28 INTER ACTIONS S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS. ACM.

ORG
 C O V E R S T O RY

Does HCI Scale?

Scale Hacking
and the
Relevance of HCI
Barry Brown, Stockholm University
Susanne Bødker, Aarhus University
Kristina Höök, Royal Institute of Technology (KTH)

Insights HCI has had a massive impact on of users, across diverse contexts. How
→→ The use of design the world through streamlining can we understand and design for this
fictions, mid-range and enabling millions of interfaces complex of interconnected uses? Put
theories, software and on billions of devices. As we face simply, does HCI scale?
hardware populations, the potential of a tenfold increase We discuss three different scales: the
and data-driven design in the number of devices and their number of users, the different contexts
tools and research complexity, it is worth asking about of use, and the multitude of systems and
methods is a start on the relationship between HCI and technologies. Emergent phenomena
addressing scale. scale. Do the tools and research like filter bubbles or “fake news” can
→→ Is addressing scale methods we currently deploy scale to exist only when millions of people
possible without access to the millions of future interfaces and are sharing and interacting online on
millions of users or millions systems, used by billions of people, a daily basis. While approaches like
of different systems?
IMAGE BY NAL APHOTOS

across multiple contexts? In this article statistical generalization offer the

we outline how we see the challenge possibility to predict the reliability of
of scale. By scale we mean how certain phenomena, they can do that
technology is used in large networks of only with phenomena that already
interconnected systems, with billions exhibit themselves—predicting and

INTER ACTIONS. ACM.ORG S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS 29

cover story
when we see hybrid forms (such as, say,
Snapchat’s combination of photography
and messaging), HCI is usually rather
late to understand them and thus needs
to develop its understanding afresh.
What then are the tools that can help
us understand systems across different
settings? Universal technologies may be
one answer, yet other research points to
specific use activities where the context-
crossing has often been ignored.
A related issue comes when we need
to reason about usage where the number
of users expands beyond what can be
considered for a practical trial or study.
In experimental and explorative HCI,
there is a tendency to focus on a limited
number of users, but this offers little
help when scaling to mass usage. While
statistical generalizations can give
confidence in phenomena that manifest
themselves at this small scale, they
cannot identify phenomena that would
occur only at the bigger scale.
In turn, there is a tendency in
interaction design to focus on one
technology at a time. We design the piece
but apply it in a whole, into contexts
where there are already other artifacts
understanding new phenomena is much WHAT IS SCALE? and systems. For research this means
harder. In turn, while we can examine There are many ways in which HCI has that we need to address the wholes of
and generalize from current-day, moved beyond one user/one device, infrastructures and artifact ecologies
large-scale system use (such as studying and others where it has not yet done and not simply the singular systems and
Wikipedia or Facebook), this does so. One part of this puzzle is being designs. Indeed, the innovative systems
not tell us how these systems change able to reason across different settings. we study today are perhaps like jigsaw-
over time. In contrast, qualitative More and more, technologies are puzzle pieces, the significance of which
analysis can detect future trends or being applied across contexts, such as will not be apparent until other parts of
help us to understand and design new when work technologies are brought the puzzle are adopted. How then can we
systems, but this too can be hampered into the home, or when people use understand a single piece, without access
by the difficulty in predicting what is technologies in different places or in to the whole puzzle? The application in
fundamental (and will remain) and what different forms of collaboration with the whole may not simply be a smooth
is passing or unimportant. others. Mobile devices are perhaps transition of the artifact ecology. As a
Focusing on HCI, how do prevailing the most obvious examples of this, matter of fact, a new design may disrupt
popular HCI research methods, in but social media, ubiquitous text the ecology in ways that cause tensions
particular research through design messaging, eHealth, and even casual and more profound changes, perhaps on
(RtD), critical design, interface- gaming have insinuated software into a longer time scale (see, e.g., [1]).
technique-driven development, and its current almost ever-present form. Indeed, ecologies of artifacts,
data-driven design methods, address While of course there are studies of services, and data may lead to
work at scale? We suggest here four each of these, the challenge of scale here unexpected and unpredictable
different “hacks” that can help with the is that we still lack an understanding interactions. Layers of systems on top
challenges to these methods. across these different contexts. Indeed, of one another may also relate to scale:
Through a touch of the iPhone, we may
change the heating or lighting of a whole
house, steer a ship, or move an atom in

Layers of systems on top of one another a quantum computer. In this manner,

that touch—the human-computer
may also relate to scale: Through a touch interaction—cuts across human scale as

of the iPhone, we may change the heating

I M A G E B Y FA B R I K A S I M F

well as super- or supra-scales [2].

or lighting of a whole house, steer a ship, SCALE IN HCI

or move an atom in a quantum computer.

