Kim Sauvé, Steven Houben
We are in the midst of a data revolution. Large amounts of data can be collected about ourselves and our environment. Modern technologies such as wearables, sensors, and crowdsourced tools are revolutionizing the way datasets can be accessed and used by larger audiences than ever. This availability of data has the potential to empower individuals, groups, and communities and lead to personal, communal, environmental, and even political change as these technologies become tools for reflection, discussion, and decision making.
The way in which such widespread data availability can be leveraged or embedded into everyday life, however, is an open and difficult challenge. In recent years, perspectives such as human-data interaction (HDI) , data commons, and data humanism  proposed to shift the ownership, actionability, and interaction with data toward people themselves. But democratizing the access and presentation of data in a meaningful context to nonexperts remains a problem, as there is a systemic lack of tools, visualization approaches, and conceptual and interaction models targeted at nonexpert groups.
We propose that meaningfully embedding data into everyday experiences through artifacts is a way of bridging this gap between people and datasets. Literature and our own experiences with constructing physical data artifacts show that for data to become meaningful to people, there needs to be a strong and direct connection to their own experiences, activities, or situations. Bringing data close to those personal experiences leads to interactions or understandings that leverage the data to create more informed and actionable outcomes (i.e., enjoyment, reflection, or behavior change). As this translation from data to designing everyday experiences is a fast-emerging research topic, we reflect on challenges and opportunities learned through our own experience of creating and deploying physical data artifacts in the wild [3,4,5,6].
What are physical data artifacts? Physical data artifacts are interactive physical artifacts that: (i) represent data through physical and material properties, and (ii) allow for meaningful interaction, configuration, or interrogation of that data. In this context, data includes a wide variety of data types and granularity. For example, from direct "raw" data (e.g., a sensor-reading), to aggregated datasets (e.g., a snapshot of air pollution in London), to experience sampling (e.g., bullet journaling), and derived data (e.g., quality of sleep through a motion sensor). Broadly speaking, physical data artifacts constitute any designed physical object that presents one or more datasets to a person, group, or community within a specific context. In our views, this embeddedness or situatedness is fundamental to co-constructing meaning or actionable outcomes from data.
Physical data artifacts have a rich history, adopting many shapes and forms and evolving over centuries . While other work provides a complete overview , here we highlight some forms of physical data artifacts as observed within our own work. The first type is ambient data visualizations, which are abstract event-based communicators of changes in data that are important but not essential to the viewer (e.g., LOOP  and Physikit ; Figure 1A). The second type, data sculptures, are artistic representations of socially relevant issues (e.g., Econundrum ; Figure 1B). Finally, public data installations are often larger interactive systems that are designed as kiosks for data inspection and exploration on a larger scale (e.g., Roam-IO ; Figure 1C). These broad categories are not mutually exclusive or exhaustive, but rather are demonstrative of a spectrum of different form factors and different ways of interacting with data.
|Figure 1. Different forms of physical data artifacts.|
In this section, we outline key elements to consider when translating data to everyday design, such as the selection and translation of the dataset, the construction of the physical artifact, and the use and implication of the physical data artifact in the real world.
How to select data. The goal of physical data artifacts is to translate complex problems or situations into a representation that leads to an awareness, understanding, or reflection, which in turn enables meaningful outcomes or change. For example: How can we understand the climate impact of dietary choices? or What is the impact of pollution in cities on everyday life?
The first consideration when designing a physical data artifact for a particular context is what is the actual information people need and in what form to tackle the topic at hand? We often see data communicated through classic UIs with numbers and graphs. However, people are generally more interested in relative changes, not in absolute values. Second, we have to consider how to communicate the information. There is a difference between communicating "if something is happening" versus visualizing "what is happening." We want to utilize physical space and the design to communicate data changes and relations, not only states. Moreover, the target audience can have different goals and information needs that change over time. Important data elements that need to be defined are granularity (what level of detail does the audience need?), actionability (how can they use it in daily life?), and temporality (in what frequency over what time span?).
In our work, we observed different strategies when it comes to dataset selection. One can take either a user-centric or a domain-centric approach, by directly engaging the target audience (e.g., interviews with users of activity trackers for LOOP) or obtaining knowledge from related fields (e.g., looking at eco-visualization literature for Econundrum). These activities can happen in a more or less participatory way, and with different stakeholders in relation to the topic (i.e., target audience or experts).
How to translate data to design. The next challenge is to translate the selected data into a meaningful design. This translation process (or representation mapping) is not straightforward, and 2D visualization methods are only partially informative for communicating information in 3D space. As there are currently no established communication tools, we need to develop a specific visualization vocabulary—a visual language that can be understood and interpreted by the audience. From our experience, we have extracted three important things to consider when constructing this language:
- Aesthetics: If something is not aesthetically pleasing (e.g., LOOP, Physikit) people don't want it in their home, or in the case of (semi-)public environments, they aren't inclined to look at it when it is not aesthetically interesting or doesn't stand out (e.g., Econundrum). Therefore, a physical data artifact should balance informative and aesthetic properties so it's not only meaningful for the audience but also pleasing to look at.
