1 Introduction
IoT devices are rapidly finding their way into people's everyday worlds. With projections of 30.9 billion units worldwide [
126] and a global market of $1.6 trillion by 2025 [
127], IoT devices—and the data they collect, use, and share—are becoming particularly entangled in everyday lives, often in intimate and private spaces like homes. While there are a multitude of home IoT devices and an even broader range of homes they exist in [
35,
88], the interfaces created for home dwellers to engage their own data remain restricted by a narrow set of techno-solutionist and capitalist pressures. Home data are often presented (and understood by home dwellers) as a protective safety net [
32], as a compromise for receiving the services provided by IoT companies [
60,
67,
99], as a tool for self-knowledge [
60,
104,
121], or as a mechanism to save money or energy [
25], etc. Spreadsheets, apps, dashboards and smart recommendations associated with IoT systems are typically grounded in an objective view of data which ignores the deeply local, interpretive, and dynamic nature of data [
36,
60,
75], qualities often more attuned to characteristics of home environments. The challenge is that these goals and orientation of data, while perhaps desirable on their own, create a closed set of design possibilities and foreclose exploration of other possible encounters with data that may be less about productivity or objectivity.
Our work builds on a long tradition of research in human-computer interaction (HCI) and design that investigates how people conceptualize and live with their own IoT devices and IoT data [
25,
26,
66,
105,
122]. In particular, the Odd Interpreters (OIs) were designed in response to recent calls within empirical and critical research that call for diversifying data encounters [
32,
88,
100] in everyday settings. In their work on Diffraction-in-Action, Sanches et al. [
100] state “
Engaging data diffractively offers a way to understand data differently and reposition it as something that is lived, situated, and contextual, making designs that are closer to the entangled phenomena of being in the world.” In other words, looking for alternatives means looking for ways to bring data back together with the material conditions of its production and the social context of its interpretation.
In this paper, we present design work that opens the door to engagements with IoT data that are open-ended, experiential, and embodied. We discuss a series of design events [
89] that unfolded throughout the making of the Odd Interpreters (OIs). The OIs are research-through-design (RtD) artifacts which create experiences for people to encounter their own home data beyond questions of self-improvement, accuracy, or usability. We focus on the process of making the OIs because the process of designing ‘against the grain’ revealed important lessons for imagining IoT data encounters otherwise. Further, centering our process as a contribution of its own responds to recent calls in RtD scholarship for better documentation of the messy RtD processes, which can allow for a better understanding of the validity and rigor of the knowledge produced through the making of artifacts [
6,
14,
24,
33,
106]. The Odd Interpreters include
Broadcast, a wall-mounted device which plays ephemeral sounds to represent data's interwoven existence within broader infrastructures beyond the home;
Soft Fading, a rotating fabric cylinder which collects traces of sunlight to explore analog, slow and imprecise data collection; and the
Data Bakery, a system which transforms smart plug data into prints of cookie recipes to activate the dynamic and human labor intensive side of data systems (Fig
1). By emphasizing these alternative sides of data, we resist positioning data encounters within technosolutionist frames with the hope of creating a space for exploratory and experiential ways of being with data. This defamiliarization is meant to draw attention to data in home contexts and to encourage reflection about the broader assemblages of data and people that constitute home IoT systems.
Our methodological commitment is to Research-through-Design (RtD) [
47,
125]: the material making of our conceptual ideas lead us to discover generative frictions while designing with home IoT data as a material and against the grain when it comes to the role data typically plays. We combined inspiration from design works as well as theoretical works by Science and Technology Studies (STS) scholars, philosophers, and media studies critiques of technology with the rich trial and error of ‘making’ with data and IoT. Our intention is to reveal what we learned about working with home IoT data in our efforts to move away from conventional data encounters centered on objectivity and productivity. We report on the making of the OIs with an attention to the process of conceptualizing, making and refining the artifacts. This approach is akin to research journeys [
111] and design events [
89], in the sense that it focuses on the ‘through part’ of RtD [
33]. By purposefully exposing the messy side of our design process, our goal is to illustrate the complexities of working with data as material—in particular when aiming to design outside of conventional tropes with data.
