© Kluwer Academic Publishers 2005.

This is the author’s version of the

work. It is posted here for your personal use only. Not for redistribution.
The definitive version published in the Proceedings of the 9th European
Conference on Computer Supported Cooperative Work, September 18-22,
Paris, France.

A Design Theme for Tangible

Interaction: Embodied Facilitation
Eva Hornecker
Interact Lab, Dept. of Informatics, University of Sussex, GB
[email protected]

Abstract. This paper presents parts of a design framework for collaboratively used
tangible interaction systems, focusing on the theme of Embodied Facilitation. Systems
can be interpreted as spaces/structures to act and move in, facilitating some movements
and hindering others. Thus they shape the ways we collaborate, induce collaboration or
make us refrain from it. Tangible interaction systems provide virtual and physical
structure - they truly embody facilitation. Three concepts further refine the theme:
Embodied Constraints, Multiple Access Points and Tailored Representations. These are
broken down into design guidelines and each illustrated with examples.

Introduction
Tangible User Interfaces (TUIs) have become a hot topic in HCI. Until recently,
research was mostly technology-driven, focusing on developing new systems. A
special issue of ‘Personal & Ubiquitous Computing’ on ‘tangible interfaces in
perspective’ (Holmquist, Schmidt and Ullmer, 2004) marks a change in focus
towards conceptual analysis. Yet, there is still a lack of theory on why tangible
interaction works so well (Dourish, 2001). Cooperation support might be the most
important, domain-independent feature of TUIs, but this issue has attracted even
less explicit attention. Many researchers agree that TUIs are especially suited for
collocated collaboration and build systems aimed at group scenarios (e.g. Stanton
et al, 2001; Ullmer and Ishii 2001). Nevertheless, conceptual papers (as in the
mentioned special issue) tend to brush over this issue by briefly mentioning
visibility of actions and distributed loci of control as collaborative affordances.
User studies focusing on group interaction are still scarce, even though we know
from CSCW research that collaborative use often poses different (and possibly
contradictory) requirements to single-user usability. We therefore lack concepts
for analyzing and understanding the collaborative aspects of tangible interaction
and design knowledge on how to design for collaboration.
This paper focuses on part of a framework that offers four themes and a set of
concepts for understanding and designing collaboratively used tangible
interaction systems (for an overview: Hornecker, 2004b). The framework builds
on results from a PhD project on the collaborative use of tangible interfaces
(Hornecker, 2004) and on recent studies in related areas (Hornecker and Stifter,
2004, Hornecker and Bruns, 2004). Just as interaction design aims to create
opportunities for experience, one can design for cooperation and create a ‘force
field’ encouraging and inducing collaboration. The framework aims to help in
creating such ‘force fields’ by offering “design sensitivities” (Ciolfi, 2004,
Fitzpatrick, 2003) and soft guidelines. The framework theme focused on here is
Embodied Facilitation. Tangible interfaces/interaction systems embody
facilitation methods and means by providing structure and rules, both physically
and procedurally. Any application can be understood as offering structure that
implicitly directs user behavior by facilitating some actions, and prohibiting or
hindering others. It thus influences behavior patterns and emerging social
configurations. With Tangible interaction systems, structure is not only in
software, but also physical. They can truly embody facilitation.
I now describe what ‘tangible interaction’ means, summarize the overarching
framework and present the Embodied Interaction theme. The following sections
deal with the concepts relevant to embodied interaction and design guidelines
derived, illustrated by examples. I conclude on open questions and related work.

A Framework for the Design of Tangible Interaction

for Collaborative Use
From the characterizations of tangible interfaces/interaction found in literature,
we can distinguish a data-centered view, pursued in Computer Science and HCI; a
perceptual-motor-centered view, pursued by Industrial and Product Design; and a
space-centered view influenced from Arts and Architecture:
• Data-centered view: Physical representation and manipulation of digital
data (Ullmer and Ishii, 2000; Dourish, 2001) or the interactive coupling of
physical artifacts with “computationally mediated digital information”
(Holmquist, Schmidt and Ullmer, 2004). Research often explores types of
coupling. These systems are usually referred to as “tangible interfaces”.
• Perceptual-motor-centered view: Bodily interaction with objects, exploiting
the “sensory richness and action potential of physical objects”, so “meaning
 is created in the interaction” (Djajadiningrat, Overbeeke and Wensveen,
2004). Design takes account of skills and focuses on expressiveness of
movement, e.g. rhythm, force and style (Buur, Jensen and Djajadiningrat,
2004). The design community prefers the term ‘tangible interaction’.
• Space-centered view: A combination of real space and real objects with
virtual displays (Bongers, 2002). “Interactive systems, physically embedded
within real spaces, which offer opportunities for interacting with tangible
devices, and so trigger display of digital content or reactive behaviors”
(Ciolfi, 2004). This is termed ‘interactive/interactivating spaces’.
The concept of tangible interaction has a much broader scope than Ullmer and
Ishii’s (2000) description of tangible interfaces: “giving physical form to digital
information” and its subsequent physical control, which is often referred to or
used as a definition (data-centered view). Tangible interaction is not restricted to
controlling digital data and includes tangible appliances or the remote control of
real devices. Because it focuses on designing the interaction (instead of the
interface), resulting systems tend less to imitate interaction with screen-based
GUIs (as does placing and moving tokens) and exploit the richness of embodied
action (Buur, Jensen and Djajadiningrat, 2004). Interaction with ‘interactive
spaces’ by walking on sensorized floors or by simply moving in space further
extends our perspective on ‘tangible’ interaction. Instead of using a restrictive
definition that excludes some of these interesting system variants, it seems more
productive to address this larger design space. Thereby we leave the somewhat
artificial confines of any definition behind, and can interpret these attempts at
conceptualization as emphasizing different facets of a related set of systems.