- Abstraction: It is important to keep in mind the difference between a metaphor and an abstraction, as they can be more or less effective in communicating information. An abstraction is general enough to make the physical data artifact blend into the physical environment (in people's periphery) so it can be ignored when not needed, but also allows for its own interpretations and metaphors when looked at. Using abstraction creates information that is easier to comprehend by people. For example, LOOP visualizes the increase of steps (progress) by abstract upward movement, whereas Econundrum metaphorically visualizes food types through graphical icons.
- Comparison: "Raw" values or data points (e.g, from an air pollution sensor) are often difficult to interpret. If we can compare data over time (historical data) or across different categories (any type of reference data) it can become more informative. Creating meaningful comparisons, either absolute or relative, between elements of a dataset allows for the audience to make sense of the data. For example, knowing the number of steps you walked is not interesting in itself, but knowing if you did better than yesterday or better than your friend allows for meaningful comparison.
How to develop and construct physical data artifacts. Another challenge is to develop the concept and build the artifact. There is no fixed infrastructure when it comes to the tools and methods for designing and constructing physical data artifacts, and it is often an intersection of many digital and physical elements, such as online data systems, sensors, and actuation mechanisms, that together form the physical representation of data.
Where do you start? It can be helpful to get inspiration from related fields as the design of physical data artifacts is at the intersection of different disciplines. We can build on knowledge from related fields (e.g., shape-changing interfaces and InfoVis) and they can support the translation process of how to visualize the information most effectively, from a mechanical, constructive, and conceptual perspective.
Who do you involve? It can be challenging to identify stakeholders and their interests; moreover, different stakeholders can have conflicting interests, which will most likely influence the creation of your physical data artifact. It is important to identify the stakeholders and users relevant to your topic and involve them in the translation process early on.
Data exploration through physical data artifacts is a collective sensemaking activity, and the basic design of these artifacts should directly support this.
What design activities can you do? Explore a variety of creative methods, either with or without users/stakeholders, as these can help you draw inspiration and insights in unique ways. Example activities are, but not limited to: analog/digital sketching (Figure 2A), rapid prototyping (Figure 2B), creative sessions with users (Figure 2C), and pilot studies (Figure 2D).
|Figure 2. Example design and translation activities.|
How do you realize it? After developing the concept for the physical data artifact, a varied skill set is needed to create the final system (Figure 3), which is an exploratory process in itself. Think of activities such as developing the UX design of the interface, mechanics for actuation, electronics for artifact behavior, but also developing the online part of the system: connecting the physical changes to digital data and bridging the physical artifact and the digital data stream.
|Figure 3. Example realization and implementation activities.|
How are physical data artifacts used? The role of physical data artifacts is to explore how people can leverage different situated representations of data in everyday situations. From our experiences with deploying physical data artifacts in field studies, we observed three main usage patterns that are central to designing new physical data artifacts:
- Exploratory awareness: As people often start with little to no insights about the data, there is an initial phase of highly explorative behavior and interactions with the physical data artifact in which people explore the changes, meaning, and understanding of the data through various mappings and actions. The more physical data artifacts enable such exploratory interactions (e.g., through reconfigurations or mapping tools), the faster people will understand the scope of the data, leading to more sustained and meaningful interactions. Physical data artifacts are a portal into an intangible data space and often become a central hub or meeting place for the discussion of the topics related to the data. While an initial understanding or awareness about the data can be valuable, concrete actionable outcomes, activities, or steps will help people internalize the importance of the data and its consequences.
- Appropriation as utility: One of the central observations across our different field studies is that once people build a basic understanding of the scope of the data, the presentation of that data becomes a platform for more-refined exploratory appropriations to bind the meaning of the data to situated activities, events, or contexts. To facilitate these explorations, physical data artifacts could explicitly support ways of appropriation through a range of basic flexible design patterns. For example, being able to move the physical data artifacts, turn them on/off, change the basic data mapping, or use them by multiple people or by remote control.
- Socialframe of reference: In all our studies, we observed how people compared their own interpretation of the data to other mappings or data representations to create a direct comparison that allows them to understand the data in relation to an external point of reference. For example, rather than trying to understand the level of air pollution in their street, people would compare the change to a location that they knew was heavily polluted. Furthermore, we consistently observed that physical data artifacts are used by groups of people (e.g., colleagues, families, passersby), not individuals alone. Data exploration through physical data artifacts is a collective sensemaking activity, and the basic design of these artifacts should directly support this.