Given these challenges to scale—in size
of data and in ecologies of artifacts—

30 INTER ACTIONS S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS. ACM.ORG

we need to ask how well HCI research role can longitudinal studies play in between particular HCI work and the
methods address them. Do our methods letting us understand scale? world in which it must ultimately find
explicitly handle big data? Sometimes Going beyond individual designs. a home? One promising, but clearly
this can be addressed by moving back Research through design (RtD) and partial, technique is the use of design
and forth between quantitative and various forms of critical design are fiction. Design fiction is an approach in
qualitative data in rapid succession, popular methods in contemporary which fictional accounts of the use or
designing and redesigning services HCI. We have seen a lively debate on presentation of particular technologies
that rely on big data (such as Spotify ways of framing design knowledge that are used to envision a new future.
or Facebook). For other problems, we can transcend different technological In our own work, for example, we
might address scale implicitly if the waves or design contexts (sometimes created a “future IKEA catalog” in
method guarantees that questions framed as mid-range theories, such as which unusual and inventive different
are asked that allow it to scale beyond strong concepts or experiential qualities types of IKEA devices were advertised
the specific, limited setting that is [3]). We have also seen a lively debate and listed in an imaginary IKEA
empirically examined. on how to frame particular critical catalog from the future [6]. Our goal
In turn, it is not enough that research designs against the backdrop of a longer in doing this was to think a little more
methods can handle many users or go history— the so-called critical thread about how IKEA furniture could
across many devices. The outcome of of research. By putting particular be hybridized with innovative new
the research process—the knowledge critical designs into a context of, for technologies, but also to think about
contribution—also has to scale and example, art styles or whole traditions how the world, business models, and
be applicable to many different design of work in the humanities, the effect of methods that IKEA works with would
situations, contexts, domains, and so on. a particular critical design project, and change and influence that technology
We characterize these attempts as scale its contribution to a high-level, ongoing in development. To take one example,
hacking—ways of conducting research debate, might scale [4]. IKEA as a company has always reduced
that can go beyond the particular Sometimes the ideas arising in a costs by moving work onto the user,
situations being addressed or studied design-driven project are organized such as transporting furniture from the
and be generalized in new ways. in whole methodologies, providing store to home, assembling furniture,
Scale has of course always been a particular design judgments or and in some cases even fitting existing
concern to HCI researchers. There has, value-based approaches to design parts of furniture together to create a
for example, always been work that work. An example is the well-known user’s own arrangement. If there are
relies on statistics and quantitative Utopia project, originally aimed at 3D printers or devices that in some way
data to statistically generalize over shaping desktop publishing with can produce raw materials, we might
populations. In the early days of HCI, laser printing in a direction that think about how IKEA might simply
through its origins in ergonomics and would keep the sensitive, skill-based offer users plans that they themselves
cognitivism, HCI researchers attempted aesthetics that typesetters had, but can print, rather than relying upon the
to find behavior patterns that would that instead brought us the whole IKEA production process. Here we take
be stable over different contexts, idea of participatory design with its a new technology and consider how it
independent of specific technological strong political values [5]. What is will develop in IKEA’s world, with the
solutions. Fitts’s law is one example; perhaps interesting here is that we kinds of constraints on its production
creating software and/or hardware seldom discuss how ideas are generated and development with which the
that is made available as open source in HCI, and how these ideas might company is familiar.
tools is another. With software, this is have purchase. We find discussions Of course, one rather serious
relatively easy, as one line of code can that address “motor themes” or problem with design fiction is that, as
be duplicated and run on thousands or concepts like interaction somehow a fiction, it is likely, perhaps fatally, to
even millions of devices, and in turn be unsatisfactory, in that they are attempts replicate the prejudices and issues of
run millions of times. to rather bluntly explore the sorts of its authors, and it perhaps lacks the
topics we address, rather than thinking generality that comes from data that
FOUR SCALE HACKS about ideas more centrally. might contradict those wishes in some
