ArticlePDF Available

Abstract and Figures

This paper develops a representation-theoretic notion of spatial context for cognitive agents interacting with spatial environments. We discuss the current state of the art in defining context as used in context-aware and/or location- aware systems. In contrast to existing approaches, we define context through cognitive processes. The term ‘invisible geography’ alludes to the fact that knowledge about geographic space develops through complex cognitive interaction and is not simply ‘out there’ to be looked at. Placing (cognitive) processes in the focus of our context definition allows for a truly user-centered perspective: conceptualizations imbue spatial structures with meaning. This allows for fixing terminological problems and relating context definitions to work in spatial information theory and cognitive science. Although we focus on spatial context, the approach is generic and can be adapted to other domains in which cognitive aspects concerning users of information systems are central. @InProceedings{freksa_et_al:DSP:2007:980, author = {Christian Freksa and Alexander Klippel and Stephan Winter}, title = {A Cognitive Perspective on Spatial Context}, booktitle = {Spatial Cognition: Specialization and Integration}, year = {2007}, editor = {Anthony G. Cohn and Christian Freksa and Bernhard Nebel }, number = {05491}, series = {Dagstuhl Seminar Proceedings}, ISSN = {1862-4405}, publisher = {Internationales Begegnungs- und Forschungszentrum f{"u}r Informatik (IBFI), Schloss Dagstuhl, Germany}, address = {Dagstuhl, Germany}, URL = {}, annote = {Keywords: Representation theory, spatial context, location aware systems} }
Content may be subject to copyright.
A Cognitive Perspective on Spatial Context
Christian Freksa1, Alexander Klippel2, Stephan Winter3
1Transregional Collaborative Research Center Spatial Cognition &
Cognitive Systems Group, University of Bremen, Germany
2Cooperative Research Centre for Spatial Information
Department of Geomatics, The University of Melbourne, Australia
3Department of Geomatics, The University of Melbourne, Australia
Abstract. This paper develops a representation-theoretic notion of spatial
context for cognitive agents that interact with spatial environments. We discuss
the state of the art in defining context as used in context-aware and / or
location-aware systems. In contrast to existing approaches, we define context
through cognitive processes. Placing cognitive processes in the focus of our
context definition allows for a truly user-centered perspective: conceptuali-
zations imbue spatial structures with meaning. This allows for fixing termino-
logical problems and relating context definitions to work in spatial information
theory and cognitive science. Although we focus on spatial context, the
approach is generic and can be adapted to other dom ains in which cognitive
aspects concerning users of information systems are central.
1 Introduction
Context has become an omnipresent notion in human-computer interaction (HCI)
research. Geographic information systems and services are concerned in particular
with context-aware or location-aware systems. The general idea of context research is
to adapt the reasoning of a system / service to current requirements (e.g. location and /
or task), and hence, to make the information generated by the system more useful for
its user.
It has been considered difficult, however, to define what constitutes context.
Popular definitions remain unspecific, and most attempts to fill the concept of context
with meaning do it by examples, or – more systematically – by a taxonomy of aspects
of context. As a consequence, the list of influencing factors of possible constituents of
context get out of hand instead of rendering the concept of context more precise,.
In the present paper we take an orthogonal approach. Rather than decomposing
context effects into the different aspects that may play a role we define context in
terms of the cognitive architecture that determines the interactions between the
components involved. As a starting point, we use cognitive processes that allow for a
Dagstuhl Seminar Proceedings 05491
Spatial Cognition: Specialization and Integration
characterization of context, in particular
spatial context. From such an operational
definition of context we expect a better understanding of context effects and on
requirements to be taken into account when dealing with context.
The paper is structured as follows: We briefly introduce the notion of context as
found in the ubiquitous computing literature and point out weaknesses of current
context definitions. On this basis we develop a cognitive architecture for wayfinding
problems to exemplify interactions that take place in human spatial problem solving.
We discuss the trilateral relationship between an environment, a cognitive agent, and
a cartographic map and the interactions that take place between these three entities.
On this basis we define spatial context. We conclude by discussing possible
applications and give an outlook how this approach can be used to overcome
deficiencies in more general context definitions.
