploration. In particular, we will show the versatility of such a toolkit for participatory research on the smart home. It allows for an empirically grounded exploration of IoT in the home and allows for participatory design work with participants of diverging technical literacy and fluency. Participants with some technical expertise came up with highly creative applications for sensors in the home. Participants with a only little technical expertise were significantly empowered in understanding the gains and risks of sensor data on the home. Students in turn were able to independently design and conduct empirical studies in the context of the home. Collectively, the diverging themes of participatory exploration unraveled in the following use cases shed some light on the values of doing research on the smart home together with people.
4.1 Use Case 1—Exploration of Usage Scenarios
Inspired by our own previous work on participatory exploration for the IoT in domestic contexts [22, 23] we wanted to investigate themes and applications emerging from the free and prolonged exploration of simple sensors in peoples’ homes. In this use case our participants were free to use the sensors where and how they wanted to. In this free exploration the participants could experiment or simply play around with devices and data. We saw it as important for this use case to happen in the real world context of participants homes. This study needed also to run long enough for participants to really experience the usage of sensors in their daily routine. The actual real sensor usage and data collection in such a context allow a direct validation of scenario ideas in means of feasibility and relevance. Despite different levels of literacy and fluency our toolkit enabled the participants to find some interesting, innovative and unexpected usage scenarios. The participants even gained literacy and fluency in using the sensors and reflecting on their meanings. The findings of context and implications of use helped us to develop and enable other use cases.
The goals of this use case were two-fold. First, we wanted to find out whether participants were able to use the sensors and their data meaningfully at all. Second, we were interested whether and how participants might come up with innovative and unexpected usage for the sensors if we engaged and fostered a free exploration. An instant camera and a booklet complemented the deployment of our sensor pack, as we wanted to focus on the experiential qualities of a cultural probe study [24]. The booklet guided participants through the first steps of the exploration with an example scenario and offered templates for structured notes of their own exploration (similar to [25]). Questions included: What do you want to try out with the sensors? What sensor values revealed certain insights? Where did you place the sensors? What does the collected data actually show? Participants were also asked to annotate graphs and document placement of sensors with photos. We conducted this use case consecutively with three groups of participants. Group A consisted of 3 master students. Group B consisted of 3 computer scientists of our department, not involved in the project but with solid technical background. Group C constituted 4 elderly volunteer users with no special technological knowledge or skills. Groups A and B set up their probe packs at home on their own, for group C this was done by one of our researchers. The participants used the sensors for about two weeks in their homes and documented their work during this time. After this time all parts were returned to us. When an exploration and data collection phase was finished we reviewed the notes of the participants. We invited them groupwise for a closing discussion to talk about their experiences and insights. In this use case the generated sensor data itself was not intended for further analysis by us, especially because the sampling context was quite diverse. We analyzed the usage reports to improve the technical feasibility and setup which also includes usability aspects.
The individual sensor exploration and data work of the participants worked very well. All participants were able to browse their data and to explore data patterns meaningfully by the provided visualizations. As expected, the participants found the visualizations of volatile values easier (e.g. light or acceleration in movement). Participants liked these measures as they associated them with »motion« and easy to distinguish changes. But also slowly changing values (e.g. temperature and humidity) were meaningful. We see some differences in literacy and fluency between the three groups. On the one hand some participants of group C had sometimes problems in understanding and interpreting even simple sensor values, on the other hand we see advanced and innovative approaches to use the sensors. Overall the different sensor functions are usually well understood. However, often movement of the sensor is mixed with the movement within a specific room. The relationship between a measurement and the according generated data is generally understood in a fundamental way.
Most participants found usages for the sensors for own purposes beyond scenarios introduced by us as examples. Some of the created scenarios were obviously and expected while some other were highly creative and even surprising. We want to illustrate this by some selected examples in different categories that we identified. Mould prevention was an ever repeating theme including reflecting on own behavior and seeking for optimizations, e.g. for the drying of laundry in the home. The determination and improvement of air exchange in the home by measuring temperature and humidity was also performed multiple times. One very popular theme was the augmentation of mundane household objects and devices to make them »smart« . One participant tried a sensor on a vacuum cleaner to detect operation and load via motion and temperature during vacuum cleaning. An augmented teapot should notify the right time for drinking the tea by measuring its temperature. The augmentations also involved furniture and home installations. A sensor was attached to a bed to monitor sleep and attached to a door to monitor movements (when someone enters or leaves). A sensor near to the radiator revealed the switch off and on time of the central heating system (landlord owned and controlled) by monitoring falling or raising temperature. Even a toilet lid was made »smart« to measure when and how often the toilet was used. The participants attached the sensors not only to objects. One sensor was even attached to a baby to monitor activities, sleeping times and positions. One participant made own little helpers to use the sensors in different scenarios, e.g. a small Lego stand or a velcro sleeve to attach the sensor to textiles.