What are the implications and consequences of data in the real world? Making data visible in everyday life brings it closer to the contexts, activities, and situations it affects. This can lead to direct and important consequences at a personal (e.g., behavior change), social (e.g., community impact), environmental (e.g., climate change), or even political (e.g., policy changes) level. However, because physical data artifacts are embedded into everyday life and experiences, there is also a danger for negative consequences:
- Lack of actionability: While increased awareness of situations through data can be empowering, the lack of actionable outcomes or steps can lead to a situation of helplessness or confrontation. For example, continuously exposing people to Covid dashboards can be anxiety-inducing. Or what happens if people understand that the air pollution in their street is consistently at a high level, but they have no means to change it?
- Privacy and sharing: We observed in our studies that physical data artifacts were almost consistently used or appropriated by groups of people, which poses interesting challenges about how data—which might be personal or identifiable—should be treated. For example, people might differ in what they feel comfortable sharing about themselves or have different reactions to the shared data of others. Particularly for (semi-)public physical data artifacts, there are open challenges about data ownership, accountability, and representations.
- Temporality: Physical data artifacts are often introduced in everyday life through temporarily in-the-wild evaluations (one week up to several months). Hence, we still don't fully understand how they would coexist with people over a longer term, when they have passed the novelty effect. One can imagine there is a saturation point of the usefulness of making information accessible to the public. Are physical data artifacts here to stay for the long term in a particular context? Or will they have a saturation point of engagement and are meant to travel around different contexts to inform multiple communities? Do we expect people to engage with them more frequently over the short term to plant a seed (multiple times a day for a week), or do they benefit from long-term exposure to possibly elicit behavioral change (daily for a couple of months)?
- Physicality: Due to the tangible nature of physical data artifacts, the availability of data gets bound to, and associated with, a physical location, which is something that has to be carefully considered when designing. Moreover, one has to anticipate the sustainability and maintenance of the physical system over time. Lastly, how is the "ownership" or responsibility for the system arranged over time?
The central goal of physical data artifacts is to provide people with tools and designs that are integrated into everyday life, which support people in building a shared understanding of data and its consequences. We increasingly see physical data artifacts emerge for topics on sustainability and climate change, personal health and vitality, city and environmental data, and even artificial intelligence. While all these domains have specific challenges for how to translate data into actionable explorations and outcomes, we hope this summary of our findings from our own experiences of designing physical data artifacts can be inspiring for future designs and conversations around the role of data in everyday life. As we are still in the early phases of physical data artifacts, we see future research challenges at a conceptual level (how to talk about and discuss the concepts of physical data artifacts), a technical level (how to construct new physical data artifacts that can be used in the wild), and an empirical level (how to further our understanding of the impact of data awareness for people's lives).
1. Mortier, R., Haddadi, H., Henderson, T., McAuley, D., and Crowcroft, J. Human-data interaction: The human face of the data-driven society. SSRN, 2014; http://dx.doi.org/10.2139/ssrn.2508051
2. Lupi, G. Data humanism: The revolutionary future of data visualization. Print Magazine 30, 3 (2017); https://www.printmag.com/post/data-humanism-future-of-data-visualization
3. Houben, S. et al. Roam-IO: Engaging with people tracking data through an interactive physical data installation. Proc. of the 2019 on Designing Interactive Systems Conference. ACM, New York, 2019, 1157–1169; https://doi.org/10.1145/3322276.3322303
4. Houben, S. et al. Physikit: Data engagement through physical ambient visualizations in the home. Proc. of the 2016 CHI Conference on Human Factors in Computing Systems. ACM, New York, 2016, 1608–1619; https://doi.org/10.1145/2858036.2858059
5. Sauvé, K., Bakker, S., and Houben, S. Econundrum: Visualizing the climate impact of dietary choice through a shared data sculpture. Proc. of the 2020 ACM Designing Interactive Systems Conference. ACM, New York, 2020, 1287–1300; https://doi.org/10.1145/3357236.3395509
6. Sauvé, K., Bakker, S., Marquardt, N., and Houben, S. LOOP: Exploring physicalization of activity tracking data. Proc. of the 11th Nordic Conference on Human-Computer Interaction: Shaping Experiences, Shaping Society. ACM, New York, 2020, Article 52, 1–12; https://doi.org/10.1145/3419249.3420109
7. Dragicevic, P. and Jansen, Y. List of physical visualizations. 2020; http://www.dataphys.org/list
Kim Sauvé is a Ph.D. candidate in the Interactive Systems group at Lancaster University. Her research focuses on exploring the underlying principles of physicalization design. She specializes in research through design, applying design practice and developing interactive research prototypes to generate new insights for human-data interaction. firstname.lastname@example.org
Steven Houben is an assistant professor in human-computer interaction at Eindhoven University of Technology. His research focuses on physical and ubiquitous computing systems. His work explores physicalizing human-data interaction to support "from sensor to physicalization" and study new cocreation processes, concepts, interaction paradigms, and data embodiments for nonexpert human-data/AI interaction. email@example.com
Copyright held by authors. Publication rights licensed to ACM.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2022 ACM, Inc.