The Three Paradigms of HCI

Steve Harrison Phoebe Sengers ABSTRACT

Department of Computer Science Information Science and Technol- Informal histories of HCI commonly document two
and (by courtesy) Art and Art ogy major intellectual waves that have formed the field: the
History Cornell University first orienting from engineering/human factors with its
Virginia Tech Ithaca, NY 14580 USA focus on optimizing man-machine fit, and the second
121 VTKW II [email protected] stemming from cognitive science, with an increased
2202 Kraft Dr – MC 0106 emphasis on theory and on what is happening not only
Blacksburg, VA 24060 USA in the computer but, simultaneously, in the human
[email protected] mind. In this paper, we document underlying forces
that constitute a third wave in HCI and suggest sys-
Deborah Tatar temic consequences for the CHI community. We provi-
Department of Computer Science sionally name this the ‘phenomenological matrix’. In
and (by courtesy) Psychology the course of creating technologies such as ubiquitous
Virginia Tech computing, visualization, affective and educational
123 VTKW II technology, a variety of approaches are addressing is-
2202 Kraft Dr – MC 0106 sues that are bad fits to prior paradigms, ranging from
Blacksburg, VA 24060 USA embodiment to situated meaning to values and social
[email protected] issues. We demonstrate the underlying unity of these
approaches, and document how they suggest the cen-
trality of currently marginal criteria for design, evalua-
tion, appreciation, and developmental methodology in
CHI work.

Author Keywords
Embodied interaction, CSCW, interpretation, reflective
HCI, paradigms
Copyright is held by the author/owner(s).
CHI 2007, April 28 – May 3, 2007, San Jose, USA
ACM Classification Keywords
ACM 1-xxxxxxxxxxxxxxxxxx.
H5.m. Information interfaces and presentation (e.g.,
HCI): Miscellaneous.
 2

INTRODUCTION ell [5], oriented around the idea that human informa-
Over the last few years, the authors of this paper have tion processing is deeply analogous to computational
become increasingly aware that the third paradigm has signal processing, and that the primary computer-
been discussed in corners and cafes with much head human interaction task is enabling communication be-
nodding at the CHI conference, but has not been intro- tween the machine and the person. This cognitive-
duced as a legitimate frame or lens through which to revolution-influenced approach to humans and technol-
view contributions. This paper is an attempt to give ogy is what we usually think of when we refer to the
wider voice to the idea, first named by Malcolm HCI field, and particularly that represented at the CHI
McCullough’s book, Digital Ground, and discussed but conference. As we will argue below, this central idea
not named in Paul Dourish’s Where the Action Is [22, has deeply informed the ways our field conceives of
8] Our name for this is the “Phenomenological Matrix.” design and evaluation.

Looking back over the history of CHI publications, we The value of the space opened up by these two para-
can see how our community has broadened intellectu- digms is undeniable. Yet one consequence of the domi-
ally from its original roots in engineering research and, nance of these two paradigms is the difficulty of ad-
later, cognitive science. The official title of our central dressing the phenomena that these paradigms mark as
conference is “Conference on Human Factors in Com- marginal. Over the last twenty years a wide variety of
puting Systems” even though we usually call it “CHI”. critiques and approaches have been emerging that ap-
Human factors1 for interaction originated in the desire pear to fit poorly the models and methods emerging
to evaluate whether pilots could make error-free use of from the cognitive revolution. These include participa-
the increasingly complex control systems of their tory design, value-sensitive design, user experience
planes under normal conditions and under conditions of design, ethnomethodology, embodied interaction, in-
stress. It was, in origin, a-theoretic and entirely prag- teraction analysis, and critical design. On the surface,
matic. The conference and field still reflects these roots these critiques appear to involve a disparate array of
in the occasional use of simple performance metrics. issues and approaches; yet we will argue that many of
these approaches can be usefully seen as elements of a
However, as Grudin documents [17], CHI is more third (“3rd”) paradigm, which treats interaction not as a
dominated by a second wave brought by the cognitive form of information processing but as a form of mean-
revolution. HCI adopted its own amalgam of cognitive ing making in which the artifact and its context at all
science ideas centrally captured in Card, Moran & New- levels are mutually defining and subject to multiple in-
terpretations. Meaning making is entailed by the ana-
lytic frame employed by the designers and analysts,
1
Coming originally from “scientific management” (i.e. Tay- and also by the users and other stakeholders in the
lorism) in the early 20th Century, human factors began as
situation of use.
an attempt to increase production and reduce injury. By
the time of computers, it had moved on to concerns with
“critical incidents”.
 3