Sensor mounted on the top lid detecting the filling level of an aquarium by measuring the reflected light
Sensor on a small stand made of Lego by the participant to bring the sensor in optimal position in front of the tv set
A playful exploration of sensors and their data goes beyond the possibilities of an ideation that often only theorizes about what a sensor is useful for or where its limit are. We have seen examples where the exploration allows to experiment with the sensors to verify assumptions as well as examples of serendipity where the playfulness lead to unanticipated results of possible sensor usages. Our toolkit worked very well in means of supporting creativity and satisfying human curiosity by exploring the possibilities of usage. Not only the participants with more knowledge and higher skills ideated and tested interesting scenarios.
No matter how good the initial literacy and fluency on the usage of the sensors and interpretation of their data initially were, the participants gained competences and understanding during the usage. This encouraged us to use the sensors also for other use cases. This comprises e.g. further usages in student projects as well as focussing on the critical reflection of these devices in the home in means of privacy threats and surveillance.
4.2 Use Case 2—Researching and Prototyping in the Wild
We used the toolkit in an interdisciplinary teaching and learning project, which was conducted in two iterations with students of computer science and students of cultural sciences. Based on our toolkit, the students developed their own research interests and empirical case studies for Internet of Things applications in different social contexts, e.g. shared flats, co-working spaces, bars or even horse stables.
»HCI in the Wild« was first held in spring 2017 with 15 students of Chemnitz University of Technology and University of Leipzig and repeated in fall 2017 with 12 students of Leuphana University Lüneburg. The aim of the project was to let students experience both the opportunities and challenges of research and development of IoT » in the wild« . Therefore the students had to form mixed groups to choose a research interest and a method to approach the usage of our toolkit in a self-chosen social environment. The students’ projects ranged from cultural probe studies [24] to exploring living worlds like shared-flats, up to the development of own functional prototypes supplementing our toolkit. To enable the students to explore and adapt the toolkit for their own purpose, we provided a range of didactic means. Firstly we established a mutual base of knowledge by reading and discussing research literature on smart homes (e.g. Tolmie et al. [21]), research in the wild (e.g. Kuutti and Bannon [26]) and methods (e.g. Graham et al. [27]). Secondly we ran hands-on workshops with the toolkit and the data visualization by Grafana to empower all students to adapt the technology as well as the data to their research interests. Followingly the student groups started their own research projects which we accompanied through weekly consultations and the students’ online » research diaries« . This version of a learning portfolio was structured by milestones that instruct a small-scale research project, e.g. documenting research interest, developing a research question, choosing a method, contacting people in the field, gathering data, etc. Thereby the students’ projects could be supervised and mentored by us without determining upfront, what exactly would happen with the toolkit.
In the result, we were awarded with a variety of concrete use cases, explorations of diverging application contexts and even empirical user studies designed and conducted by the students. Before providing examples of this work by the students, we want to highlight the generative quality of the toolkit as proven in this use case of researching and prototyping in the wild. Even undergraduate students with background in humanities were able to use the toolkit as a technological basis for comparatively complex empirical projects within a time frame of just 10 weeks. We were not only impressed by the productivity the toolkit enabled, but also by the quality of inquiry the students conducted with it. Whereas we learned through use case 1 to narrow down specific sensor and data visualizations combinations to enable an explorative use of the toolkit, we learned through »HCI in the Wild« how it proves within different social contexts of use and diverging research perspectives. We will see in the examples that the interdisciplinary student groups did not just apply the toolkit, but came up with own modes of researching the social IoT.