The Design Framework Themes

The framework (Hornecker, 2004b) is structured around four themes, which are
not mutually exclusive, but interrelated, offering different perspectives. Each
theme consists of three or four concepts, which are broken down into concrete
guidelines. In this section will present the four overarching themes and later focus
on one. For each theme a short argument is given as to why it is relevant for
tangible interaction (referring to the definitions given above).
Tangible Interaction Systems for collaborative use should carefully exploit:
• Tangible Manipulation: Tangible Manipulation is bodily interaction with
physical objects. It is interacting with hands and the body. Tangible
interaction is observable and legible, allowing for implicit communication
and peripheral awareness. The objects react in a physical, material way.
Design can deliberately exploit tangibility, emphasizing the direct
interaction with physical objects, which have distinctive material qualities.
• Spatial Interaction: Tangible interaction is embedded in real space. We are
spatial beings; we live and meet each other in space. Our body is a reference
 point for perception. Spatial qualities have psychological meaning. Real
space is inhabited and situated. Real places have an atmosphere. Spatial
interaction is observable and often acquires performative aspects. Design
can exploit the qualities of space and the resources it offers.
• Embodied Facilitation: With tangible interaction we act/move in physical
space and in system space (software). Software defines virtual structure,
determining the interaction flow. Physical space prescribes physical
structure. Both types of structure allow, direct, and limit behavior. Tangible
interaction systems embody structure. Design can enforce social structure
and we can learn from facilitation and pedagogical methods how to do this.
• Expressive Representation: Tangible Interaction is about physical
representation of data. Hybrid representations combine tangible and virtual
elements. These communicate to us and have expression. In interaction we
‘read’ and interpret representations, act on, modify and create them. We
share externalizations of our thinking, which provide shared reference,
remember our traces and document common ground. Design can create
legible, expressive representation.
The framework is organized on three levels of abstraction. The themes offer
perspectives (or viewpoints) and argumentation of an abstract, theoretical level.
They define broad research issues such as the role of space for tangible
interaction. Themes are each concretized with a set of concepts. Concepts provide
analytical tools for describing empirically found phenomena and help to
summarize generic issues, to pinpoint design mistakes and successes. However,
concepts are quite abstract and employing them to support design necessitates
understanding the argumentation behind them. For a design framework, a level of
more directly applicable design guidelines is needed. These should be easily
communicable and comprehensible for people working on practical design
projects, but not interested (or not having time) for the underlying theory.
Furthermore, different researchers and research communities might focus on
different levels. To explain general phenomena or analyze empirical studies,
themes and concepts might be most useful. When designing systems, one might
experimentally follow some guidelines, testing their usefulness and exploring the
design space. To quickly enable people to roughly understand what the more
abstract concepts mean, ‘colloquial versions’ have also been developed. It should
be emphasized that these are not strict rules, but rather soft guidelines, close to
Ciolfi’s (2004) “design sensibilities” or Fitzgerald’s (2003) sensitizing concepts.

Embodied Facilitation
We can interpret systems as spaces or structures to act and move in, thereby
determining usage options and behavior patterns. They enforce social
configurations and direct user behavior by facilitating some movements and
hindering others. Thus, they shape the ways we can collaborate; they can induce
us to collaborate or make us refrain from it. From pedagogy and facilitation we
can learn about how structure, both physical and procedural, can be shaped to
support and direct group processes. With tangible interaction systems, which are
embedded in real space and physically embodied, this space is both a literal one
(physical space and objects) and metaphorical one (software determining action
spaces). Tangible interaction systems can thus truly embody facilitation.
The background that underpins this approach is an exploration of analogies
between interaction design and group pedagogy or facilitation (for details see:
Hornecker 2004c). Both interaction design and facilitation/pedagogy can be
interpreted as the design of ‘spaces for human communication, interaction and
experience’. Similar to architectural spaces, these are appropriated and inhabited
by users. They furthermore offer and prescribe structure, predetermining feasible
adaptation and movement paths. Interaction design cannot ‘design experiences’
just as the structure provided by facilitation can only foster certain experiences or
processes, but not automatically produce them. I became aware of what can be
learned from facilitation and pedagogy for interaction design when evaluating a
system in a group setting (Eden, Hornecker and Scharff, 2002). Seemingly trivial
design decisions (such as system size, placement and number of tools) had a huge
impact on group behavior, session dynamic and atmosphere. My knowledge of
facilitation methods helped to explain these phenomena and informed the systems
redesign. With the theme of Embodied Facilitation, I propose to utilize this
analogy by intent and to apply ‘facilitation knowledge’ to interaction design.
As stated previously, this paper focuses on the Embodied Facilitation theme.
Each theme (offering a specific perspective on tangible interaction) is elaborated
by a set of concepts. The three concepts related to embodied facilitation are now
summarized as a question in colloquial language to give a quick, but rough idea
of what they are about. Then the concepts are explained in detail and the
corresponding design guidelines are presented and illustrated with examples.
Embodied Constraints: Does the physical set-up lead users to collaborate by
subtly constraining their behavior?
Multiple Access Points: Can all users see what’s going on and get their hands
on the central objects of interest?
Tailored Representation: Does the representation build on users’ experience?
Does it connect with their experience and skills and invite them into interaction?