These critiques and approaches not only focus on dif- scribes not an accretive model of scientific knowledge,
ferent topics and questions, they also suggest alterna- but one of successive and overlapping waves in which
tive metrics and methods for design and evaluation that ideas are fundamentally re-framed. Canonical examples
can be difficult to reconcile with ones emerging from of such paradigm shifts include the acceptance of conti-
the first two paradigms. Their clash with some of the nental drift by earth scientists and the shift from a me-
central assumptions and understandings of CHI as con- chanically elegant Newtonian physics to the messy and,
stituted so far had led to a variety of fates. Some ap- at times, counter-intuitive relativistic physics. In many
proaches, such as affective computing, have found cases, including that of HCI, new paradigms do not dis-
ways to back-fit new phenomena under study to the prove the old paradigms, but instead provide alterna-
information-processing model common in CHI. Some, tive ways of thinking. They often co-exist.
such as ethnographic approaches, have been amalga-
mated to CHI in an uneasy marriage. Some, such as Following Kuhnian lines, a scientific paradigm in HCI
ethnomethodological concerns about the centrality of would contain the following elements:
practices outside those formalized in CHI, have been
heard but not fully worked through, spawning alterna- ! a common understanding of the salient properties
tive fields such as CSCW outside of CHI. In all these of interaction
cases, when force-fitting new insights to old paradigms ! types of questions that appear to be both interest-
CHI fails to capitalize on the full value of these ap- ing and answerable about those properties of interac-
proaches. tion
! a set of broad procedures which can be used to
We will use the rest of this paper to argue that (1) the provide warrantable answers to those questions
commonly acknowledged waves of influence into HCI ! a common understanding of how to interpret the
can be usefully seen in terms of paradigm shift, (2) the results of these procedures
elements of a 3rd paradigm, that of the phenomenologi- These four elements are interdependent and grounded
cal matrix, are in place, (3) the lack of clarity about the in a deeper common conceptualization. For Kuhn, who
epistemological distinctions between paradigms is a derived his theory from analyzing physics, the deeper
limiting factor in the development of the field, and (4) common conceptualization is embodied in the paradig-
CHI can and should make a concerted effort to incor- matic examples that are used in schools to teach the
porate the third paradigm in explicit and programmatic field. A paradigm shift, then, is accompanied by a shift
ways. In order to make this argument, we first need to in the examples which are considered to be central to
define what paradigms mean in the context of HCI. the field. Because of the enormous range of topics
taught as “HCI” and the dearth of classical reproducible
Understanding Paradigms experiments and demonstrations in our field, paradigm
The term ‘paradigm’ as a way to describe waves of re- shifts must be tracked in another way; following Agre’s
search in a field derives from Thomas Kuhn’s theory of theory of generative metaphors in technical work [1,
the structure of scientific revolutions [20]. Kuhn de- pp. 33-48], we argue that paradigm shifts can be
 4

traced in HCI by tracing shifts in the underlying meta- and issues that used to be marginalized have moved to
phor of interaction. the center.

In particular, Agre argues, following a long line of re- The First and Second Paradigms
search in scientific metaphor, that technical fields tend Using this model, we can now characterize the first two
to be structured around particular metaphors which waves of research in HCI. The 1st paradigm, an amal-
suggest the questions that are interesting to ask and gam of engineering and human factors, saw interaction
methods for arriving at answers to them. So, for ex- as a form of man-machine coupling in ways inspired by
ample, the metaphor underlying cognitive science – industrial engineering and ergonomics. The goal of
that human minds are like information processors – work in this paradigm, then, is to optimize the fit be-
suggests questions it could be interesting to ask - how tween humans and machines; the questions to be an-
humans process their input, how they represent infor- swered focus on identifying problems in coupling and
mation internally, how they access memory, etc. - and developing pragmatic solutions to them.
also suggests methods for finding answers to those
questions, for example that we can effectively model The 2nd paradigm, in contrast, is organized around a
human mental activity using computational code and central metaphor of mind and computer as symmetric,
validate these models by comparing computational and coupled information processors. At the center is a set of
human input and output. An important attribute of information processing phenomena or issues in com-
these metaphors is that while they by no means strictly puters and users such as ‘how does information get in’,
dictate what is done in a field, they do bring certain ‘what transformations does it undergo’, ‘how does it go
phenomena into the center of investigation, while mar- out again,’ ‘how can it be communicated efficiently’ etc.
ginalizing others. In cognitive science, for example, it To appropriate Flyvbjerg’s characterization of the state
is relatively straightforward to analyze intellectual, ab- of modern social sciences, it raises “rationality and ra-
stract skills, but it has been more difficult for the field tional analysis to the most important mode of operation
to model embodied skills. for human activity” [11, p. 23]. Left at the margin are
phenomena that are difficult to assimilate to informa-
Following Agre, we argue that central to each paradigm tion processing, such as how people feel about interac-
in HCI is a different metaphor of interaction. Each such tion, the place of a particular interaction in larger sys-
metaphor introduces ‘centers’ and ‘margins’ that drive tems of use, and elusive and enigmatic aspects of
choices about what methods are appropriate for study- everyday life such as “what is fun?”. The point is not
ing and designing interaction and for how knowledge that the margins can’t be talked about - you can make
claims about interaction can be validated. A paradigm an information-processing model of any phenomenon -
shift, then, could be said to occur when a new genera- but that things at the margin are likely to be under-
tive metaphor is driving new choices of what to re- recognized and, when recognized, are likely to pose
search and how, and can be identified when problems persistent problems that are difficult to solve.
 5

This description of the two paradigms that have been information-processing metaphor and the need to de-
dominant in HCI is not intended to imply that all re- velop alternatives.
search projects or researchers fit neatly into one of
these two categories. For one thing, alternative con- One set of issues arises out of work in ubiquitous com-
structions of paradigms are certainly possible. Our goal puting which suggests a renewed centrality for the use-
here is primarily to argue for the existence of a specific, context of computing. While context could be sidelined
additional paradigm for the purposes and goals of un- to some extent in looking at the desktop interface, the
derstanding HCI. appearance of computing embedded everywhere in
both work and personal life has raised the context of
Neither do the paradigms necessarily contradict one computing to a central problem for ubicomp design.
another. Work may be done that cuts across the para- Some methods of dealing with this context follow di-
digms or that exists outside of them entirely. Rather, rectly from the 1st and 2nd paradigms, notably ones that
the paradigms provide broad perspectives that are attempt to identify and optimize information flow be-
useful for sorting out what problems are interesting and tween mobile and ubiquitous devices and their context.
likely to be solved, and to suggest success criteria for These approaches model use-context as yet another
finding their solution. source of information which can be formalized and
transmitted to machines. But approaches to ubicomp
Of course, when paradigms clash, problems may arise. that derive from disciplines such as ethnography, de-
An example of such a clash is the ‘Damaged Merchan- sign, and the arts are based on the idea that use-
dise’ controversy in the mid-‘90’s, in which Gray and context is, in the end, fundamentally unspecifiable and
Salzman argued not only that pragmatically-oriented must be dealt with by other means [e.g. 9].
approaches to usability evaluation are invalid, but also
that usability can only be validated through the scien- Another set of issues arises out of workplace studies,
tifically and theoretically grounded methods of the sec- which focus on the social situation of interaction. These
ond paradigm [15,16]. Similar clashes, we would ar- perspectives have often been hard to reconcile with
gue, are appearing now. In the next section, we de- CHI, leading to their parallel exploration in CSCW. In
scribe emerging strands of research that poorly fit to particular, the centrality of social, situated actions in
the two dominant paradigms at CHI and suggest that a explaining the meaning of interaction is at odds with
3rd paradigm is at hand. the an information-theoretic view of social interaction
that is the core of the 2nd paradigm. [27] Activity the-
Evidence of An Emerging Paradigm ory, for example, is incorporated to the extent that it is
Following our definition of paradigms, a paradigm shift used to create accounts of an existing situation, but not
can be tracked by noticing when phenomena that used in discussions of design or evaluation.
to be at the margin have moved to the center of atten-
tion. In this section, we describe some of the contem- A third set of issues is represented by the situation of
porary strands of research that suggest limits to the learning environments and the politics of their evalua-
 6