Sensor equipped dog
The monitoring of animals links to the second dominant category of use cases the students came up with, the monitoring of behavior. Other than in checking rather passive values like air pressure, the monitoring of movement or action has to deal with meaning and interpretation. This has to be done when data is collected—not only when the students researchers finally analyze the data. An example for such projects is the self-monitoring of a student, to track her productivity over the day by operationalizing different spots in her flat as places of leisure or places of work. The research question as well as the toolkit itself implied that she reflectively explored her everyday routine in a rather ethnographic manner. Another group equipped a co-working space with multiple SensorTags to measure data like light values and humidity in order to correlate it with the co-workers perception of well-being and productivity, which was gathered by a daily questionnaire the participants had to fill out.
Handing the toolkits to students with a rather open-ended narrative, to use them for their own interests, has not only led to interesting use cases. Whereas the basic purposes of the SensorTags as device is to monitor environmental conditions, the students projects showed that the designed toolkit provokes deeper engagement with IoT in social contexts. Firstly, all projects included an explicit analysis and reflection of the »application context« , which was not derived from a given task, but following the criteria and relevances of the researched life-worlds—even if it is due to the relevance of the students themselves. Secondly, all projects included an additional form of (mostly qualitative) data gathering and analysis. Those ranged from ethnographic writing, over questionnaires, up to interview studies, which were in some projects even combined. Even the owner of the tomato plant was interviewed, whether she thinks the monitoring helped her and/or the plant to thrive.
The interdisciplinary teaching project »HCI in the Wild« confirmed two major implications of social IoT we wanted to address with the design rationale of our toolkit. Above all, IoT devices in general and research artifacts in specific have to be open for appropriation by the people using it. An IoT product sensitive for and meaningful in social contexts like the home cannot be derived from an upfront defined problem-solution pairing. It should rather provoke and enable people who use it to question such pairings and/or develop their own. Secondly, the student projects highlighted the importance of data work. The students experienced that all data can become relevant if put in in a particular context, and that this context needs to be brought up when interpreting IoT data. Especially human behavior and its traces cannot be gathered and understood without knowledge about these practices, their intention and their context. This importance of data work for the social IoT encouraged our third and final use case.
4.3 Use Case 3—Critically Reflecting Implications of Social IoT
Encouraged by the participants’ and students’ appropriation of the toolkit, we decided to deploy it as a research artifact in homes for longer periods. Within a series of three field phases we conducted studies using the toolkit in nine households in two mid-sized German cities. The main focus thereby was to understand the social implications of IoT data and its visualization for the inhabitants and to enable reflection on privacy threats and surveillance through IoT in the home.
For this third use case the toolkit was gradually developed further. Especially the deployments of the students helped us to improve the technical robustness e.g. by enabling secure remote management and recovery from undefined states caused by various influences. Furthermore we had to create a methodical frame to enable the use of the toolkit as a research artifact to evaluate how participants deal with IoT data in their home. Therefore, we narrowed down the participant’s instruction to use the toolkit compared to the very free explorations in use case 1 and 2. Mainly, we predetermined the position of the sensors in the homes (on hall door, on fridge door and opposite to tv set) in order to achieve comparability between the cases and to promote the interest in the data. When defining those spots we built on the findings we obtained earlier, especially to generate data, that shows traces of everyday activities, rather than focusing on environmental conditions. When using the toolkit as a research artifact, we thus initially restricted the application by the participants, but simultaneously left open, how the data could be interpreted and encouraged the participants to find their own purposes for it.
Altogether we ran a series of three field phases over the course of eight months in 2017 with participants living in different household-constellations, ranging from a 74 year old living alone to a family with two teenage sons and two dogs. We introduced the toolkit to them as ready-made artifact in the manner of a probe pack [24, 27, 28] The ready to use aspect was very important as we were primarily not interested in usability aspects but the social situatedness of IoT data in the home. The study concept did not only consist of our toolkit and the request to document the data but also of a group discussion format to reflect on the data.
The first part of the field phases was the actual data collection in the homes, which took 10–14 days depending on the participants’ individual schedules. In the beginning of a data collection phase, a team of two to three researchers set up the toolkit in the home and explained the data visualization software. During this phase, the participants had full access to the data visualizations and were encouraged to perform »data work« , by browsing and selecting interesting graphs on a daily routine. In a second step we conducted group discussions with the participants from different households. Thereby we would not only be able to gather insights into individual experiences [21] but in the collective dimensions of making sense of data from the home. We hoped for a discussion that benefits from multiple participants’ experiences with the probe packs and the individual data work. We also wanted to empower participants to reflect critically on IoT in the home by choosing such a collective discussion method. We called the format »Guess the Data« .