Concept: Embodied Constraints

Constraints restrict what people can do and thereby make some behaviors more
probable than others. Embodied constraints refer to the physical system set-up or
configuration of space and objects. They can ease some types of activity and limit
what people can (easily) do. Thereby they determine probable trajectories of
action. Some embodied constraints provide implicit suggestions to act in a certain
way. Others require people to collectively work around them, leading to the
adoption of interaction patterns that indirectly foster collaboration. Using such
subtle mechanisms, we can encourage and induce people to collaborate. Shape
and size of interaction spaces e.g. act as embodied constraints, which bring
groups together, focusing on a shared object, or which hinder communication.
The design guidelines are:
• Exploit constraints that require groups to:
- distribute the task - help each other out - coordinate action
• Provide a shared ‘transaction space’

Guideline: Exploit Constraints that Induce Helping and Coordination

Sometimes constraints that at first sight seem restrictive and hinder usability have
positive effects on social interaction. In evaluating and redesigning the
Envisionment and Discovery Collaboratory (EDC) (Eden, Hornecker and Scharff,
2002, Hornecker, 2004) we started to use the term ‘embodied constraints’ to
understand and pinpoint some of these phenomena.
The EDC was developed at the Center for Lifelong Learning and Design to
support co-located participatory urban planning (Arias, Eden and Fischer 1997). It
provides an augmented game board and allows tangible interaction with
computational simulations projected upon an aerial photo. We assessed two
system versions by having two groups use them in a role-play of a neighborhood
meeting on re-design of a local bus route. The sessions and subsequent
discussions were videotaped and an interaction analysis was carried out. One
system version uses a horizontal SMARTBOARD™ that allows drawing with
fingers to create, move or delete objects and pen sketching, but cannot handle
simultaneous interactions or detect physical objects. The second system version,
the PITA-BOARD, is based upon a chessboard grid (http://www.dgtprojects.com)
that registers RFID tags embedded in objects. Thus it comes closer to the vision
of a tangible interface with tangible manipulation.

Figure 1. Embodied constraints by structure and size of EDC SMARTBOARD version: (a) helping
each other to change interaction mode (menu in front) and (b) handing over of tools (a pen).
 During analysis we found that constraints forced participants to coordinate
actions, and as a result fostered group awareness and cooperation. Such
constraints can consist of shared or restricted resources that must be coordinated,
or of structures encouraging reciprocal helping. Examples are a menu for
selecting interaction modes on the SMARTBOARD (create, move, delete…) or a
limited supply of tangible tools. The sheer size of the SMARTBOARD necessitated
mutual helping and handing over of tools (figure 1), indirectly fostering
collaboration and awareness. It also made it physically impossible for one person
to take over control of the entire interaction space. Participants found these to be
valuable effects; they advised us to keep the system that large. With the much
smaller PITA-BOARD we observed markedly less of these behaviors. From group
dynamics it is known that situations requiring coordination and help do improve
reciprocal liking and group cohesion. Such situations occurred at the very
beginning of the session and initiated content-neutral cooperation, possibly
making people more willing to cooperate on more salient issues later-on. Working
with interaction modes (one global menu with create, move, delete… tools) had
negative effects from a task-oriented view and led to frequent breakdowns, but
required participants to be highly aware of each other and to coordinate activity.
Here the annoyance was higher than the benefits. Nevertheless, participants could
imagine employing similar (less disruptive) constraints to foster collaboration.
Physical or system constraints requiring coordination and sharing of resources
thus embody facilitation methods that foster cooperation and structure group
processes. From a viewpoint of task analysis, constraints seem counterproductive.
However, easing the task is not the most important goal for all situations; less
straightforward social or cognitive effects may be more critical. Nevertheless, as
the modal interaction example demonstrates, constraints need to be carefully
chosen so as not to disturb and irritate participants. Lessons learned for re-design
included enlargement of the PITA-BOARD, so people would be forced to help
each other and could not control the entire board. We also consciously provided
enough tools for several participants to be active at once, but only a restricted
number of each, so they would need to help each other and coordinate use.

Figure 2. The size of the CLAVIER necessitates several people for a more complex soundscape.
 A further example for embodied constraints originates from a very different
system. Seven installations created by students were shown on three nights in
summer 2002 at a public festival in a park in Bremen. A description and analysis
of the SENSORIC GARDEN, using concepts on interactivity to explain why some
installations successfully attracted visitors’ engagement and what made others
fail, is given in Hornecker and Bruns (2004). Here I focus on the CLAVIER: a
walkway with light sensors triggered by walking across it (figure 2). Colored
spotlights reacted where one put one’s feet. Triggered midi drums and beats
produced an ambient sound environment. Visitors danced to the music, jumped
from light to light and created music. This installation attracted many interactors
and a constant gathering of observers. Some people even danced with umbrellas
in the rain. Others used umbrellas and other objects to trigger multiple sensors.
In several ways the system encouraged people to implicitly and explicitly
cooperate. Visitors, by inadvertently passing, interacted musically with
intentional interactors. Furthermore, its size necessitated the activity of several
people to produce a complex soundscape, as a single person could only trigger a
few adjacent sounds. The installation in this way encouraged group creativity.
While the CLAVIER exemplarily illustrates the spatial interaction theme, these
effects also make it a good example for the embodied constraints given by the
sheer size of an interaction area. Additionally, by necessitating large-scale bodily
interaction, it transforms interaction into a public performance (a concept from
the tangible manipulation theme), makes actions visible, and supports full body
interaction (concepts from spatial interaction theme). This shows how the themes
are interconnected, offering different perspectives on related phenomena.
There is considerable evidence that the physical set-up affects social
interaction patterns, an issue getting relevant in research on distributed displays.
E.g. Rogers and Rodden (2003) found that groups tend to nominate one
participant for writing on a white board and line up before it. When sitting around
a table, roles are more flexible. The physical constraints of a white board mean
that standing in front blocks view and physical access for others. Only one or two
persons can simultaneously have physical access. A point on a table can be
accessed by more people. Buur and Soendergaard (2000) observed different
behaviors and discussion styles for various room set-ups. Needing to stand up and
go to a wall to show a video made people refrain from it. Discussions tended to be
abstract and general. Being able to show clips while staying seated, people would
quickly do so and referred more to concrete video clips and specific observations.