tion. K-12 learning goals are quite specified, but met- While each of these issues – and probably quite a few
rics such as user satisfaction and even performance are more - can be seen as a separate critique of what is
only partial indicators of the phenomenon of central marginalized in the prior paradigms, in this paper we
interest, learning. Tutorial programs that supplant the will argue that, taken as a whole, many of these forms
classroom are quite consistent with the 2nd paradigm, of refocusing HCI form a coherent 3rd paradigm based
tying learning tightly to information transfer. However, on several core principles. Next, we delineate those
classroom level interventions that utilize sophisticated, principles and the ways in which they drive research
interdependent claims about fit have by-and-large questions and methods for arriving at their answers in
moved to ICLS which allows discussion of broader con- different ways from the first two paradigms.
texts and goals.
Describing The 3rd Paradigm
A fourth set of issues arises out of the domain of non- We begin with the recognition that one of the themes
task-oriented computing. These approaches tend to that underlies the 3rd paradigm is a focus on embodied
be bad fits to the 1st and 2nd paradigms, whose meth- interaction. Embodiment, of course, also plays a role
ods tend to require problems to be formalized and ex- in the 1st and 2nd paradigms. In human factors, atten-
pressed in terms of tasks, goals and efficiency - pre- tion is paid to such factors as the fit of a mouse to the
cisely what non-task-oriented approaches are intended human hand or the amenability of particular font sizes
to question. It is difficult, for example, to apply usabil- to be easily read. Cognitively based work in HCI has
ity studies to ambient interfaces, since standard laid out physical constraints that usefully inform inter-
evaluation techniques are ‘task-focused’ in the sense of face design such as the speed at which humans are
asking users to pay attention to and evaluate the in- able to react in various situations. Embodiment in the
terface, precisely what the system is devised to avoid. 3rd paradigm is based on a different, central stance
drawing on phenomenology: that the way in which we
Last, yet another set of issues arise out of the mar- come to understand the world, ourselves, and interac-
ginalization of emotion in classic cognitive work. A wide tion derives crucially from our location in a physical and
range of approaches to emotion, notably those of social world as embodied actors.
Picard [25] and Norman [24], has been inspired by
more recent cognitive psychology, which argues that Many in HCI have been introduced to aspects of em-
emotion plays a central role in cognition and models bodiment with Paul Dourish’s book, Where the Action Is
emotion as a type of information flow. But other ap- [8] It emphasizes the concept of ‘engaged action’ as
proaches to affective computing reject the equation of critical to the enterprise: “Embodiment is not a prop-
emotion with information and focus instead on the in- erty of systems, technologies, or artifacts; it is a prop-
terpretation and co-construction of emotion in action in erty of interaction…. In contrast to Cartesian ap-
ways analogous to situated action approaches in work- proaches that separate mind from body and thought
place studies [e.g. 4]. from action, embodied interaction emphasizes their
duality.”
 7

Focusing on embodied interaction substantially changes knowledge is seen as embodied in situated human ac-
what we take as central to interaction. Klemmer, tors. This position leads to a number of intellectual
Hartmann, & Takayama [19], for example, in a review commitments that contrast with those taken by the first
of the literature on embodiment, highlight five central two paradigms.
implications an embodied stance has for the way we
think about interfaces. A focus on embodied interac- THE CONSTRUCTION OF MEANING
tion moves from the 2nd paradigm idea that thinking is The 1st paradigm tends to take a pragmatic approach to
cognitive, abstract, and information-based to one meaning, ignoring it unless it causes a problem, while
where thinking is also achieved through doing things in the 2nd interprets meaning in terms of information
the world, for example expression through gestures, flows. The 3rd paradigm, in contrast, sees meaning and
learning through manipulation, or thinking through meaning construction as a central focus. It adopts the
building prototypes. It suggests that our GUI interfaces stance that meaning is constructed on the fly, often
place too little emphasis on the differential abilities of collaboratively, by people in specific contexts and
the human body, overemphasizing seeing, hearing, and situations, and therefore that interaction itself is an
motor control of our hands, while under-supporting essential element in meaning construction. Meaning
other senses and our physical abilities such as action- derives from information, of course, but in this per-
centered skills and motor memory. It refocuses atten- spective cannot be summed up by mapping information
tion from the single-user / single-computer paradigm flow; it is, instead, irreducibly connected to the view-
that has dominated the 1st and 2nd paradigms towards points, interactions, histories, and local resources avail-
collaboration and communication through physically able to those making sense of the interface and there-
shared objects. It highlights the importance of risk as fore to some extent beyond the reach of formalization.
a positive aspect of embodied practice; there is no Thus, for example, we see research on the value of
undo button in the real world. Finally, it reminds us ambiguity, notably the heavily cited work of Gaver,
that while, under the 1st and 2nd paradigms we have Beaver, & Benford [13].
tended to focus on aspects of activity that are easily
automated, real-world practice is complex and rich, PUTTING USERS IN THEIR PLACE
interleaving physical activity and awareness with ab- If meaning is in some ways irreducibly local, then
stract thoughts, rituals, and social interaction in ways knowledge is strongly situated as well. Following
that defy a purely informational approach. Haraway’s definition [18], situated knowledge refers to
the idea that people’s understanding of the world,
The Essence of the Third Paradigm themselves, and, in the case of HCI, interaction is
Despite the centrality of embodied interaction to the 3rd strongly influenced or perhaps even constructed by
paradigm, it would be a mistake to take physical em- their varying physical and social situations. The move
bodiment – i.e. having a body - as its central, defining to embodiment is consequently a shift to recognizing a
characteristic. Rather, what is central is a phenome- plurality of perspectives. Designing interaction, in turn,
nological viewpoint, in which all action, interaction, and moves from attempting to establish one correct under-
 8