We use the term scale hacks here, as When we discussed our definition way. Moreover, like fiction, much of
our suggestions are not complete of scale above, we realized that HCI it is at times poor. But we see at least
methods in their own right but rather suffers from relatively stunted ideas of the possibility of being able to scale in
are opportunistic modifications of how culture at scale can influence the new ways with design fiction, to think
existing methods to deal with scale. As development of technology, and how about longer-term processes that take
one would expect with hacks, these are technology at times develops in quite design innovations but twist them into
not complete solutions. Instead, they unexpected directions. For something new shapes. One way to challenge the
are more a way of beginning to deal like touch technology, for example, design-fiction method could be to take a
with and engage with the problems that the extent to which it was of interest scale perspective. That is, using fictions
scale presents. First, can we think about to HCI was as an input technology for to ask questions like: Would this
the role that culture plays in design? large displays. In reality, clearly its true scale to millions of users and still be
Second, can we think across interaction home was in enabling input to smaller sustainable and ethical? If there were to
techniques to broader interaction displays. How could we have foreseen be many interfaces and interactions like
gestalts? Third, how can we change this misdirection? How can we think in these, what social or societal changes
design to make use of data? Last, what innovative ways about the connections could follow?
INTER ACTIONS. ACM.ORG S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS 31
cover story
app stores as such mediate the choice
and sharing of apps [8].
If we think of the longer-term
deployment of applications, we might
also reflect on their life cycles—in
particular the ways in which they
may be updated to new versions, or
how applications might die or be
reborn in new forms. To add to this
complexity, there are also questions
of how applications might depend on
each other or create ecological niches
for other applications. The markets
in filters for Photoshop, for example,
or new software keyboards for mobile
phones, exist as niches in the world that
is created by other software products.
Data as a design material. When
discussing how to make HCI research
methods scale, data-driven research
methods are perhaps the first to come
to mind. Data is often discussed in
terms of its potentially enormous
value in modeling and predicting
behavior. This in turn can be used to
create various actuations in the world:
controlling processes in industry,
organizing logistics, handling so-
called smart cities and smart homes,
Ecologies of software and replicates and spreads, sometimes and augmenting our bodies. It seems
artifacts. User interface software in autonomous ways, or at least clear to both academia and industry
and technology (UIST) is concerned autonomous to the software developers alike that patterns in data will be
with a broad range of technical issues [7]. Large-scale system design has to the basis of a whole thriving field
in interfaces, such as tools for user some extent addressed these issues, of innovations, novel services, and
interface development, advanced but with much less attention to the novel interactions. But in order to
interaction technologies thriving case where the software is produced by serve that role, it is not enough to just
off of new materials, and software different authors, and where systems engage with techniques of statistical
architectures of interactive systems. have complex interactions between generalization or inference. Rather,
While UIST research engages with the the different parts. Thinking about we are thinking about how data can
problems of understanding the user’s software (and users) in this way raises support the actual intellectual process
position when systems scale, scale a number of interesting questions. Not of idea generation. While this might
presents challenges there with the much HCI work has focused on the seem like a suspect idea, in the world
complexity of systems when they scale, interaction between apps, even in the of design there is increasing use of
and the role of humans in this. hands of one user. In the contemporary machine-learning techniques to
One interesting approach is work situation where there are multiple apps generate possible design solutions.
on software populations, and studying with similar functions available in app Simply gathering data without
software ecologically as we might study stores, we encourage studies of how knowing what actuation that data can
animals in an environment. Studying such apps interact, how one gets chosen drive is, if not meaningless, at least
software as a population focuses over another by users, and how users misguided. The design process that
on how different software interacts arrange their uses of several apps over makes use of data as a material is not
in specific environments, and even time and activity, not to mention how one of finding some unknown pattern
in a preexisting dataset that surprises
the design team, allowing them to
innovate new functionality. The