2 The notion of context in ubiquitous computing
Spatial context came into the focus of research with the concept of ubiquitous
computing (Weiser, 1991). Ubiquitous computing aims to provide services
everywhere and at any time that take into account features of the actual environment
and situation; hence, ubiquitous computing requires information about the
environment as well as about the situation and goals of the cognitive agent. For that
purpose, ubiquitous computing concepts employ sensors that collect data on the user’s
location as well as environmental parameters. Interface design research has been
aware of the separation of the physical environment and its representation in digital
space for a long time (cf. Ishii & Ullmer, 1997). However, this separation usually
reflects the provider’s perspective and ignores the individual user with her knowledge,
abilities, focus of attention, or emotions. For example, a mobile navigation system
that only considers the environment and its representation in digital space would
neither take into account the cognitive map of the user nor her spatial abilities.
The term ‘context’ itself has become ubiquitous in the research literature. It is used
in combinations such as context-aware’ systems (Abowd et al., 1997; Dey, 1998;
Kjeldskov et al., 2003), or, for specific contexts such as location, in respective
combinations such as ‘location-aware’ systems (e.g., Nicklas et al., 2001; Want &
Schilit, 2001; Winter, 2003). The frequently cited survey by Chen and Kotz (2000)
clarifies that context-awareness means that applications have to adapt to changing
context instead of producing prefabricated content. For example, a (location-aware)
mobile navigation system adapts automatically to the changing location of the mobile
user and specifies route directions with respect to this location without further user
interaction. In contrast, a web service for locating street addresses might come up
with a similar map, but will not be considered context-aware since all parameters have
to be explicitly specified by the user without taking into account the location at which
the query is specified. Other authors distinguish between reactive systems that adapt
to the current context, and proactive systems that anticipate future context (Mayrhofer
et al., 2003). In any case, time, location, and change play an important role for
In this literature, almost all authors agree that it is difficult to def
ine the term
‘context’. A generally accepted definition does not exist and the term is frequently
used with unspecific meaning. According to Dey (1998), context is “any information
that can be used to characterize the situation of entities that are considered relevant to
the interaction between a user and an application, including the user and the
application themselves.” To precisiate the concept of ‘context’ the literature has
developed taxonomies of aspects that together form context. The influential taxonomy
by Schilit et al. (1994) names spatial context (where you are), social context (who you
are with), and computing context (what resources are nearby), a taxonomy that is
widely considered incomplete (see, e.g., Chen & Kotz, 2000). Alternatively, Dix et al.
(2000) distinguish infrastructure context, system context, domain context, and
physical context. This illustrates that the categorization of the notion ‘context’ in turn
depends on the specific context for which the notion is used; there seems to be no
natural categorization of context. Categorizations of context frequently are made ad-
hoc without formal methodology, and hence without proof of completeness or
Some of the aspects of context identified in ubiquitous computing now receive
attention in the spatial information theory literature. Among the first is cultural
context in cross-language studies (Levinson, 2003; Mark, Skupin, & Smith, 2001;
Mark & Turk, 2003). Another one is temporal and spatial context in characterizing the
salience of spatial features (Elias, 2003; Winter et al., 2005). This literature typically
avoids defining or categorizing specific features of context.
Dix et al. (2000) acknowledge that context-awareness is not a question of a system
interface, but of the broader circumstances in which the system is applied, including
the physical environment. From that perspective they focus on location. They
consider location and environment in terms both of the physical space and its
representation in the map system. They implicitly introduce what we will call
‘environment’ by means of nearness and set up an algebraic specification for the type
space, consisting of location, nearness, and regions, and for the type world, consisting
of spaces and bodies. With these elements they set up a kind of top-level ontology of
spatial context in the environment in the mind of a wayfinder or in the system that
computes maps. So far, this is the only formal approach to define spatial context in
ubiquitous computing.
Instead of providing a new taxonomy for contexts we will investigate in the
following sections how contexts are created and used; this will help us to provide an
operational definition of context. When we focus on spatial context, we follow Dix et
al. in their argumentation that spatial relations form a fundamental aspect of context
for location-aware systems.
3 The relation between spatial environment, cognitive agent, and
cartographic map
Let us now render the notion of spatial context more precise. In contrast to
approaches that are concerned with the potential factors contributing to context, we
will detail the cognitive and computational functions of negotiating knowledge in a
complex system. This system distinguishes the major comp
onents in which these
different factors may play a role. We will discuss the roles of spatial contexts in the
framework of the trilateral relationship between a spatial environment, a cognitive
agent interacting with this environment, and an external representation of that
environment (specifically: a map) that the agent may use to support this interaction.