standing and set of metrics of interaction to studying constituting the 3rd paradigm is remarkably catholic,
the local, situated practices of users, taking into ac- ranging from the arts to sociology to policy. The goal
count but not adjudicating the varying and perhaps does not appear to be to establish one of these disci-
conflicting perspectives of users. Aoki & Woodruff, for plines as the gold standard. Indeed, one characteristic
example, argue for the value of CMC systems accom- of the 3rd paradigm is a preference for multiple inter-
modating multiple understandings of what is happening pretations that give a rich sense of the site of interac-
in a relationship [2]. tion over a single, objective description of it [26].

PUTTING INTERFACES IN THEIR PLACE EXPLICIT FOCUS ON VALUES IN DESIGN

One result of a viewpoint that takes situated embodi- Given that the phenomenological perspective highlights
ment as crucial is a renewed emphasis on the impor- the variety of potentially valid viewpoints, evaluation of
tance of place in computing. For example, what makes a system a success can no longer be
McCullough’s Digital Ground [22], which treats ubicomp rooted a priori in measures said to be universally valid.
from an architectural perspective, analyzes the signifi- Instead, we must ask questions about what it means
cance of technologies becoming designed for or de- for a system to be ‘good’ in a particular context – a
signed to adapt to specific locations, times, social question that quickly brings us to issues of values.
situations, and surrounding systems. Broadly, ‘putting Value-based approaches to HCI such as participatory
interfaces in their place’ is grounded in the recognition design and value-sensitive design have come into use
that the specifics of particular contexts greatly define to establish new criteria of success - and therefore of
the meaning and the nature of an interaction. Since all decision-making - in system design and evaluation.
possibilities cannot necessarily be designed for, one [12] All call for some form of explication and explicit
design strategy is to make the computation and the negotiation. Instead of being marginalized, the context
interface embodied. By designing the externalities of of design is brought back as central and filled with
the interface in much the way that robotics has em- questions such as “Who is making the design deci-
braced the idea (and of course, drawing on the ideas of sion?”, “Who is paying for it?”, “What is this saying
the embodied human mind), the device or system does about the user?” and so on. Likewise, in aesthetic
not have to model every contingency. Other strategies evaluation of interfaces, “elegance” is no longer exclu-
include location awareness or situation awareness, for sively premiated; it is just as likely that “appropriate”
example cell phones knowing if they are in a movie or “appropriable” are central aesthetic requirements.
theater or if their owner is in the middle of non-phone
conversation. THE CENTRALITY OF CONTEXT
The 1st and 2nd paradigms acknowledge context primar-
PUTTING RESEARCHERS IN THEIR PLACE ily as “those non-technological factors that affect the
If users’ knowledge is situated, so is that of the re- use of the technology.” Under the 3rd paradigm, we ask
searchers studying them. Compared to the 2nd para- not “how does context give our design meaning?” but
digm, at least, the range of disciplines and perspectives instead “how does our design accommodate the con-
 9

text?” This latter question includes what we do not put the totality of coexisting facts which are conceived of as
into our design, our restraint, or “zensign.” It also en- mutually interdependent” [21, p.240] to explain, pre-
compasses the possibility that the technological system dict, and influence human behavior and experience. In
is reported not because it is particularly unique or at- a curious way, the 3rd paradigm resembles the 1st in its
tractive, but because of how it fits into the particulars ability to recognize issues phenomenologically. How-
of a complex situation. A consequence of this is that ever, rather than eschewing theory, it adopts multiple
context is a central component not only to the problem theories or stances and considers them non-exclusively.
(if any) but also to design and evaluation.
Different Ways Of Seeing
The 3rd Paradigm, Defined To clarify the differences among the paradigms, as well
We are now in a position to define the 3rd paradigm as the ways they can co-exist, let us take a simple and
more precisely. It contains a variety of perspectives hopefully well-known interface example. In the 1960’s,
and approaches whose central metaphor is interaction the United States Air Force developed automated cock-
as phenomenologically situated. The goal for interac- pit warning systems to alert pilots to hazardous condi-
tion is to support situated action in the world, and the tions. The systems used recorded voices to tell pilots to
questions that arise revolve around how to complement turn, climb, or dive to avoid head-on collisions, among
formalized, computational representations and actions other things.
with the rich, complex, and messy situations at hand
around them. The three paradigms are compared in The 1st Paradigm.
Table 1. The situations that drove the initial system design were
classic examples of “critical incidents” [10]. The Air
Because of its emphasis on multiple perspectives, the Force realized they needed to quickly gain the pilots’
3rd paradigm does not espouse a single, correct set of attention. At the time, all pilots and fight controllers
methods or approaches. Instead, as discussed previ- were male, so someone had the bright idea of using a
ously, we see a variety of approaches that are embed- woman’s voice so that it would be immediately identi-
ded in a similar epistemological substrate, like a bio- fied as the “emergency voice”. This was clever and
logical matrix. For this reason, we suggest the term worked well.
the, phenomenological matrix, a multidimensional
characterization of concerns in which relationships and The 2nd Paradigm.
sequences can be defined as a name for the 3rd para- Of course, thinking about it terms of information the-
digm2. It fulfills Kurt Lewin’s demand that we “draw on ory, this not only reduced errors (a fundamental value
of the 1st paradigm), it transmitted information more
efficiently. It is easy to see that there could be a taxon-
2
The name of the paradigm seems to distress many reviewers; omy of voice types created based on cognitive load and
the authors are quite open to alternative names for the para- response times.
digm. In fact, we see the indeterminacy of the name to reflect
the emergent nature of paradigm.
 10