Guideline: Provide A Shared Transaction Space

Kendon (1990) introduced the term transaction space in his explanation of the F-
formation. A persons’ transaction space is formed by the half-circle before the
upper body, that (s)he can see and act within. It is framed by body orientation and
posture. "An F-formation arises whenever two or more people sustain a spatial
and orientational relationship in which the space between them is one to which
they have equal, direct and exclusive access“ (Kendon, 1990, p. 203). If people
stand in a circle or surround a table, their transaction spaces overlap and create a
shared one. Kendon found that establishing, changing and leaving an F-formation
correlates with beginning, participating in, and ending social interaction and that
changes of the configuration give subtle social signals (see also: Suzuki and Kato
1995). As people seem to interpret its establishment as indication that social
interaction is appropriate, implicit creation of an F-formation might stimulate
group interaction. This can explain why surrounding an image on a table produces
a different atmosphere and interaction style than the same image on a wall.
A shared transaction space provides shared focus (if a representational object
attracts attention), while allowing for peripheral awareness. Systems that render
sides of a table unavailable to users affect the shape of transaction spaces (Scott,
Grant and Mandryk, 2003) and thereby the interaction. A transaction space, by
providing exclusive access, also limits communication to those sharing it. There
is a natural limit to its size determined by visibility and audibility.
The focus-providing effect of the EDC’s shared transaction space can be seen
well on evaluation videos. Even from only a bird’s eye view of the table, one can
discern from the rapid activity and gesturing on the SMARTBOARD that people
mostly look at the aerial photo. Nevertheless, the fluidity of interaction and
conversation demonstrates high awareness. Figure 3 a shows a group surrounding
the enlarged PITA-BOARD highly focused on the map and on group activity.

Figure 3. (a) The enlarged PITA-BOARD provides a shared transaction space. (b) Size and form of
the Electrical Telegraphy hands-on exhibit support small group interaction.

An evaluation of a museum exhibition in Vienna on media evolution provided

further examples of the effects of specifically formed transaction spaces
(Hornecker and Stifter, 2004). The exhibition combines traditional object
exhibits, computer-augmented hands-on exhibits, touch screens, interactive
installations and computer terminals. Evaluation combined logfile analysis with
qualitative observation and visitor interviews. Observation revealed interesting
differences in interaction patterns with installations types, in particular in terms of
group sizes. While most touch screens or computer terminals tended to be used by
one visitor and only rarely by two, interactive installations were often surrounded
by groups of up to five persons. Figure 3 b shows a family exploring a hands-on
exhibit on electrical telegraphy. The image illustrates how its size and form limit
the number of people able to focus on it. By providing a hands-on device in the
foreground (not visible: Morse ticker and letter wheel) the screen is moved to the
rear; focus shifts between device and screen. Size and form of an interaction
space (or system) act as a specific type of embodied constraint delimiting access.

Concept: Multiple Access Points

Access points refer to the options to access and actively manipulate relevant
objects. Access is an issue of power, highly influencing group dynamics. We can
analyze systems in terms of the resources they offer for accessing and interacting
with the objects of interest and in terms of privileges and limitations of access.
Restricted resources affect the power play and may even entice people into
conflict and competition for control. Sufficient resources and non-privileged
access create a more egalitarian situation, allowing everyone to participate and to
have a say (abstracting from factors such as hierarchies), making it difficult for
individuals or subgroups to take over control. Access points determine the
opportunities to observe and to become involved hands-on with relevant objects.
Researchers comparing single and multiple mouse conditions for children’s
games found different interaction structures (Stewart et al, 1998). In multi-mouse
conditions significantly more cooperation and communication took place, conflict
was reduced, children interacted more, were more on equal terms and did not
drop out of the activity as much as in the single mouse set-up. Stanton et al (2001)
conclude on a study with tangible props for children’s storytelling: “If everyone
has a prop, then everyone has a vote”. Multiple input devices allow for
simultaneous action, easing active participation, reducing time constraints and
supporting fluent switches between individual and group work (Stewart et al,
1998). By allowing parallel and non-verbal contributions they shift power away
from the verbally articulate, aggressive or self-assured members of groups.
Observational studies of design sessions often find fine-grained
synchronization of simultaneous multimodal activities. Simultaneous activity not
only speeds up interaction, it also displays shared understanding and distributes
ownership (Hornecker, 2004). Visible representations provide focus and shared
reference; they anchor discussions (Arias, Eden and Fischer, 1997, Henderson
1999). Public interaction triggers communication and negotiation. Access points
are influenced by size and form of artifacts and shared space (Scott, Grant and
Mandryk, 2003), determining the physical configuration or arrangement of a
group and affect audibility, visibility and manual accessibility.
The design guidelines are:
 • Give multiple points of interaction
• Allow for simultaneous action
• Give equal access - no privileges