Why are maps useful for our spatial orientation in an environment in which we ar e
immersed and to which we have direct visual access? To answer this question, we will
look at the kinds of entities and structures that are involved in solving orientation and
wayfinding tasks. We will distinguish between the spatial environment E, in which
the orientation or wayfinding task is to be carried out; the cognitive agent A – a
person or a robot who carries out the task; and the map M that serves as a tool for
performing the task. These three entities are involved in rather sophisticated cognitive
interaction processes when we use maps to solve orientation or wayfinding problems.
To reduce the complexity in presenting this trilateral relationship, we will carry out
a Gedankenexperiment involving the three entities E, A, and M in the paradigm of
synthetic psychology as introduced by Braitenberg (1984) and discussed for spatial
communication with maps, for example by Frank (2000). We will begin with a simple
configuration and analyze interaction processes that may take place in this
configuration; we then will gradually augment the configuration and we will
investigate from a knowledge representation-theoretic perspective in which ways the
augmentation influences the interactions. On this basis we develop a representation-
theoretic characterization of spatial context applicable to spatial reasoning and spatial
3.1 An agent without cognition in a spatial environments
The Gedankenexperiment starts by considering a spatial environment E and primitive
agents A (amoebas or other abulic agents) to whom we would not concede any
cognitive capabilities. How do amoebas move in a spatial environment? In a
structured environment, their tracks will not be equally distributed random spatial
configurations; rather, the tracks will be influenced by the initial position and by the
physical structure and the physical forces acting in the environment. For example, if
the environment consists of hills and water streams, the movements of the amoebae
are guided to follow the spatio-temporal course of the water streams. Those parts of
the spatial environment that influence the motion of the amoeba belong to the spatial
situation context of the amoeba. The abulic agent completely depends on the
affordances of the environment that will determine where they move (cf., Gibson,
1979). This can be regarded as a weak version of “knowledge in the world” (e.g.,
Norman, 1980; Raubal & Worboys, 1999); it also can detail the very origin of this
Although we discuss movements in our Gedankenexperiment, we will not
emphasize the issue of temporal context and of changing knowledge about the
environment, here,
A cognitive agent without mental representation of its spatial environment
How does the situation change when we replace the primitive agent by a cognitive
agent – specifically by a human being or a cognitive robot? The physical affordances
of the environment will still determine to a large extent where the agent will move
(see Fig. 1): one of the fundamental aspects of affordance, of course, is gravity; it will
keep the agent on the ground, for the most part. Other aspects are passages that are
easy to traverse and obstacles that will prevent the agent from moving to certain
places. Besides the affordances imposed by the environment certain affordances are
imposed by size, shape, and abilities of the agent: the agent can perform certain
movements on the basis of its anatomy and physiology; certain other movements are
not possible. In general, affordances are determined by the interaction between agents
and their environments: for instance, the size of an agent interacts with the size of a
passage: the relationship between these sizes will determine the affordance of certain
movements between the agent and the environment.
Fig 1. Affordances emerge in the interaction between environment (E) and agent (A).
Let us now consider the cognitive side of the agent: the agent wants to move to some
specific location in the environment, say to the exit of the building he is in. Agents
with low-level cognition (e.g., insects or reactive robots) may employ rather primitive
reactive mechanisms to move to their destination that do not require an internal
representation of their environment in their minds (e.g., Brooks, 1991); thus, simple
cognitive affordances relating to the agent’s perception and action capabilities can be
engaged in addition to the purely physical affordances discussed in the previous
3.3 A cognitive agent with mental representation of its spatial environment
Higher cognitive animals like rats, humans, or cognitive robots build up internal
representations of their environment (frequently referred to as cognitive maps) that
help them to plan and c
ontrol their movements in space (see Fig. 2). If a wayfinder’s
destination is not directly accessible to perception, a mental representation that
functions as a memory for the structure of the environment is necessary to plan and
carry out actions that will get the agent to its destination. Certain aspects of the spatial
structure of the environment are represented in the memory of the cognitive agent to
allow him, her, or it to reflect about the world and to plan actions to be carried out in
the spatial environment.
Why is it economical to represent aspects of the world in which we are immersed?
The answer is simple: if we can use a mental representation as a model of an
environment we can carry out certain operations mentally that would otherwise
require physical actions in the spatial environment itself. Besides the savings of
physical energy and time through mental operations, there may be advantages due to
suitable representation structures (Sloman, 1985), as these mental structures are not
replicas of the environment. In addition, mental operations may be much less
dangerous and harmful than the corresponding physical actions.