Paradigm 1 Paradigm 2 Paradigm 3

Metaphor Interaction as Interaction as information Interaction as phenomenologically situated

of interac- man-machine communication
tion coupling

Central Optimizing fit Optimizing accuracy and effi- Support for situated action in the world
goal for between man ciency of information transfer
interaction and machine

Typical How can we ! What mismatches come up ! What existing situated activities in the world
questions fix specific in communication between should we support?
of interest problems that computers and people? ! How do users appropriate technologies, and
arise in inter- ! How can we accurately how can we support those appropriations?
action? model what people do? ! How can we support interaction without
! How can we improve the constraining it too strongly by what a computer
efficiency of computer use? can do or understand?
! What are the politics and values at the site
of interaction, and how can we support those in
design?

Table 1: Paradigms compared

The 3rd Paradigm. lers and pilots, this 1st and 2nd paradigm strategy
From the outset, there were design issues based on the ceased to be effective. It also caused interface design-
meaning of this approach. The particular female voice ers to explore new meanings of the gender of the voice.
was reputed to have been selected for its sultry and
seductive tone3. This quality reinforced the idea of the While different paradigms focus on different problems,
space of the cockpit being “male,” echoed in movies like we can see that all three of these perspectives run in
Top Gun. Of course, as women became flight control- parallel – that whatever the solution, pilots should be
warned of peril in a timely fashion, that measurable
improvement in this context is better, and that the
larger issues of the construction of problematic mean-
3
One interesting side effect was to gender popular ing also matter. As we will describe next, the situation
media representations of flight control automata as becomes more challenging when the paradigms come
female. Particularly notable is the original StarTrek in competition.
computer.
 11

Challenges of A “New” Paradigm sions of fourteen models, theories or frameworks of

The description of the 3rd paradigm should not sound HCI, about half of which cannot by themselves lead to
new – many researchers in HCI are already working out CHI publications.
of this framework, although it has not been systemati-
cally recognized as such. One goal of this paper is sim- That is, we find many techniques used in requirements
ply to bring what already appears to be happening in development, but not in the conceptualization or
CHI to the surface for conscious consideration. Indeed, evaluation of the resulting system – at least as repre-
a survey of the 151 long and short papers at CHI 2006 sented in CHI papers. It is as if physicists said “Now
shows that 30 could be thought of as developed from that we have shown that we can create linear accel-
the phenomenological matrix (3rd paradigm). But erators, the findings from these are irrelevant to re-
deeper issues and concerns are involved, as well. search.” And because techniques arising from the 3rd
paradigm are not seen as inherently valid, methods and
The fact of multiple simultaneous paradigms is not in insights from alternative perspectives are often simply
itself a problem – a new paradigm does not disprove an amalgamated to informational or engineering perspec-
old one, instead providing an alternative perspective tives, without recognizing or dealing with the very real
that highlights and addresses alternative phenomena. incompatibilities between these perspectives. The no-
The primary challenge, however for the 3rd paradigm to tion of communities in the CHI conference may be seen
fully bloom is to break out of the standards which have as a reflection of the notion that some new perspec-
been set up by incompatible paradigms. Doing so is not tives ought to be acknowledged.
easy; the result is a series of misappropriations, mis-
understandings, and rejections of work resulting from There are three recurrent, pragmatic difficulties in get-
the 3rd paradigm because it poorly fits ideas of method ting a paper through the CHI review process: (1) the
and validity arising from previous paradigms. legitimacy of only certain kinds of measures of success,
(2) limited understanding of validity of methods outside
Dourish, for example, argues that 20 years after the a limited canon, and (3) insensitivity to important inno-
introduction of ethnography into the HCI canon it is still vation. They are symptomatic of the tensions between
systematically misunderstood as a method for extract- the 1st and 2nd paradigm methods and values, and the
ing user requirements rather than a discipline that actual approaches that pervade HCI today.
analyzes the entire site of human-computer interaction
[7]. Thus, an ethnography, by itself, does not consti- Measures of Success
tute a legitimate CHI publication without an additional In the 2nd paradigm, acceptable measures of success
instrumental component such as user requirements or focus on measuring the comparative effectiveness and
an evaluation of the interface using information- efficiency of information transfer. User self-reported
processing criteria. More recently, we see that Carroll’s satisfaction might suffice, but is seen as a poor cousin
Models, Theories and Frameworks of Human-Computer to efficiency. Measures of success from the 3rd para-
Interaction [6] presents for students extended discus- digm fare a variety of fates when reviewed from this
 12