It is not enough to just engage with choice of data sources is not objective,
but rather constitutes a choice in
techniques of statistical generalization the design process that in turn will
IMAGE BY AFRICA STUDIO

or inference. Rather, we are thinking

determine what services or technical
actuation can be done. It is a chicken-
about how data can support the actual and-egg problem.

intellectual process of idea generation.

Once there is a first application
where data is used in interesting ways,
32 INTER ACTIONS S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS. ACM.ORG
such as in the early experiments with is Kraut et al.’s Homenet studies run in interaction design research. ACM
movielens.org or Minecraft, then in Pittsburgh, which pioneered the Transactions on Computer-Human
new functionality and new possible deployment of Internet technology Interaction 19, 3 (2012), 23.
4. Bardzell, J. and Bardzell, S. Humanistic
interactions can be imagined. For [10]. Two innovative aspects of this
HCI. Synthesis Lectures on Human-Centered
example, the crowdsourced city work are that the original deployments Informatics 8, 4 (2015), 1–185.
planning in Minecraft could not have were of sufficient scale for complete 5. Kyng, M. Bridging the gap between politics
been imagined before Minecraft communities to be established, hence and techniques: On the next practices of
existed as a multi-player game. In generating new types of interactions participatory design. Scandinavian Journal
our view, these exemplify interesting between participants. The Homenet of Information Systems 22, 1 (2010), 49–68.
6. Brown, B., Bleecker, J., D’Adamo,
processes of crowdsourcing design trial ran for 20-plus years—to the
M., Ferreira, P., Formo, J., Glöss, M.,
in and through massive amounts of point where the Homenet systems were
Holm, M., Höök, K., Johnson, E.C.B.,
data. Partially, though, this gives us superseded by market-based Internet Kaburuan, E., and Karlsson, A. The IKEA
some arguments for attempts (such technology (at least partially). This Catalogue: Design fiction in academic and
as app-store trials) to expand the user longitudinal component gives us one of industrial collaborations. Proc. of the 19th
base involved in particular trials, the few cases where we can contrast how International Conference on Supporting
thus overcoming the bootstrapping market innovations can replicate but Group Work. ACM, 2016, 335–344.
7. Higgs, M., Morrison, A., Girolami, M., and
problems. It may be that certain also fail when compared with academic-
Chalmers, M. Analysing user behaviour
phenomena appear only when systems produced innovations. through dynamic population models. CHI
have a certain complexity or a certain 2013 Extended Abstracts. ACM, 271–276.
size, and the right combination will CONCLUSION 8. McMillan, D., Morrison, A., Brown, O.,
unlock interesting results. HCI Our original question was: What Hall, M., and Chalmers, M. Further into
researchers have occasionally made are the tools/research methods we the wild: Running worldwide trials of
sure their designs can find a significant can use to reason about millions of mobile systems. Proc. of the International
Conference on Pervasive Computing.
audience through the development of interfaces, thousands of systems,
Springer Berlin Heidelberg, 2010,
proper mobile apps or Internet-based used by billions of people? Clearly our 210–227.
applications [8]. Once you have a stable suggestions here do not fully address 9. Gaver, W., Boucher, A., Jarvis, N.,
group of users, you can test theories this question. Adding design fictions, Cameron, D., Hauenstein, M., Pennington,
by altering the design for one group of mid-range theories, software and S., Bowers, J., Pike, J., Beitra, R., and
users and then statistically comparing hardware populations, or data-driven Ovalle, L. The Datacatcher: Batch
deployment and documentation of 130
the behavior changes. design tools and research methods is
location-aware, mobile devices that
Scaling over longer time periods. only a flawed start. Perhaps if we had
put sociopolitically-relevant big data in
Last, we might think about how we can access to millions of users and millions people’s hands: Polyphonic interpretation
simply continue or expand our existing of interfaces under our control, we at scale. Proc. of the 2016 CHI Conference
methods but prolong them through could ask really interesting research on Human Factors in Computing Systems.
longer studies or more industrial-style questions and then use quite different ACM, 2016, 1597–1607.
processes. This might be characterized research methods to figure out the 10. Kraut, R., Scherlis, W., Mukhopadhyay, T.,
Manning, J., and Kiesler, S. HomeNet: A
as “trying harder”—simply taking what answers. Yet for most research studies,
field trial of residential Internet services.
we already do but attempting to prolong this is beyond what is possible or Proc. of the SIGCHI Conference on Human
or expand its scale. One example of practical. This means that we need our Factors in Computing Systems. ACM, 1996,
this is the exploration of batch design existing methods, and new methods, 284–291.
methods, where they produce a hundred to scale up. We should not be content
versions of one artifact rather than just a simply to figure out limited domains Barry Brown’s recent research has
few [9]. By expanding the length or scale or sites. Instead, we should ask more focused on the sociology and design of leisure
of a particular technology trial, there is generic questions about how best to technologies—computer systems for leisure
and pleasure. In recent years he has used video
the possibility that we might be able to integrate new things into the ecologies
methods to study mobile phone and Apple
gain a more diverse perspective on use. of artifacts we already live with. What Watch use; he is currently working on an “at
One challenging question here, is needed are new research methods scale” study using similar methods.
though, is whether this does actually that “talk back,” provide alternative →→ barry@dsv.su.se
address the analytic problem of scale. futures, and push back on our thinking,
Susanne Bødker is well known for her early
What we are looking for is not just instead of scaling by just adding access
work on participatory design and then later on
more of the same but instead to actually to data or ecologies of artifacts. for activity theory, as well as for framing the
get a handle on the sorts of issues that third wave of HCI.
are generated when there are larger Endnotes →→ bodker@cs.au.dk
interactions (be they between users 1. Bødker, S. and Klokmose, C.N. Dynamics
or different systems). The challenge in artifact ecologies. Proc. of NordiCHI’12. Kristina Höök is active in interaction design
448-457. and is currently focusing on somaesthetic
then is to pass on doing existing trails
2. Bødker, S. and Andersen, P.B. Complex design and issues around Internet of Things
with more people in favor of systems mediation. Human Computer Interaction development. Her earlier work focused on
where the larger number of users has 20, 2 (2005), 353–402. social navigation, seamfulness, and affective
a material effect on what each user 3. Höök, K. and Löwgren, J. Strong interaction.
does. One interesting example of this concepts: Intermediate-level knowledge →→ khook@kth.se

DOI: 10.1145/3125387  © 2017 ACM 1072-5520/17/09 $15.00

INTER ACTIONS. ACM.ORG S E P T E M B E R – O C T O B E R 2 0 17 INTER ACTIONS 33