Fig. 2. Two worlds and a representation (correspondence) relation (cf. Palmer, 1978). In this
example the arrows indicate the relation ‘further north than’.
In short: we have information about the spatial environment twice: in the world and
in the mind. The two information sources are connected through a representational
correspondence (Palmer, 1978). Certain tasks are achieved more economically by
taking a ‘shortcut’ through the mental representation than by taking action in the
3.4 A cognitive agent with mental representation of its spatial environment
and a map
From the perspective of cognitive architecture the situation becomes much more
complex when a map as a third element is integrated. If information about the
environment is available in two incarnations in the environment and in its mental
representation why do we need maps to find our way? A map is a third source of
information about the spatial environment besides the two we dealt with in the
Mental Representation
previous section. The answer is simple again: a map enables a cognitive agent to solve
spatial problems that he can solve neither by inspecting the environment nor by
inspecting its mental representation.
A map can replace neither the environment nor its mental representation; however,
a map can extend our cognitive capabilities in certain settings (e.g., Scaife & Rogers,
1996): (1) a map can provide information about environments which we are not
immersed in and / or which we have never seen before; (2) a map can provide
information about environments we have seen before but whose details have escaped
our mind; and (3) a map can provide information about environments we are
immersed in for which it may be difficult or impossible to get an overview; it enables
us to get a global view of the environment that allows us to apply certain spatial
reasoning mechanisms. Thus, a map can extend our mental representation of an
environment and our mental representation can interact with this external
representation to extend the range of problems we can solve.
We now have three sources of information about the spatial environment: the
environment E itself, the mental representation of the agent A, and the external
representation in form of a map M. The three information sources each are in a
correspondence relation to the other two; these relations are depicted in Fig. 3:
Fig. 3. The relation C1 establishes a correspondence between the environment E and the mental
representation of agent A; C2 establishes a correspondence between this mental representation
and the map M; and through composition of C1 and C2, C3 sets some aspects of the
environment E in relation to the map M.
The correspondences between the three sources of spatial information are established
in different ways: C1 is more or less hardwired by means of the agent’s sensory
organs and / or established in early phases of getting acquainted with spatial
environments (e.g., Clark, 1973; Wilson, 2002); in humans, the correspondence
established by our perceptual / cognitive machinery becomes so strong that we
sometimes are unable to distinguish between ‘what is out there’ in the real
environment and what we know about it; a single correspondence relation is
The relation C2 is different in that it is not specified by the development of the
perceptual apparatus: making and interpreting external representations of mental
images is an art cognitive agents develop much later and there is not one ‘natural’
way of externalizing mental images or of interpreting external spatial representations.
A variety of correspondence relations can be established here. A map can be
conceived of as an abstract picture that can be interpreted in different ways and
independently of the represented environment. A cognitive agent can use a map to
reason about spatial relations even if no corresponding spatial environment exists; the
relation C2 can be substituted for the relation C1 in such a way that the map becomes
the target spatial environment or the representation of a fantasy world.
The correspondence relation C3 between the environment and the map
representation is different again: it is established by the agent on a high cognitive
level by composing the relations C1 and C2. An external depiction of something is
never a representation by virtue of the intrinsic properties of the ‘something’ and the
depiction; it becomes a representation by explicitly establishing correspondence
relations (Furbach et al., 1985; Palmer, 1978).
As we pursue a cognitive perspective, we will only consider cognitively relevant
aspects of these three entities with respect to the spatial tasks to be solved. In the
environment E, these are the locations of physical objects present, their spatial
relations with respect to each other, their shape and other properties of appearance,
their visibility, their uniqueness and / or their distinctiveness in the environment, and
possibly further aspects. Regarding agent A, we are concerned about (1) perceptual
spatial abilities (specifically vision and audition, and possibly the sense of smell); (2)
spatial memory abilities (specifically the ability to remember previously perceived
environments and / or representations of environments); (3) abstraction abilities (in
particular: abilities to develop a mental image of a real environment, to generate and
interpret maps, and to relate the different representations); (4) imagery abilities
(mental ‘visualization’ of those memories); (5) mental reasoning abilities
(transformation of perception and memories into other forms); (6) spatial action
abilities (the transformation of new insights about the spatial environment into
physical actions in the environment; (7) spatial interaction abilities (abilities to
communicate with other cognitive agents about spatial situations); these abilities
involve (8) abilities to relate and integrate spaces of different scale and type: table top
space, vista space, environmental space, ... (Montello, 1993) and (9) abilities to
employ different spatial reference systems (Levinson, 1996) and to transform from
one reference system to another. In the map M, we are concerned about adequate
symbols and relations for depicting and interpreting spatial relations in a consistent
and unambiguous way.