perspective. Some criteria, such as delight, are not Furthermore, even experiment-based theories that
seen as legitimate criteria at all. Other criteria, such as grapple with highly contextual content are seen as in-
provoking ideas or causing the reader to consider new sufficient, because they are difficult to apply without
possibilities, are not considered sufficient criteria of training and thought. Monk, for example, concludes his
success. Furthermore, balancing the concerns of differ- discussion of Clark’s theory of language as follows: “In
ent stakeholders in a clever way, or enabling activity an ideal world, a theory should be encapsulated as a
that would otherwise simply not be possible are not by- set of guidelines or rules that could be used by a de-
and-large sufficient measures of success. signer with very little background in human factors of
human communication. Falling this, the theory should
To compensate for this, as Grudin [17] has docu- be formalized as principles…. the theory is only really
mented, we see the rise of specialty research commu- usable by researchers….” [23, p. 288]. Insofar as HCI
nities such as ICLS bearing no relationship to the offi- claims to be a scientific discipline, this is a surprising
cial communities in the CHI conference with their own declaration. Insofar as it is an engineering discipline,
conferences and publications. They do not, as they we note that civil engineers are required to have a con-
might, form new sub-disciplines with a more particular siderable understanding of basic physics, followed by
set of methods, values and aesthetics that derive from considerable instruction in how that physics relates to
the CHI paradigm, but rather must adopt independent real materials and conditions before they are certified
standards. to build bridges. It is not the theory’s job to be simpler
than the phenomena it describes. In any case, such
Methods limited guidelines or rules run counter to understand-
Many in HCI bemoan the fact that CHI is poorer for not ings of the complexity of interaction that arise from the
understanding the values or implications of alternative 3rd paradigm.
perspectives, but place the onus on the HCI research-
ers in a potential sub-discipline to analyze the results in Recognizing Innovation, NOT
2nd paradigm terms. As represented in accepted papers, If we wish the field to be consequential, we must ex-
CHI holds controlled experimentation with a few kinds plain important questions. However, many questions
of quantifiable outcomes in extraordinarily high es- cannot be addressed within the 2nd paradigm frame-
teem4. The canon of acceptable methods is even more work. For example, in the 2nd paradigm, there is no
confined than that in psychology since many of the explanation for why people play games or why there
most famous psychological studies involve quasi- are more Windows machines than Macintosh’s. A nice
experimental or demonstration designs [14]. looking interface cannot be evaluated in its own terms,
but rather in functional terms. Don Norman has to cite
4
studies showing that good-looking interfaces produce
By rough count, at least 90 of the long and short papers in
more efficient outcomes to give legitimacy to the notion
CHI 2006 reported quantified results. We cannot, of
course, know how well the accepted papers represent the of emotional design [24]. Furthermore, there are le-
rejection criteria. gitimate questions about equivalency of designs rather
 13

than differences between them that cannot be well ex- tions of what it means to know something is true. Our
plored using statistical methods. goal in this section is to outline the ways in which the
3rd paradigm’s epistemological commitments contrast
For the most part, CHI missed the rise of the Internet; with those of the 1st and 2nd paradigms; these differ-
this is old news, often attributed solely to CHI’s focus ences are summarized in Table 2.
on the very detailed aspects of interfaces and browsers’
ability to present information in different formats. From Objective vs. Subjective Knowledge
the stand-point of a strict cognitive approach, there is The 1st and 2nd paradigms emphasize the importance of
limited language that would describe the general phe- objective knowledge. The 3rd paradigm, in contrast,
nomena of a unified information browsing-, socializing-, sees knowledge as arising from situated viewpoints in
retail-, play-, educational-, and work-environment. the world and often sees the dominant focus on objec-
From a 3rd paradigm point of view, we would not de- tive knowledge as suspect in riding roughshod over the
mand a single unified language but take each of these complexities of multiple perspectives at the scene of
and their confluence as significant. action. As Bannon expresses it in the case of CSCW:
“Our goal was to develop a case against an objective
Different Ways of Knowing reality that can be usefully captured in a model and
The three issues described previously – limited and subsequently used as a sufficient basis on which to de-
inappropriate measures of success, acceptable meth- velop a computerized system” [3]. A number of HCI
ods, and recognition of innovation – can be traced to a researchers have taken it a step further, recognizing
lack of awareness of the epistemological distinctions the subjectivity of the researcher and the relationship
between the paradigms, as a consequence of which 2nd between the researcher and the researched; where
paradigm measures, methods, and phenomena are of- issues of intersubjectivity are common in anthropology,
ten taken as applicable to all forms of CHI work. But they are remote and difficult to address in the 2nd para-
the difference between paradigms is not only one of digm.
different core phenomena, but also different concep-
 14

Paradigm 1 Paradigm 2 Paradigm 3

Appropriate Engineering, Laboratory and theoretical behav- Ethnography, action research, practice-
disciplines for programming, ioral science based research, interaction analysis
interaction ergonomics

Kind of meth- Cool hacks Verified design and evaluation A palette of situated design and
ods strived for methods that can be applied re- evaluation strategies
gardless of context

Legitimate Pragmatic, Objective statements with general Thick description, stakeholder “care-
kinds of knowl- objective de- applicability abouts”
edge tails

How you know You tried it You refute the idea that the differ- You argue about the relationship be-
something is out and it ence between experimental condi- tween your data(s) and what you seek
true worked. tions is due to chance to understand.