Guideline: Give Multiple Points of Interaction

Multiple interaction objects distribute control in a group, make it difficult for
individuals to take over control, and lower thresholds for shy or timid persons to
become active. Whereas in the original PITA-BOARD version, bus stops were
‘stamped’ with a tool onto the map, the new version provided as many stop
tokens as could be used (figure 4 a). This made it easier to relocate stops and to
keep track of ‘unused’ stops. At the same time, it became difficult for a single
participant to remain in control and set all stops.
There is reason to believe that touching objects creates a sense of ownership
and aids cognitive and emotional appropriation (cp. Buur and Soendergaard,
2000). When distributing creation and manipulation of representations over a
group, these can thus become truly shared objects. This belief was strengthened
by observing the SMARTBOARD-group taking turns in drawing the final bus route
at the end of the session, explicitly involving everyone. Members of a workshop
using a redesigned PITA-BOARD version did the same. While access to the modal
menu on the SMARTBOARD was limited to those next to it (an embodied
constraint enforcing coordination and help), access to the board for other actions
was not restrained. Not being forced to aggressively acquire control over
interaction devices lowers thresholds. Even though there was no equal
distribution (achieving this is probably illusionary), the more quiet or shy group
members gestured lively and made important contributions in manipulating items.
The CLAVIER installation from the SENSORIC GARDEN provides another
example for multiple points of interaction (figure 4 b + c). While here the visitors’
bodies constitute interaction devices, input points are distributed, allowing several
persons to be active without being in each other’s way. This allowed for
incidental simultaneous activity and for cooperative dancing and composing.
The setup as an embodied constraint often also limits access points. In the
exhibition evaluation (Hornecker and Stifter, 2004) it was observed how different

Figure 4: (a) Many interaction objects (PITA-BOARD) and (b, c) input at various loci (CLAVIER)
Figure 5: (a) Terminals suit single users. (b) The ORF-ARCHIVE is used by up to two visitors.

types of installations attracted different visitor constellations. Computer terminals

were almost always used by single persons (figure 5 a), as screens and seating suit
this best. Although of same screen size, the ORF-ARCHIVE (radio and TV clips)
was quite often occupied by pairs. The seat and the small screen allow up to two
people to see and be active. Having only two of these stations gave an incentive,
and the seat seemed to provide a sense of intimacy while being comfortable
enough for two because of sideward space. In contrast, hands-on installations
were frequently surrounded by groups with several people interacting. Several
visitors can move the physical beads of the ABACUS (figure 6 a) at once and the
set-up provides space for observers. The large screens of the GLOBAL STORAGE
(figure 6 b) installation are interacted with via laser beam pens. The large
projection affords many observers and a number of laser pens are attached to long
strings, allowing multiple visitors to move about and be active.
The idea of analyzing size and form of systems in terms of providing access
points originated from observing students work with LEGO MINDSTORMSTM. In
several groups of five, two people only observed and soon got distracted. It was
salient that more than three people can simply not touch the robots simultaneously
(limiting participation in building and testing) and block view for others. The
option to touch something can thus be a scarce resource to start with. Small
objects or surfaces make it difficult to reference via gestures for large groups.
Large objects on the other hand may provide many access points. Yet, their size
means that one can only access a certain subset at a time. The CLAVIER provides
an example where this effect is positive in fostering cooperation.

Figure 6: Hands-on exhibits (a) ABACUS and (b) GLOBAL STORAGE afford small groups
Guideline: Allow for Simultaneous Action
Multiple points of interaction ease simultaneous interaction, but do not
necessarily permit it. Often systems provide several input devices, but require
sequential input, ignoring parallel events or reacting delayed. The PITA-BOARD
allows for simultaneous interaction, while the SMARTBOARD does not. Having to
alternate and sequentialize actions caused multiple breakdowns, even though
participants were highly aware of each other. Alternating actions was felt to be
demanding. Simultaneous interaction speeds up work that can be done in parallel
and thereby helps the group to concentrate on issues requiring negotiation and on
developing shared understanding. It also allows less vocal group members to have
a say, as they do not need to wait for a free time slot or need to interrupt.

Figure 7: Simultaneous action on new PITA-BOARD (a) introductory phase (b) mapping land use

Most examples given in the previous section for multiple points of interaction
apply here as well. Simultaneous interaction thus supports multiple points of
interaction. Yet, it is not a design guideline that should be followed slavishly.
Physical constraints that sequentialize actions can serve to give necessary order to
an interaction process or to ensure equal rights (e.g. a waiting queue).

Guideline: Give Equal Access – No Privileges

Privileged access to system features naturally gives more power to those
privileged. Besides affecting the interaction process it changes the atmosphere by
evoking certain assumptions and expectations, in particular by delivering implicit
social signals on hierarchies and expertise. Equal access refers to giving
everybody equal options; it does not mean everybody should have one of every
tool or that all interaction devices should provide the same functionality.
In assessing the EDC (Eden, Hornecker and Scharff 2002) we found that
privileged access of facilitators to system functions affected the power play of
sessions. Facilitator access to PITA-BOARD features via mouse and keyboard,
invisible and unpredictable to participants, made them feel as guests, not allowed
to ‘own’ the system space. In comparison, the SMARTBOARD group quickly
learned how to close error messages (appearing on the table) and took over this
Figure 8. (a) A menu pops up unexpectedly within participants’ manual space, who cannot see
facilitator actions. (b) The new PITA-BOARD version has an extra ‘admin-space’.

task. Making the means of controlling the system invisible and non-observable
does not enable users to learn and become ‘experts’. Providing privileged access
to some group members gives implicit signs of ownership. When re-designing the
PITA-BOARD, we eliminated privileged access by providing means to control the
simulation by manipulating objects on the game board (Figure 8 b). Combined
with other improvements, the new version provided a much better experience,
allowed for equal access, and enabled everybody to take over system control.
Another kind of privilege relates to optimal viewpoints, due to e.g. a vertical
screen next to a table (cp. Scott, Grant, Mandryk, 2003). With the EDC, there was
no optimal, and therefore privileged place, as most icons were easily identifiable,
even if upside down, and the aerial photo has no implicit orientation. Orientation
and positioning are much more critical for text. Privileged viewpoints are a result
of the type of representation used as well as size and form of interaction spaces.