3.5 Cognitive processes related to the three spatial information sources
So far, we discussed representational correspondences as if they were static
relationships. However, in as far as they are established by cognitive processes, we
should point out the importance of dynamic aspects in establishing representational
correspondences in spatial domains. The affordances of the spatial environment
determine to a large extent how people perceive an environment, as we interpret the
world largely in terms of its functions and its presumed ‘purposes’. They are
important to generate expectations about what will or what might happen next. For
humans, a small pathway in a meadow may be more salient and memorable than large
branches of a tree while for birds and monkeys it may be the other way around; thus
actual or potential actions and events structure the environment into relevant and
irrelevant aspects (e.g., Richardson & Spivey, 2000).
The actions and events that may take place in the environment are reflected in
mental capabilities in human mental representations: we can imagine the same type of
actions and events mentally; in fact, it is much easier for us to imagine realistic events
like a person walking down a pathway by mental simulation than fictional events like
the disintegration or reconfiguration of the environment. Similarly, we use external
maps to physically simulate actions like journeys by moving our finger across the line
symbols that correspond to the pathways of the journey in the real environment or we
mentally simulate such actions by traversing these line symbols with our visual
perception and attention apparatus.
These examples and other evidence suggest that the spatial correspondence
between different information sources is particularly useful for establishing a direct
process correspondence between processes in the spatial environment, perceptual
attention processes in vista space and table-top space, mental imagery processes, and
manipulation processes in table-top space (Freksa, 2004).
4 Spatial context
In the foregoing sections we established a general representation-theoretic framework
involving a spatial environment, one or more cognitive agents, and external
representations of the spatial environment; we will now use this framework to
characterize various types of spatial context without differentiating between different
aspects that may be involved in these contexts; instead, we will distinguish types of
contexts on the basis of their role in the framework.
It is evident that in different cognitive domains different aspects are relevant and
therefore different contexts apply; in our case we are interested in contexts in the
environment, in the mental representation of the agent, in the external map
representation, and in their mutual interactions. In the following, we will briefly
sketch examples of such contexts. Again, we will stress the interaction between
different entities in the architecture of a complex cognitive system involving the
spatial environment, the cognitive agent, and an external representation structure. In
this way, the agent will be the focus of attention as he contributes the cognitively
active parts of the overall system. Contexts are then determined through cognitive
functions that are involved in the interactions of the architectural components. To this
end it becomes possible to relate the definition of (spatial) context to work in
cognitive science and approaches of research on ontologies. A process-oriented
characterization of context is a necessary requirement for the integration of context
concepts in modern information systems. We will provide a short classification of
spatial contexts and exemplify their role for the domain of wayfinding.
Situation context
The spatial situation context of an object or of an agent is the spatial structure in the
physical environment that this object or agent is embedded in. The available physical,
perceptual, or cognitive processes will determine which structures influence a given
situation and thus must be considered as part of the respective spatial context. For
example, the relevant situation context for the movements of an amoeba consists of
those spatial structures in the spatial vicinity of the amoeba that affect its motion
pattern. For cognitive agents, the relevant spatial situation context varies depending
on the focus if interest: are we interested in visual influences, auditory influences, or
in the agent’s disposition with respect to air flow and its temperature, or in a
combination of various relevant factors. From a cognitive perspective we would argue
that the relevancy of factors is determined by their role for the conceptualization
process of the cognitive agent, i.e. the instantiation of a representation that takes into
account several sources. This form of representation has been discussed under various
names, for example, current conceptual representation (Habel, 2003), conceptual
structure (Jackendoff, 1997), current spatial representation (Klippel et al., 2003). We
will briefly describe the approaches from the domain of wayfinding and route
directions that exemplify a formalization of simple conceptualization processes.