Values ! reduce ! optimization ! Construction of meaning is intrinsic

errors ! generalizability wherever pos- to interaction activity
! ad hoc is sible ! what goes on around systems is
OK ! principled evaluation is a priori more interesting than what’s happening
! cool better than ad hoc, since design at the interface
hacks desired can be structured to reflect para- ! “zensign” – what you don’t build is
digm as important as what you do build
! structured design better than ! goal is to grapple with the full com-
unstructured plexity around the system
! reduction of ambiguity
! top-down view of knowledge
Table 2: Epistemological distinctions between the paradigms

Generalized vs. Situated Knowledge tions, it has a greater appreciation for detailed, rich
The 2nd paradigm values generalized models such as descriptions of specific situations. In part, this refers
GOMS. But because the 3rd paradigm sees knowledge back to the arguments around situated action, which
as arising and becoming meaningful in specific situa- argued that while abstract knowledge and formalisms
 15

are certainly useful, they do not directly drive or ex- “Clean” vs. “Messy” Formalisms
plain our activity in the world. In order to better un- The 2nd paradigm, reacting to the a-theoretical orienta-
derstand what people are doing, we need to track the tion of the 1st paradigm, values clean, principled, well-
situated contingencies and strategies people use to ap- defined forms of knowledge. The 3rd paradigm, in con-
ply this abstract knowledge in real situations. Where trast, sees the practical trade-offs in design as more
the 2nd paradigm down-played whether an office had often “messy” rather than principled. Paradigmatic for
books in it or that a computer sitting under a desk pro- the 2nd paradigm, for example, are design spaces,
duced lots of heat when analyzing mouse performance, which are, as Tatar argues [27], clean, mathematical
we all now recognize that “externalities” are often cen- representations of what is at stake in design and sug-
tral figures in the understanding of interaction. gest that design decisions can be made independently
of each other and with little regard for context. Tatar
Information vs. Interpretation contrasts design spaces with ‘design tensions’, a series
The 2nd paradigm arises out of a combination of com- of (non-orthogonal) axes laying out conflicting design
puter science and laboratory behavioral sciences that opportunities that come out in practice, the contextual
emphasize analytic means such as statistical analysis, issues that they impinge upon, and the ways in which
classification and corroboration in making sense of what they may be practically negotiated. The difference be-
is going on at the site of interaction, often under con- tween these ways of thinking is rooted in whether re-
trolled conditions. As Sengers & Gaver argue, how- searchers place the cleanliness and certitude of formal
ever, new approaches to CHI see interaction as stimu- models at the center of their thinking or whether they
lating multiple interpretations in concrete, real-world instead place an appreciation for the complexity of real-
situations, and the job of the evaluator to identify and world, messy behavior and activity at the center.
track those interpretations, often in collaboration with
their ‘subjects’ [26]. The epistemological stance Where’s The Science?
brought to this site is generally hermeneutic, not ana- From a 2nd paradigm point of view, the contribution of
lytic, and focuses on developing wholistic, reflective HCI may be thought to rest on empirical, generalizable,
understanding while staying open to the possibility of scientific results. The 3rd paradigm does not promise to
simultaneous, conflicting interpretation. As Bannon address these. Yet a careful look at the state of the 2nd
writes, “Our critique relied on the centrality of inter- paradigm identifies several kinds of needs for 3rd para-
pretation in the conduct of work, and also on the fact digm thinking.
that the development of computer-based applications
requires the collaboration or involvement of a variety of First, many fields that feature empirical investigation
distinct communities.... [characterized by an] essential such as that advocated by the 2nd paradigm also build
incommensurability of their world views and lan- on a substantial tradition of systematic observation of
guages”. [3] phenomena similar to that advocated by the 3rd para-
digm. For example, the Linnean classification of or-
 16

ganisms was a major empirical contribution to biology paradigms, but we also trust that the reader recognizes
though not, in origin, experimental. the elements of their own work that are in each.

Second, the empirical status of 2nd paradigm thinking is We are not arguing that the 3rd paradigm is right, while
itself subject to question. Critics raise the question of the 1st and 2nd paradigms are wrong. Rather, we argue
whether true scientific theory is possible in the social that paradigms highlight different kinds of questions
sciences on which much 2nd paradigm epistemology is that are interesting and methods for answering them.
based. Flybjerg, for example, argues that “the problem Paradigms frequently co-exist and researchers may
for social studies is that the background conditions work within multiple paradigms. Even so, we believe it
change without the researcher being able to state in would be wise to recognize the differences and incom-
advance which aspects one should hold constant in or- patibilities between paradigms that make them amena-
der for predictions to continue to operate” [11, p. 45]. ble to different sorts of problems; so that, for example,
it would probably be unwise to attempt to uncover the
Third, unlike scientists, even under the 2nd paradigm, rich appropriations of a situated technology with an
we in HCI are not pursuing abstract truth in general, objective laboratory test.
but rather in more particular, technologically defined
ways. We are interested in generalizability, but gener- We also believe it is important for CHI to understand
alizability of meaningful design decisions. For example, that, sometimes, paradigms do clash; those clashes
we no longer do research on emacs keystrokes because may appear in the form of a debates in the field about
the emacs text-editor is no longer widely used. Our proper methodology, validity of results, etc. Work in
principles are almost always local and provisional. one paradigm can easily look invalid to someone work-
ing in another paradigm, because it is based on quite
Thus, in some sense, the science in the 3rd paradigm different notions of what knowledge is and how it is to
bears a similar uneasy relationship to science in the 2nd be generated. Or it may seem valid but beside the
paradigm. Both are ways of coming to know about the point, since the driving questions are different.
world, and both require continual reflection about
goals, purposes, assumptions and legitimacy. And when paradigms clash, the overlap of ways of
seeing taken with conflicting epistemologies results in a
Conclusion miasma of legitimacies. HCI has always been a hybrid
In the opening chapter to HCI: Models, Theories and discipline and therefore has used either the intersection
Frameworks, Jack Carroll describes HCI as a multi- or union of legitimate practices from its constituents.
disciplinary science [6]. By ordering the disciplines into Thus the 2nd paradigm defines legitimacy as measur-
three paradigms, it is our desire to bring some clarity to able utility, and it is this standard to which 3rd paradigm
the field, and begin mapping the relations between work tends to be held. But that is not a priori the defi-
them. We may have used some radical language to nition of legitimacy; to allow the 3rd paradigm to bear
clarify the breaks we see between the 1st, 2nd and 3rd full fruit, we need to recognize and accommodate its
 17