Concept: Tailored Representations

The concept of tailored representations refers to a different type of access, which
is cognitive and emotional instead of manual or visual. Discussions of tangible
interfaces often highlight the intuitiveness of interaction. Focusing only on
intuitiveness neglects the skills and knowledge of people (cp. Buur, Jensen and
Djajadiningrat, 2004) and may result in systems that don’t scale up to experienced
users and complex domains. While intuitive usability is important in giving new
users access, we also must consider expert users and specialists. Representations
that connect with users’ experiences and skills invite them into interaction and
empower them. Representations that do not connect do exclude and silence users,
who cannot relate, understand, and contribute. Representations need to be
adequate for the task, the domain and the user group. Intuitiveness is thus relative.
Nevertheless, it is important to ease initial access on the basic level of
manipulating relevant objects. If we cannot figure out how to interact with a
system, it is of no help if the representation is legible. Users should be able to
quickly explore the basic syntax of interaction. Over time they might acquire the
more complex syntax of advanced interaction (learnability). Experience-
orientation thus refers more to the semantics of interacting with a representation.
 Representations that build upon users’ experiences can become tools for
thought as thinking prop or external memory, and can complement verbal
communication by allowing people to gesture, refer to visible objects and
manually demonstrate something (Norman, 1994; Hutchins and Klausen, 1998).
Adequately chosen representations thereby ease participation in discussions.
Henderson (1999) describes the gatekeeper function of design representations that
control access, invite or discourage participation and define the range of allowed
actions. Representations can privilege perspectives (different notations being
easier to read and manipulate for specific professions) and become symbolically
owned territory. Good representations offer several layers of legibility, are
accessible for people with differing knowledge areas, and provide a shared
reference. Then they can serve as boundary objects (Star and Griesemer, 1989).
Another aspect of representations and materials is that these trigger people’s
imagination and creativity (Rettig, 1994). The selection of materials provides a
trajectory for thought, for the positive or the negative. What is not available or not
visible will be thought of less. Similar to facilitators, system designers should be
aware of the responsibility they carry in deciding upon available materials and
representations, as these might affect the decisions of people using them.
The design guidelines are:
• Build on the experience of the group and its members
• Make the interaction intuitive enough for easy access
• Allow the semantics to rely on specific knowledge

Guidelines on Intuitiveness and Experience-Orientation

Interaction with the PITA-BOARD tokens was perceived by all participants as
intuitive. In analyzing the videos, no interaction problems could be detected after
an initial phase of finding out how to place tokens on the board. A new
introductory phase for exploring the system in a playful way (figure 7 a) gave
participants the opportunity to get accustomed to its reactions. Many common
methods for citizen participation in urban design use aerials as maps distort and
abstract geographical relations. Furthermore, map reading must be learned, it is an
acquired skill. Aerials might relate more to inhabitants’ experience, with
landmarks being easy to identify and street shapes visible. Ernesto Arias (personal
communication) emphasizes the importance of selecting an appropriate level of
abstraction in participatory urban design, such as very literal, figurative building
blocks: “Some laypeople need a tree-tree, not a green general block”. After a
while categories become well known and more abstract blocks can be introduced.
The systems introduced so far have all been of the ‘walk up and use’ kind,
meant for public places or participatory meetings. A recent study provides a better
example of the difference between intuitiveness and experience-orientation.
Together with students I carried out a user study on the TANGIBLE IMAGE QUERY
Figure 9. (a + b) Study participant at the TANGIBLE IMAGE QUERY with query results

(Matkovich et al 2004). This system offers architects inspiration through

serendipitous searching in collections of images. Users define a search by laying
colored objects onto the input area. The underlying algorithm searches for color
distributions. The study participants were architecture and computing students.
While manipulation of input and querying was intuitive, the search results
required habituation and were initially irritating. One major finding was that the
attitude of participants towards the system depended largely on their relation with
images (as architects, art lovers or avid photographers) and their ability to find
value in being inspired and surprised (instead of finding “what I searched for”).
Examples for this concept are not yet sufficient, as the systems studied so far
did not address specialized and experienced professional users and the TANGIBLE
IMAGE QUERY has no collaborative use context. Nevertheless the concept is
important to Embodied Facilitation, and therefore needs to be presented here.