Our contributions are as follows: (1) the OIs are unique design exemplars that pluralize and expand home data encounters, contributing to a growing repertoire of artifacts [
16,
116] that start to define that design space ; (2) we articulate design events that were particularly telling with regards to revealing data's invisible infrastructures and assumed objectivity; (3) we conclude with a discussion around the need for pluralizing home data encounters, the tactic of designing between illusion and precision, and a reflection on living with the prototypes while designing.
3 Methodological Approach
We see the OIs as artifacts that invite the experience of data encounters that are intimate, polysemic, and multi-sensory in home environments. Our goal is to create space for engagements with data that allow home dwellers to consider data as a multifaceted phenomenon that goes far beyond their home or the IoT service they use. We specifically focus on revealing data's material infrastructure and hidden labor, as well as data's entangled relation to the common imaginary of data as objective and productive.
Our methodological approach builds on the traditions of discursive design [
112], speculative design [
5,
40,
81] and research-through-design (RtD) [
47,
110,
125], but also responds to newer calls for making speculation more consequential and experiential [
23,
43,
117]. In this paper, we share our process of designing and making the OIs. We particularly focus on the ‘through’ parts of RtD, as Desjardins and Key [
33] note, to work through segments of our process that did not feel like a straight path yet yielded insights about how to design with IoT data as a material. The ability to discuss the messy, trial and error side of RtD is crucial when working against the grain of convention, as friction may be the most interesting part of the process. We also use Oogjes and Wakkary's idea of ‘design events’ as a scaffold for our analysis, as design events focus on particular moments within a project, without “
structure[ing] them by finished designs or samples.” [
89]. Our approach responds to recent calls for more transparent and honest documentation of RtD processes [
6,
24,
98] and clearer articulation of intermediate level knowledge within design projects [
6]. Within these calls, researchers have argued that rigor, in RtD, comes from a strong “chain of reasoning”[14,24,29], which connects design decisions towards the making of an artifact. While researchers have proposed methods for documenting RtD processes (e.g. critical journal / contextual portfolio [
98], annotated portfolio [
16], design events [
89], etc.), this form of dissemination for RtD work still remains rare.
In addition, we interweave stories of iterative making with moments of living with the prototypes in our homes while we were refining them, further foregrounding our actual RtD process. Similarly to Odom et al.’s approach of the designer-researcher [
87], and autobiographical design [
30,
33,
83], our approach allows us to acknowledge and interpret our first-hand experiences with the artifacts. While autobiographical design relies on designing for a researcher's genuine need, our approach is closer to Odom et al.’s idea of designer-researcher which harnesses a “
first-hand account of key insights emerging through our process”[87]. We chose a first-person research methodology [
30,
42,
76,
83] because it allowed us to immerse ourselves in and document the messy RtD journey we were on. It allowed for a constant awareness of the process as we continued tinkering, debugging, and making adjustments to the Odd Interpreters [
30,
83] while living with them. This offered an opportunity for our team to observe the designed behaviors unfold over long periods of time and to surface and work through unanticipated ethical considerations with the systems we designed [
103] (see 3.3). A first-person approach also satisfied our own curiosity: while we had spent months developing the Odd Interpreters, it was hard to imagine how it would be to experience them over time, as part of our everyday routines.
3.1 Project background and context
Our team at Studio Tilt, a design research studio at University of Washington, Seattle, USA, is composed of a professor in interaction design, three graduate students in design, information science, and engineering, and five undergraduate students in design. During the concept development and prototyping phases, our studio was split into three sub teams, each responsible for one OI. In addition to team work sessions, we followed the format of a Directed Research Group (DRG) [
114] where we met weekly for two hours to share progress, conduct critique and make decisions on next steps.