notions of validity. And a cost of work in the 3rd para- 3. Bannon, L. J. 1995. The politics of design: repre-
digm is the need to explicate what is legitimate in the senting work. Commun. ACM 38, 9 (Sep. 1995), 66-
68.
3rd paradigm enterprise. We would expect that any
submission in the 3rd paradigm would explain its phe- 4. Boehner, K., DePaula, R., Dourish, P., and Sengers,
P. 2005. Affect: from information to interaction. In
nomenological matrix and explain (rather than argue
Proc. on Critical Computing ‘05. ACM Press, New
for) its measures of success. York, NY, 59-68.
5. Card, S, Newell, A., & Moran, T. (1983) The Psy-
We trust that if these arguments resonate with the
chology of Human-Computer Interaction, Lawrence
reader, they will take the time to consider alternative Erlbaum Associates, Inc., Mahwah, NJ.
names for or constructions of the 3rd paradigm.
6. Carroll, J. M., (2003) HCI Models, Theories, and
Frameworks: Towards a Multidisciplinary Science
We would expect that calling out the underlying para- Morgan Kaufmann, San Francisco, CA.
digm will become a standard part of every publication
7. Dourish, P. 2006. Implications for design. In
in our field. Thus, we will not be forced into the sort of Proc.CHI '06. ACM Press, New York, NY, 541-550.
pro forma corners that Paul Dourish warned us about at
8. Dourish, P, 2001. Where the action is: the founda-
CHI 2006. [7] Further, it is also reasonable to expect tions of embodied interaction, MIT Press, Cam-
that evaluation of research and new interface ideas will bridge, MA.
become more nuanced and situated, and that richer 9. Dourish, P. 2004. What we talk about when we talk
descriptions (no matter what the paradigm) will be- about context. Personal Ubiquitous Comput. 8, 1
come the standard. In this way, we hope that the 3rd (Feb. 2004), 19-30.
paradigm, just as the 1st and 2nd, can be allowed to 10. Flanagan, J.C. 1954. The critical incident technique.
make a permanent contribution to the field. Psychol Bull. Vol 51, No 4, pp. 327-58.
11. Flyvbjerg, B. 2001. Making Social Science Matter:
ACKNOWLEDGMENTS Why social inquiry fails and how it can succeed
The authors would like to thank Malcolm McCullough for again. Cambridge: Cambridge University Press.
being the first to recognize these three paradigms and 12. Friedman, B. (Ed.) (1997). Human values and the
thank each other for contributing equally to fleshing out design of computer technology. New York: Cam-
this idea. This work was supported in part by NSF bridge University Press and CSLI, Stanford Univer-
sity.
awards IIS-0238132 and IIS-0534445
13. Gaver, W. W., Beaver, J., and Benford, S. 2003.
REFERENCES Ambiguity as a resource for design. In Proc. CHI '03.
1. Agre, P.E. 1997. Computation and Human Experi- ACM Press, New York, NY, 233-240.
ence, Cambridge, UK, Cambridge University Press. 14. Gerrig, R, and Zimbardo, P. G. (2002) Psychology
2. Aoki, P. M. & Woodruff, A. 2005. Making space for and Life (16 th ed.) Boston: Allyn and Bacon.
stories: ambiguity in the design of personal commu- 15. Gray, W. D. 1995. Discount or disservice?: discount
nication systems. In Proc. CHI '05. ACM Press, New usability analysis--evaluation at a bargain price or
York, NY, 181-190. simply damaged merchandise? (Panel session). In
 18

CHI ’95 Conf. Companion. ACM Press, New York, NY, 22. McCullough, M. 2004. Digital Ground. Cambridge,
176-177. MA: MIT Press.
16. Gray, W. D., & Salzman, M. C. 1998. Damaged Mer- 23. Monk, A. (2001) Common Ground in Electronically
chandise? A Review of Experiments that Compare Mediated Communication: Clark's Theory of Lan-
Usability. Human-Computer Interaction, 13(3), 231- guage Use. In Carroll, J. M., (Ed) HCI Models, Theo-
261. ries, and Frameworks: Towards a Multidisciplinary
17. Grudin, J. Three Faces of Human-Computer Interac- Science Morgan Kaufmann, San Francisco, CA
tion." IEEE Annals of the History of Computing. Vol. 24. Norman, D., 2004. Emotional Design: Why We Love
27, no. 4, Oct.-Dec. 2005. pp 46 - 62. (or Hate) Everyday Things. Basic Books, New York.
18. Haraway, D. (1988). Situated Knowledges. Femi- 25. Picard, R., 1997. Affective Computing. MIT Press,
nist Studies, Vol. 14, No. 3., pp. 575-599. Cambridge.
19. Klemmer, S. R., Hartmann, B., and Takayama, L. 26. Sengers, P. and Gaver, B. 2006. Staying open to
2006. How bodies matter: five themes for interac- interpretation: Engaging multiple meanings in de-
tion design. In Proc. DIS '06. ACM Press, New York, sign and evaluation. In Proc. DIS '06. ACM Press,
NY, 140-149. New York, NY, 99-108.
20. Kuhn, T.S. 1970. The Structure of Scientific Revolu- 27. Suchman, L. (1987). Plans and Situated Actions.
tions, (2nd edn). University of Chicago Press, Chi- Cambridge: Cambridge University Press.
cago. 28. Tatar, D. (2007) Design Tensions. Journal of Human
21. Lewin, K. (1951) Field theory in social science; se- Computer Interaction. Forthcoming.
lected theoretical papers. D. Cartwright (ed.). New
York: Harper & Row.