Conclusions and Outlook

In this paper I presented a theme for design and analysis of collaboratively used
tangible interaction. Tangible interaction encompasses a broad scope of system
and interfaces sharing aspects of tangibility, physical embodiment of data, bodily
or embodied interaction and embedding in real space. It encompasses approaches
from HCI, computer science, product design and interactive arts. Following a
short summary of the overall framework, I focused on the Embodied Facilitation
theme. Its basic idea is that tangible interaction systems provide procedural as
well as physical and spatial structure, which shapes the ways we act. It can induce
collaboration, foster it or make us refrain from it. Thus, tangible interaction
systems embody styles, methods and means of facilitation.
The theme of Embodied Facilitation was broken down into three concepts.
These were concretized with design guidelines and illustrated with examples.
Embodied Constraints are aspects of the physical set-up that subtly constrain
peoples’ behavior or provide implicit suggestions for action, encouraging
collaboration. The guidelines suggest (a) employing constraints that require
groups to distribute the task, to help each other out and to coordinate action, and
(b) providing shared ‘transaction spaces’. The concept of Multiple Access Points
makes us consider systems in terms of how many people can see what is going on
and lay hands on the objects of interest. The guidelines suggest (c) giving
multiple points of interaction, (d) allowing simultaneous action and (e) giving
equal access, not privileging some users. Tailored Representations take account
of users’ experiences and skills, inviting and empowering them. The guidelines
suggest (f) building on experience and (g) making interaction intuitive enough for
easy access, but (h) allowing the semantics to rely on specific knowledge.
In its current state, the overall design framework should be read as a proposal,
backed by examples and arguments. There are several directions for future
research. To demonstrate its utility as a design framework, practical design
studies employing concepts and guidelines are required. These could involve
design of new and redesign of existing systems, or adjusting a previously single-
user system to collaborative use. Other studies could systematically explore the
design space given by specific (sets of) guidelines. A further direction for
research relates to the frameworks’ general applicability to CSCW. Illustrative
examples so far stem predominantly from entertainment, design and negotiation
support. To demonstrate its general utility, examples from other application areas
are required. It is furthermore open whether the framework covers collaboration
distributed over time and space. Further research questions concern the relations
between some of the concepts and guidelines. Transaction spaces and access
points are clearly positively related. Seen as absolutes, multiple access points and
constraints are in tension. Different configurations may prompt different
interaction patterns, such as providing a tool for every second or third group
member. Similar questions could be studied in detail empirically. The concrete
influence of size and form of interaction spaces or the number of access points is
still unclear. Is there a systematic relation between task, access points, number of
actors and evolving interaction patterns? Considering the number of guidelines in
the overall framework there will be many more detailed research questions.
It is important to remember that the guidelines are meant to sensitize designers,
not to be slavishly followed. While it is tempting to make concepts operational,
we need to be wary of transforming analytic terms, meant to sharpen perception,
into rules and measurements. Design needs sensitivity and judgment. Sometimes
it might even be best to temporarily discourage collaboration, prevent observation
and restrict access, turning the guidelines from do’s into don’t’s. Which
guidelines should be applied and which take precedence over others, will depend
on the task and the larger context of an activity, requiring further investigation on
indicators for the applicability of guidelines and priorities in-between guidelines.
The contributions of this paper towards understanding the relation of embodied
interaction and collaboration consist of: framework themes and concepts which
support high-level analysis; complemented with guidelines to support design; and
a research agenda. The framework is illustrated with several examples. It
furthermore contributes to research on interactive exhibits, where space is an
intrinsic issue (e.g. Ciolfi, 2004), as these served as major illustrative domain.
To round up, I will put my own framework into the context of related work.
There are several frameworks aimed at the design for social interaction and a
number of frameworks on tangible interfaces/interaction. With its soft guidelines
and ‘design sensitivities’ my framework shares characteristics with others that
offer concepts as ‘sensitizing devices’ and support designing for social interaction
(Ciolfi, 2004, Dourish, 2001, Fitzpatrick, 2003). These frameworks are not
prescriptive, do not offer recipes, and thus need to be interpreted and appropriated
in response to concrete situations. Although operationalized to a greater extent,
the framework presented here is meant to be continually evolving and open.
Previous frameworks on tangible interfaces/interaction have focused mainly on
defining terms, categorizing, and characterizing systems (e.g. Ullmer and Ishii,
2000, some articles in ‘Pervasive & Ubiquitous Computing’ special issue 2004).
While supporting the structural analysis of systems and detection of uncharted
territory, these approaches offer little advice when designing for specific real
world situations. Furthermore, these frameworks seldom address the human
interaction experience or are restricted to solitary users. Suzuki and Kato (1995)
and Arias et al (1997) did pioneering work on acquiring a better understanding of
how tangible interaction affords social interaction and collaboration, but found
few followers. Even though many TUIs supporting collaborations have been
developed and some field-tested, analysis often remains domain-specific and
yields few generalizable concepts (for a literature overview see Hornecker 2004).
This framework contributes to the larger research agenda of Embodied
Interaction. While sharing the goal of understanding tangible interaction with
Dourish (2001), my view on embodiment is more in line with Robertson (1997,
2002). Dourish’s perspective on embodiment focuses on the social construction of
meaning, whereas Robertsons starting point (in the tradition of French
phenomenologist Merleau-Ponty) is the living, feeling, responsive body as our
primary means of experiencing the world, the world being its milieu. In embodied
interaction the living body encounters and enters into dialogue with the world.
Dourish (2001) states that social action is embedded in settings, which are not
only material, but also social, cultural and historical, focusing his analysis on the
latter. While the social has been elaborated, materiality has been less discussed.
Understanding system embodiment in the sense of being physically manifested
takes materiality seriously. I aim to unfold these aspects, inquiring into the
interweaving of the material/physical and the social. Similar to Robertson (1997)
and Fitzpatrick (2003) I am interested in how we accomplish communication and
collaboration and how designed environments can support this.
Several framework themes and concepts not focused upon within this paper
relate to topics discussed by other authors. E.g. social and atmospheric qualities
of places (Ciolfi, 2004, Dourish, 2001) are part of the spatial interaction theme.
The concepts of non-fragmented visibility and performative action are related to
Dourish’s (2001) discussion of accountability and observable action and build
heavily on work from Robertson (1997). The concept of embodied constraints is
at the same time related to and in intrinsic tension with configurability, focused
on by other authors as an important system quality, but often with little reference
to collaboration (Dourish, 2001, Jaccucci, 2004). This is a productive tension, as
understanding the effects of embodied constraints makes the needs for
configurability apparent. Moreover, it may give us insight on where exactly
configurability is desirable and where (and how) system designers should provide
structure – at least initially – in order for social processes to start evolving (cp.
Hornecker 2004c). That “by configuring space in different ways, different kinds
of behaviours can be supported” has often been stated (e.g. Dourish 2001).
However discussion usually stops here. There have been only few attempts (e.g.
Rogers and Rodden, 2003) to dig deeper and understand these relations.
Affordances as ‘exploitation of physical constraints’ are often merely seen in
terms of usability and provision of legible cues. With my framework and in
particular with the theme of Embodied Facilitation presented here I extend the
analysis to less straightforward, indirect (or second-order) social effects.

Acknowledgments
Jacob Buur pushed and helped me to go on, identify core ideas, fuse them into
structure, focus on the essentials, and keep it simple. Thanks for the challenge and
encouragement. Years ago Michael Heger gave me and my HDA-TG teammates
a hands-on education in facilitation that imbues this work. Mark Stringer, Paul
Marshall and Geraldine Fitzpatrick commented on different versions of this paper
and helped polishing it. Special thanks to Geraldine for getting me to Sussex.