In addition to being inspired by STS literature broadening common understandings of data (as outlined in 2.3), we were intrigued by Desjardins et al.’s descriptions of novel data encounters [
32] as well as provocative views of what alternative IoT could be [
35,
66,
68,
71,
88]. From this scholarship, we chose three concepts as starting points for sketching:
Diffuse ways of noticing, which refer to slow, imprecise, and maybe unproductive ways of capturing data;
Imaginary leaps, which reveal the broader material infrastructures data traverse during their journeys to and from a home; and
Performative ways of knowing, which explores active ways of engaging the human labor behind data. We conducted three rounds of open and divergent sketching (Fig
2) adding up to over 100 sketches before doing a session of editing where we chose three concepts to develop:
•
Broadcast: using sound to depict imaginary leaps beyond the home, focusing on data travel.
•
Soft Fading: an analog data collection device to notice sunlight on faded fabric
•
Data Bakery: an embodied and performative act for translating data into baked cookies.
With three directions in mind, our studio spent multiple group sessions dissecting, contextualizing, and adjusting our concepts so that they could stand alone but be companions to one another. Our making process moved from sketches and cardboard prototypes to hardware, code, and 3D prints (Fig
2).
3.2 Documenting the RtD process
Following Research-through-Design (RtD) practice [
33,
47,
125], we documented the design process and the making process through weekly notes, photos of sketches and prototypes, videos of prototyped interactions, share-out presentations, and notes on debugging. After the prototypes were completed, we created a template for taking notes three times a week on our experiences living with the OIs (inspired by [
12,
69,
98]), including a prompt for writing and a space for images and captions (as per Mackey's commitments when doing first-person research [
80]). Along the way we shared our experiences and concluded with an exchange with the rest of our research studio to discuss our final impressions. Inspired by duoethnography [
101] (a type of first-person method that invites researchers to compare their first-person experiences via dialogue), we organized our experiences of living with the prototypes in parallel so we could create dialogues between our households. Duoethnography is based on dialogic relations between the lived experiences of two or more researchers. The intent is to juxtapose various voices to highlight similarities and reflect on differences in experiences.
Each discussion was recorded, transcribed, and analyzed alongside our written/photographic notes. Prior to living with the OIs, we sought approval and guidance from our university IRB board and used consent forms within our own team to make sure studio members were comfortable with documenting their at-home experiences. We also prepared a consent form with information about the project for the people living with us at home (partners, roommates, family members). While they were not interviewed or required to reflect on the OIs, we sought consent in case we discussed their reactions while living with the OIs or if they wanted to provide feedback on their experience.
To ground our experiences living with the OIs, Table
1 describes our team's positionality and living situations at the time of the study. As a team, our experience with IoT devices at home ranged from no devices at all to a small ecosystem of smart bulbs and outdoor camera. Our own experiences of home, data, and IoT are partial and subjective, and influenced how we designed and experienced the OIs.
3.3 Ethical considerations while working with personal and home data
In our design process, we were intentional about the way the OIs would collect home IoT data and aimed at emphasizing privacy and security in the systems we built. Soft Fading is simple; it does not connect to WiFi and does not collect digital data. Broadcast only records a 30 second snippet of encrypted packet metadata (the timestamp and the source and destination addresses) related to a connected smart speaker before deleting it. Data Bakery is a more complex system: data is recorded by using the third-party service IFTTT [
63] and a Google Spreadsheet. We investigated these services’ policies and consulted with our technology department to assess what protections and risks were involved when using these services. For both Data Bakery and Broadcast, we needed to collect a home WiFi's name (SSID) and password, while Broadcast required the MAC address for the smart speaker. Each of these data were saved locally on each device and deleted after engagements were completed. Finally, Data Bakery required remote login to the networked printer, which was performed by a studio member with consent from the home dwellers.