In preparation of all three sessions we performed data work ourselves. We browsed the data as a researchers’ group and looked for interesting patterns, whether recurring (regular activity) or outliers/anomalies (special events). We used our experience on working with the sensors and their data, as well as our everyday knowledge of common activities to interpret the data. Thereby it was inevitable to include knowledge about the context and the situatedness of the data, gained from previous contacts and home visits. We selected 10–12 data sections per discussion session. We printed the selected data sections on large format paper sheets. We used the Grafana graphs format that the participants were already familiar with. We anonymized the graphs by removing all markings that directly hinted at the creatorship of the data.
Guess the Data—participants discussing the printed data graphs
The mechanisms and horizons of participants’ sensemaking of their domestic data we observed corresponds with similar studies for individual participants [19, 21]. Daily routine and implicit, situated [30] knowledge become a backdrop for interpreting data from the home. When discussing the individual data work in the group, especially sensitive activities and behaviors became visible. Even simple sensor data can become critical for residents’ privacy not only by pointing to possibly sensitive information like sleep rhythm or wasteful behavior. Our study design showed furthermore the critical dimensions of collective sharing and sensemaking of such data. Eventually the participants realized the surveillance potential of gathering data in the home by own practice: The active usage of the sensors and data work led to forms of surveillance of other family members.
»He said ‘I have been in the garden all the time’. And there I laughed and said: ‘This cannot be true, because the apartment door only opened at 17:30.’ And he said ‘Really? […] Were you watching me?’«
It becomes apparent that the participant used the sensor system asymmetrically. During the group discussions it became clear that all of the participants understood that this system could be used to surveil partners or children.
The use of the toolkit in homes shows its generative potential in hindsight to a critical reflection of IoT technology. The participants proved and realized that simple environment sensors can reveal a lot of sensitive and personal information. Combined with situated knowledge about the housing situation this data can be used to identify a certain person and recognize their activities within the household. Combined with the discussion format »Guess the Data« the—sometimes ethically problematic—implications of IoT in the home became tangibly aware to the participants and us as researchers.
5 Results
We built a functional IoT research device for participatory exploration of smart sensors in the context of the home. Using it in three different modes with people encouraged our initial design rationale, to enable an empirically grounded exploration of the design space for IoT in the home. The tangible exploration of usage scenarios proved that participants with diverging technical literacy and fluency were able to use the toolkit in their own means. Furthermore we could observe that the participants gained competences in data interpretation during the usage. This encouraged us to use the sensors also for other use cases. By using the toolkit as a research artifact for student projects, we tested its versatility and robustness for prototyping and appropriation in the wild. This exploration showed that the toolkit enables people who use it to question problem-solution pairings of IoT for the home and to develop their own. Use cases 1 and 2 highlighted importance of data work, while situated knowledge of practices in the home was topic of use case 3. Here the deployment in homes over two weeks sensitized the participants that simple environment sensors can be used to identify a certain person and recognize their activities within the household. We further developed the discussion format »Guess the Data« to gain insight into the participants’ evaluation of the social and ethical implications of IoT in the home.
Through these three case studies we have shown that Sensing Home allowed us to adapt the toolkit with relative ease to diverging methods and applications. All these in-the-wild studies could only be conducted because Sensing Home was technically functioning. The toolkit allows for independent studies for longer periods of time without having researchers present. It thus allowed participants to gain first hand expertise through usage and subsequently empowered participants to engage in informed and critical discussion about the potential gains and risk of smart sensors in the home.
Still, some of the ideated scenarios were beyond the technical limits of the current implementation, e.g. due to the limited range of the sensors. Evolved sensors might solve this issue and will allow the usage in the greater context of home in the (semi-)public space of the building and maybe even outside the building without the need of additional gateways. SensorTag are versatile, but sometimes not versatile enough. Sensor dimension and sensor placement are an issue for deployment. Yet, the possibility to add extension modules with more or specialized sensors for specific purposes needs to be explored. As well as longer unmaintained deployment of the sensors could be enabled by slightly larger housings that holds larger batteries for a runtime of several months. This would offer the possibility to make the devices part of the daily life of the participants without any need for attention during long term studies.
Acknowledgements
This research was funded by the German Ministry of Education and Research (BMBF) under grant number FKZ 16SV7116.