Duckham and Kulik (Duckham & Kulik, 2003) expand an approach by (Mark,
1985) on calculating a simplest paths. The general idea is to find a path in a network
of paths that matches the criterion of being easy to describe. This approach is in
contrast to other approaches that calculate, for example, the shortest connection
between two locations. Conceptualization processes are a precondition for (verbal)
descriptions of routes and verbalizations can be used as a window to these
conceptualizations. The formalization of the “ease of description” by Mark (1985) can
therefore be seen as a formalization of a conceptualization process, hence providing a
formal description of a spatial context in the sense used in this article. The frame with
slots that Mark uses for the characterization of actions at intersections and
corresponding descriptions is the specification of a spatial context.
A similar yet antipodal approach is taken by Richter and Klippel (2005). Instead of
providing a single description for finding the best matching paths in a network, a
variety of descriptions is given to find the best conceptualizations for a given route.
Each conceptualization is suited to identify and characterize a spatial context.
4.2 Mental context
Processes in the mental representation of spatial environments are affected not only
by perceptions of the environment but also by activated memory contents (e.g.,
Baddeley, 1986). For example, if a human cognitive agent has been mentally engaged
in the dangers of wildlife, she or he will be more likely to suspect and associate
dangerous creatures with natural spatial environments than otherwise. This mental
spatial context may be activated by certain features in the environmental context, but
it is independent of the fact whether or not wild animals actually exist in the perceived
Map context
Accordingly, map context relates to spatial entities in the map that may affect the map
generation and interpretation processes. For map generation, these may be entities that
influence the map generalization process; for map interpretation, these may be entities
that capture the map reader’s attention.
4.4 Other contexts relevant in spatial reasoning, action, interaction
Our representation-theoretic characterization was restricted to the three sources of
spatial information E, A, and M. However, if we augment the model, for example to
include natural language as a source of information about space, a language context
will get involved.
Not only the system components themselves but also their interactions can be used
to define contexts. For example, an experienced map reader will apply different map
interpretation processes than a novice; thus, the substructure (Frank, 2000) defines a
context in which certain interpretations are generated while others are not. Or in the
communication between two agents that each engage their own language with
personal vocabulary and personal background knowledge, a specific communication
context is created in which certain types of exchanges that are suited to this context
are generated while others are not.
5 Application
In a more or less static environment the location of the wayfinder will change during a
physical wayfinding process; thus, at least the location is of spatio-temporal nature. A
location-aware, location-adaptive, or location-based system interacts with the
wayfinder and reacts to a change of her location at the same time. This means that
location is used with the characteristics of a context as defined in the ubiquitous
computing literature.
Location per se is not a context. We can assume location to exist independently of
a perceiving mind. But it is the perceiving mind that identifies gestalt and affordance
in the signals of perceptions of the environment and applies cognitive processes to
focalize (Bal, 1997) experience of space either in internal cognitive representations or
in external representations. In other words, the focalizer uses location to create a
spatial context. Thereby the same location can be used to create different spatial
We will now discuss the creation of spatial context from location by our process-
oriented perspective, applying the principles introduced above.
5.1 Location and environment
Each subject (wayfinder) and object (e.g., her mobile device) has a unique location in
the physical world at each point in time. This location can be specified in terms of a
-dimensional body in relation to the rest of the spatial environment. Each body
can be imposed with axes, giving it an orientation, and a center (frequently abstracted
as position of the entire object). We consider location to specify a relation to other
objects, while position specifies a relation in an (otherwise empty) reference frame.
Location and position can be determined by perception and / or computation. Bodies
can move, and thus, location can change. Individual body movements can be quite
complex, and hence, are typically generalized and abstracted in mental representations
(e.g., {activity, start, end} or {activity, start, direction}) and external representations
(e.g., trajectories).
5.2 Location and wayfinder’s mind
The perceiving mind of a wayfinder focalizes location into a spatial context of I-am-
here. The internal spatio-temporal concept of I-am-here-now extremely depends on
other contexts. A child playing hide-and-seek will have a relatively detailed idea of I-
am-here, and a person experiencing ‘Europe-in-ten-days’ will have a relative coarse
idea, maybe two-dimensional if not one-dimensional (see also Read & Budiarto,
2003), even if both share the same location. They have a different perspective on their
immediate environment, and they perceive different entities in their environment, in
terms of potential activities. The same is true, for example, for a pedestrian and a
bicyclist, who are at the same location.