3.4 Analysis
We analyzed our templates and our discussions from living with the OIs, as well as our weekly meeting notes, in progress presentation slide decks, sketches, photos, and videos from our design and making processes. Our review process allowed us to trace back key conceptual, functional, and technical decisions and tease apart lessons learned when designing alternative encounters with IoT data. Throughout our analysis, we flagged data that helped us answer the questions: How do the OIs help imagine IoT data encounters differently? How do the OIs create friction with existing IoT systems and data structures? We conducted one round of thematic analysis [
20] to identify main themes and followed a process of open and then axial coding. Below, we present the Odd Interpreters, the design rationales that led our process and the design decisions we made as we were designing alternative encounters with home data.
5 Experiences in Making and Living with the Odd Interpreters
In this paper, we do not focus on the full process of designing and making the OIs. Instead, in this section, we highlight six design events [
89] which help us qualify our designed alternative encounters of home IoT data. As Oogjes and Wakkary state, design events allow us to stay with the “ongoing-ness and dynamic nature” [
89] of design research, qualities that are particularly relevant for our work, as the design, making and living-with were deeply intertwined. Further, this focus on the ‘through’ part of RtD [
33] is relevant to demystifying both the conceptual, but also material and infrastructural complexities of designing data encounters that go beyond a technosolutionist framing. It is through these precise descriptions that we demonstrate the validity and rigor of our RtD work, and that we offer solid anchors for others to build upon.
5.1 Richness in the process: Anticipation while capturing sunlight
One of the first things we noticed while living with the Odd Interpreters was a de-emphasis of the final data archive and a renewed attention to the active process of capturing data, which also came with a new sense of
anticipating data. It became apparent to us that data were being collected but we did not have access to them at all times. While some IoT services focus on always-available data, restraint (also see [
94]) proved to be a generative tool for bringing more awareness to data. A clear example of anticipating data is illustrated in Soft Fading. Since occupants are unable to see the effects of the sunlight until weeks or months later, they spent more time imagining what the fabric would eventually look like. For months Ruby, Audrey, and Chandler pondered on the subject of sunlight; its locality, its intensity, and the movements they made within their home. They questioned data, even worrying about their fabric's outcome, as Ruby said, “My biggest worry was that the sun would fade the fabric beyond any recognition, so much so that no lines of data would be present.” Further speculation involved directing attention to one's environment, as Ruby writes,
“
I had been imagining the fading pattern on Soft Fading to be sort of geometric, with gradients of stripes. But seeing this leaf's shadow makes me wonder about how my little desk environment will imprint its data onto the fabric? Looking more closely at this photo [ Fig 13], I notice that my plant's pot is reflecting light onto the window. I forgot that light can bounce around! And even the little folds and wrinkles on the fabric are creating shadows and patterns.”
By having Soft Fading withhold instant access to data it forced Ruby to confront ‘data’ as it was happening, long before seeing the actual imprint on the fabric (which we get to in section 5.4). Audrey had a similar encounter with data, one that happened when she was not even in the home. She wrote, “It was so hot during our vacation, the sun was really strong, it made me wonder a bit - how much is this intensity impacting Soft Fading? What will the fabric look like?” These anticipatory and imaginative moments cast data in a new light, one where occupants need to carefully consider the conditions in which data are produced to envision data coming into existence.
5.2 Revealing infrastructure: Unpredictable patterns with Broadcast
While most IoT devices aim at hiding background processes to prioritize end user desires and needs, with Broadcast we intentionally augment smart speakers with a new device whose only function is to show when data are traveling to and from the device. This idea offers an important departure from common narratives of seeing data as ‘in the background’ [
60,
75] and instead use this characteristic of data as a jumping off point to connect singular IoT devices to the broader (data) systems they exist in.
Both Audrey and Chandler described anxiety over “missing” sounds while they were out of the home, and rushing (sometimes running!) to the device when they heard the static or noticed the flashing light indicator. This combination of ephemerality and unpredictability led occupants to describe their interactions with Broadcast as “catching” sounds and data. While some sounds were heard clearly, other times “it was so quick. I only caught the last glimpse and then it was gone” (Audrey). This example expresses a sense of resignation: that nothing can be done to listen longer or catch a sound better, because the data represented has already moved on, and the intersection with the human world has disappeared.