References
Arias, E., Eden, H. and Fischer, G. (1997). ‘Enhancing Communication, Facilitating Shared
Understanding, and Creating Better Artifacts by Integrating Physical and Computational
Media for Design’. In Proc. of DIS '97, ACM. pp.1-12.
Bongers, B. (2002): ‘Interactivating Spaces’. In Proc. of Symposium on Systems Research in the
Arts. Informatics and Cybernetics.
Buur, J., Jensen, M.V. Djajadiningrat, T. (2004). ‘Hands-only scenarios and video action walls:
novel methods for tangible user interaction design’. In Proc. of DIS’04. ACM, pp. 185-192.
Buur, J. and Soendergaard, A. (2000): ‘Video Card Game: An augmented environment for User
Centred Design discussions’. In Proc. of DARE’00, ACM. pp. 63-69.
Ciolfi, L. (2004): "Situating 'Place' in Interaction Design: Enhancing the User Experience in
Interactive Environments". Ph.D. Thesis, University of Limerick.
Djajadiningrat, T., Overbeeke, K. and Wensveen, S. (2002): ‘But how, Donald, tell us how?’ In
Proc. of DIS'02, ACM. pp. 285-291.
Dourish P. (2001): Where the Action Is. The Foundations of Embodied Interaction. MIT Press.
Eden H., Hornecker E. and Scharff E. (2002): ‘Multilevel Design and Role Play: Experiences in
Assessing Support for Neighborhood Participation’. In Proc. of DIS'02, ACM. pp. 387-392.
Fitzpatrick, G. (2003) The Locales Framework: Understanding and designing for Wicked
Problems. Kluwer Publishers
Henderson, K. (1999): On Line and On Paper. Cambridge, MIT Press.
Holmquist, L., Schmidt, A. and Ullmer, B. (2004): ‘Tangible interfaces in perspective: Guest
editors’ introduction’. Personal and Ubiquitous Computing 8(5). pp. 291-293.
Hornecker, E. (2004): Tangible User Interfaces als kooperationsunterstützendes Medium. PhD-
thesis. University of Bremen. Dept. of Computing, July 2004.
Hornecker, E. (2004b): ‘A Framework for the Design of Tangible Interaction for Collaborative
Use’. In Proc. of Danish HCI Research Symposion. University of Aalborg. pp.57-61.
Hornecker, E. (2004c): ‘Analogies from Didactics and Moderation/Facilitation Methods: Design-
ing Spaces for Interaction and Experience’. Digital Creativity Vol. 15 No. 4. pp. 239-244
Hornecker, E. and Bruns F.W. (2004): ‘Interactive Installations Analysis - Interaction Design of a
Sensory Garden Event.’ In Proc. of IFAC/IFIP/IFORS/IEA Symposium on the Analysis,
Design and Evaluation of Human-Machine Systems.
Hornecker, E. and Stifter, M. (2004). Evaluationsstudie Ausstellung medien.welten Technisches
Museum Wien. Unpublished project report. TU Vienna & TMW
Hutchins, E. and Klausen, T. (1998): ‘Distributed Cognition in an Airline Cockpit’. In Engeström,
Middleton (eds.) Cognition and Communication at Work. Cambridge Univ. Press. pp. 15-34.
Jacucci, G. (2004): Interaction as Performance. Cases of configuring physical interfaces in mixed
media. PhD thesis, University of Oulu
Kendon, A. (1990): ‘Spatial organization in social encounters: The F-formation system’. In
Kendon (1990). Conducting interaction. Cambridge Unviversity Press. pp. 209-237.
Matkovic, K., et al (2004): ‘Tangible Image Query’. In Proc. of Smart Graphics. pp. 31-42.
Norman, D. (1994): Things that Make Us Smart. Addison Wesley, Reading. Mass.
Rettig, M. (1994): ‘Prototyping for Tiny Fingers’. Communications of the ACM 37 (4), pp.21-27.
Robertson T. (2002). ‘The Public Availability of Actions and Artefacts’. CSCW 11 (3-4), 299-316.
Robertson T. (1997). ‘Cooperative Work and Lived Cognition. A Taxonomy of Embodied
Actions’. In Proc. of E-CSCW'97, pp. 205-220.
Rogers, Y. and Rodden, T. (2003): ‘Configuring spaces and surfaces to support collaborative
interactions’ In: O' Hara, et al (eds.) Public and Situated Displays. Kluwer. pp.45-79
Scott, S. D., Grant K. D and Mandryk R. L. (2003): ‘System Guidelines for Co-located
Collaborative Work on a Tabletop display’. In Proc. of E-CSCW'03. pp. 159-178
Stanton, D. et al. (2001): ‘Classroom Collaboration in the Design of Tangible Interfaces for
Storytelling’. In Proc. of CHI'01, ACM. pp. 482-489.
Star, S. L. and Griesemer, J. (1989): ‘Institutional Ecology, "Translations” and Boundary Objects’.
Social Studies of Science 19. pp. 387-42.
Stewart, J., Rayborn, E., Bederson, B. and Druin, A. (1998): ‘When Two Hands are Better Than
One’. In Proc. of CHI'98, Extended Abstracts, ACM. pp. 287-288.
Suzuki H. and Kato H. (1995): ‘Interaction-level support for collaborative learning: Algoblocks -
an open programming language’. In Proc. of CSCL 1995, pp. 349-355.
Ullmer B. and Ishii H. (2000): ‘Emerging frameworks for tangible user interfaces’. IBM Systems
Journal 39(3-4), pp. 915-931.