We initially envisioned Broadcast would only actuate when an occupant interacted with a smart speaker—an assumption based on common promotional and instructional materials from Amazon and Google and shared with many users of voice assistants [
73]. However, implementing and debugging Broadcast introduced practical and technical revelations. During debugging, we found that voice assistants consistently and periodically send data packets, often to addresses related to the service provider (e.g., Amazon or Google) or to other network devices such as the router. While these were likely routine synchronizing activities, we also found some voice assistants send a packet after the device appears to have completely powered down. These discoveries challenged both our understanding of how networks and devices operate and required us to revisit Broadcast's design.
At home, this discovery made the constant communication of networked devices material by actuating at unexpected times. This uncertainty led occupants to periodically attend to Broadcast's light (Audrey: “giving little glances”) or to intentionally trigger Broadcast by interacting with the smart speaker. While Broadcast reliably captured packets after interactions with the smart speaker, many packets were not generated by human activity but by machines. Through Broadcast, occupants were able to both confirm that the smart speaker was sending packets without human activity, then to speculate at both the data and devices implicated in streams of “invisible” data (Chandler). Audrey was reminded that, even when out of the house, “the house is still there, doing its thing, sending and receiving data, no matter what we do”. Through both designing and living with Broadcast, the persistent motion of network traffic packets were made experiential.
5.3 Embodied and experiential: Making cookies for meaning making
Once data are collected, a question of ‘what to do with the data’ remains. It is often understood that data are cleaned, prepared, organized, and visualized to generate a certain meaning. With the Odd Interpreters, we aimed at making that process of manipulation tangible in interactions with the artifacts. The most obvious example is the manipulation of data in Data Bakery—once a recipe is printed, it is up to the user to perform the recipe to the best of their ability (often inserting new human interpretation or error in the process).
Data Bakery offers home dwellers a novel route for understanding, and manipulating, their home data. Even interpreting a new Data Bakery recipe first requires a step of manipulation, since without the context of the recipe book the print feels nonhuman and ambiguous. Numbers float intermittently across the page and are disrupted with two words only: the topping type and the seasoning spice. After receiving her first print, Jena wrote “The first thing I looked for was topping and mix-in type. I found myself placing more priority on their relative plugs than others. I laughed at raisins, because I really wanted a wacky topping and I don't hate raisins, but I don't love them.”
When placing the recipe into the book to reveal the correlation between data points and ingredients, Jena reflected “
What did I do to get half (0.51) an egg? Need to invent special tool for egg halving!” In her process of interpretation, Jena annotated the recipes directly in the book (Fig
14), to make sense of the strange quantities presented on the print, but also to make these abstract numbers into actionable instructions in the kitchen. While baking is often seen as a precise activity, it isn't at the level of precision required by the Data Bakery print. In her annotations, we can see adaptations: “
0.76-ish tbsp vanilla extract”, where the “ish” suggests an approximation. We also see replacements: instead of unsalted butter, Jena used salted butter because this is what she had at home. Again, the rigidity of the data is confronted with the realities of her home, forcing further manipulation and making tangible the process of data cleaning or adjusting.
5.4 Aiming for Softer Edges: The Challenges of Precision
Part of our goals with revisiting home data encounters was to create space for capturing data that is not ‘hard’ but instead that can represent the affective and lived experiences within a home. With Soft Fading, we particularly aimed for slow impressions of sunlight. The making process of Soft Fading revealed an important dilemma in the fabrication of a tool for data collection. While our intention was to create an artifact that could generate soft, imprecise, and open to interpretation data (hence our choice of an analog approach), we were still faced with a need for a high level of precision to accomplish the basic functionality of our artifact. For instance, it was critical that the same fabric section to be exposed at the same time everyday. Our stepper motor has exactly 200 rotor teeth, meaning that each revolution would need to take 200 steps at precisely 1.8º rotation per step. In order to have the fabric line up after 24 hours (360º) we had to have the motor take one step every 7.2 minutes (432 seconds). While in theory this should work, in reality a small misalignment over a long period of time (weeks or months) would mean that any data interpretation would be skewed. We saw a certain irony in the need to fabricate a very precise instrument to collect something that was meant to be read and interpreted imprecisely.
Once we started to live with Soft Fading, this contrast became even more pronounced. In one of Audrey's early reflections she wrote, “I plugged in SF at my desk. I put a dot on one of the elastic bands, and noted the time, to see if it would really turn in 24 hours.” She continues “After a couple days, the cylinder was still turning on time, which was very impressive.” Curiously Audrey describes the precise turning of Soft Fading as impressive, already hinting at a doubt over the device's precision. Ruby also noted the rotating of Soft Fading, “Sometimes when I'm working I'll hear the little tsssk of the cylinder turning. […] Soft Fading is like its own little clock, telling time in 7 minute increments.” The precision of soft Soft Fading is once again emphasized, similar to a clock. Both Audrey and Ruby's attention to the precision of the device highlight the connection between the instrument of data collection and the value and accuracy of the data collected—which was necessary to attain an ‘effortless’ feeling of softness (imprecision) on the faded fabric.
Once Ruby and Audrey had taken their pieces of fabric out of Soft Fading (Fig
15), both had similar questions when looking at irregular stripe patterns: why are there so many stripes? why is this not a simple gradient at sunrise and sunset? Ruby wondered if this was caused by their plant casting shadows on Soft Fading. Audrey retraced the patterns of light coming in through her window, even noting light bouncing off the neighbor's house. They questioned the accuracy of Soft Fading. Had the cylinder stopped rotating for a few days? Is that why there was a very faded band here? Reading the final piece of fabric requires interpreting multiple layers of events: the house conditions, what Soft Fading's hardware was doing, what the participants were doing, how the UV rays were changing day to day. It becomes impossible to discern the various factors. The tension between the precise turns and the messy data collection was finally manifested.
After some investigation, we found that the revolving cylinder was slipping on the motor's axle causing a mismatch in fading. Before deploying Soft Fading into Chandler's home, we implemented a fix by designing a new hexagonal end piece for the cylinder that properly attached to the stepper motor and prevented slipping. The fabric generated from Chandler's home was closer to our original hypothesis: that sections exposed closer to noon would be most faded while areas exposed at night would remain bright yellow (Fig
16). This experience made it clear that the building of instruments for data collection are central to the data generated and impact how we might eventually interpret data. In this case, we saw how data manipulation can start even before any data are collected at all, just in the design (and glitches or mistakes) of an instrument.
5.5 The human in the machine: Layers of labor
The actions directed by the Data Bakery help us make experiential the layers of labor involved in collecting, living with, and translating data that are otherwise intentionally hidden in mainstream IoT device systems. Collecting data, according to convention, is ideally done with minimal effort and with no personal connection to the ways in which it is translated and presented back to a user [
36]. In contrast, the data collected with Data Bakery is not automatically rendered into digestible charts or visualizations, but is instead sent to Philbert who takes the time to semi-manually transform the data into a recipe: pluging the On/Off data into the algorithm spreadsheet, applying the decimal numbers to an image file, and then sending the file to be printed at the home through Data Bakery's networked printer.
While living with Data Bakery, Jena reflected on the comparison between automated and human translations:
“My expectation was that it would feel like an autonomous system. My data gets sent to Bakery and then I get a recipe and that's it. But it really became more of a pen pal system with acts of care between me and Philbert and that's where the magic happened. It wasn't because this magic algorithm was making judgments about my home dwelling, it was because I now have this very dispersed system between my home, a friend of mine, and the Bakery, which was not what I was expecting at all.”
While we originally wanted to keep an element of human labor in the project, we did not anticipate that ‘the human in the machine’ would become a feature of the system that would draw as much attention as it did for Jena. Even as we tried to ‘hide’ Philbert's labor during Jena's experience, our technical debugging sessions and troubles with WiFi connectivity made Philbert's labor visible.
When the first printout arrived, Jena reflected on additional layers of labor required in order to interpret the recipe. In reference to having to fill out the baseline guide she stated “It felt like I had to provide for the Bakery before it provided for me.” The acts of performativity we described in 5.3 are also moments of labor enacted by Jena. Not only did she need to interpret the recipe, but she had to actively go to the store and buy ingredients, physically make dough, roll it into balls, and bake the cookies.
With the Data Bakery, we were able to interrogate multiple levels of labor within an IoT system, from Philbert's processing tasks, to our team's debugging process and Jena's care for the smart plugs, the book, and the prints, and her baking activities. We can imagine in a future iteration of this project also inquiring into labor directly at the services we used (IFTTT and Google), as a way to continue mapping hidden labors in IoT.
5.6 Expansive and imagined: In contrast with the material realities of WiFi
The Odd Interpreters allowed us to enter the worlds of data and to imagine what might be happening behind the scenes. With Broadcast, Chandler and Audrey responded to the sonic representations of data by describing the sounds in terms of quality (e.g., “fast”, “energetic”, “airy”) and location (e.g., “another world”, “on a beach”, “in the water”, “in a large mechanical workshop or warehouse”). Using these interpretations, Audrey imagined her data traveling to and through unobvious locations: “what underwater cable it's taking”. Chandler noted that Broadcast emphasized that “data is on journeys not just [at] destinations… it travels far”. While we had designed two types of sounds (location and movement), the distinction between the two types was not as obvious as we had expected. Nevertheless, the images and imaginaries they provoked touched on both location and movement (as seen above).
While living with Broadcast offered an expansive view on where data might be in the world, echoing Yanni Loukissas’ point that ‘all data are local’ [
75], developing and debugging Broadcast emphasized this idea even more. As Broadcast lived in multiple homes, we realized it had different behaviors in different homes—in fact, each home had a different amounts of WiFi traffic as part of their techno-landscape.
Data's local quality became very evident in our development of Broadcast, in particular when monitoring the real movement of smart speaker data. In addition to showing how data are constantly moving in and out of smart speakers, our work with Broadcast also emphasized that home networks are deeply situational and contingent. Fabricating Broadcast required collaboration between developers and designers. As a practical consequence, Broadcast was connected and tested within homes, dorms, or apartments of designers and developers. While we imagined Broadcast would behave consistently in home networks, each WiFi network provided unique challenges during debugging, leaving a unique imprint on Broadcast's software. In homes where the smart speaker was physically distant from a network access point, we observed the smart speaker shifting between 2.4ghz and 5ghz WiFi. This required us to expand monitoring to both network bands. We also noticed performance drops in home networks with low signal-to-noise ratios (or a large number of adjacent networked devices). Exposing Broadcast to different amounts of network traffic helped expose performance bottlenecks in our software architecture. To account for differences in performance, we adjusted Broadcast to be over-responsive in homes with high noise and conversely, less responsive in homes with low noise. The experiences of designing, implementing, and living with Broadcast highlights that, while networks are built on shared protocols and technologies, each home network has unique contours and specifics that Broadcast had to accommodate.
7 Conclusion
In this paper, we presented the Odd Interpreters: three artifacts that materialize alternative ways of engaging with Internet of Things (IoT) data in home environment. As data materializations, these artifacts not only allow occupants to further understand parts of their home environments, but also jolt their sense of “knowability” towards data beyond existing conventions. By weaving imagination with precision, and illusion amongst truth, new pluralities of encounters and relationships begin to emerge. Together, the OIs render experiential some timely critiques and concerns of IoT: infrastructure invisibility, assumed objectivity, and hidden human labor. Through our RtD practice, we engaged directly in working closely with data and building new ontological bridges between home, IoT data and human. We hope that this work will further discussions within the design research and HCI communities around how we, as designers and researchers, know data (as plural, layered, diversified, part of infrastructures, and enmeshed with human labor) and how we embed data in present and future interactions.