A primitive agent without cognitive capabilities (see Section 3.1) is a purely
reactive agent. Sensed physical affordances lead to hard-wired motor actions. With no
cognitive processes involved, the agent is located, but establishes no spatial context.
A cognitive agent without a mental representation of its environment can perceive
gestalt and affordances (see Section 3.2), and will focalize perceptions at least to a
level of planning and controlling future actions. Spatial context is minimal, but
depends fully on these cognitive processes; these processes, in turn, depend on the
embodiment of the cognizer: a wheeled robot plans and controls actions differently
then a legged robot, for example.
A cognitive agent with mental representation of her environment (see Section 3.3)
focalizes perceptions to an internal representation; the agent establishes
representations of the relationships of her body to other bodies in the environment
depending on current cognitive processes. Hence, location is transformed in
(complex) spatial context.
5.3 Location and external map
An external map represents location of a moving agent typically by reducing the
notion to position. A familiar representation of position on a map is a point in a two-
dimensional space, a projection of the earth’s surface to a plane surface. The point is
characterized by coordinates in a specific spatial reference system (the mapping
system), and possibly by a covariance matrix describing positional accuracy. In an
alternative form of representation, position can be characterized qualitatively (e.g., in
sketch maps).
In these cases, location is constructed by means of positioning techniques. The
location of the mobile positioning device is only an approximation of the location of
the wayfinder, since they are two different bodies and use two different sensing
techniques to derive their position. Depending on the respective positioning
technology of a device, its position can be represented by GPS coordinates, cell IDs of
a wireless communication network, or coordinates matched to a particular travel
network (Schiller & Voisard, 2004; Scott-Young & Kealy, 2002). Furthermore, the
current location of the positioning device – which becomes the location on the map –
can differ from the location represented on the map, due to inaccurate or outdated
So far, we have considered position as a representative of location. Nevertheless,
by putting the position on a map the map making agent establishes relations between
represented objects and the current position of the agent. The agent does this by
applying cartographic variables like selection, accentuation, generalization, or
displacement. The controlled application of these variables is, again, focalization,
based on cognitive processes.
6 Conclusions
Nevertheless, the definitions of context hedge to render the term more precise and
rather add aspects to it. The current paper tried a different approach by focusing on
spatial context and relying on cognitive processes as a means for defining context.
The general approach taken is a representation-theoretic characterization of the
trilateral interactions that take place when a cognitive agent is active in an
environment aided by a map-like representation.
In this article we considered the notion of spatial context in terms of cognitive
processes involved in the interaction between cognitive agents, spatial environments,
and cartographic maps. Defining spatial context through cognitive processes
(especially spatial cognitive processes) allows for the integration of several currently
discussed topics, for example, principles of embodied cognition, such as cognitive
off-loading (Wilson, 2002), that are regarded as most useful in spatial tasks. Our
approach develops a framework for context to demonstrate the relationships between
environmental spatial context, mental spatial context, and map spatial context for a
The presented formal method constitutes an operational approach to characterize
specific spatial contexts involved in cognitive interactions. Our method does not
incorporate user studies regarding specific features or aspects that may have to be
taken into account in cognitive modeling; rather, it presents the architecture of a
model for cognitive processing into which the results of such studies easily can be
incorporated. Categories of context are formed through the type of knowledge
engaged in the cognitive processes, not by ad-hoc decisions. This representation-
theoretic approach makes the representation relations between different cognitively
relevant domains explicit and can be applied to other aspects of context equally well
and will clarify the notions of context in the corresponding approaches.
Funding by the German Research Foundation (DFG) for the Transregional
Collaborative Research Center SFB/TR 8 Spatial Cognition at the Universities
Bremen and Freiburg and by the Collaborative Research Centre for Spatial
Information, Department of Geomatics, The University of Melbourne, Australia, is
gratefully acknowledged.
Abowd, G. D., Atkeson, C. G., Hong, J., Long, S., Kooper, R., & Pinkerton, M.
(1997). Cyberguide: A Mobile Context-Aware Tour Guide. ACM Wireless
Networks, 3(5), 421-433.
Baddeley, A. D. (1986). Working memory. New York: Oxford University Press.
Bal, M. (1997). Narratology: Introduction to the Theory of Narrative (Second Edition
ed.). Toronto: University of Toronto Press.
Braitenberg, V. (1984). Vehicles. Experiments in synthetic psychology. Cambridge,
MA: MIT Press.
Brooks, R. (1991). Intelligence without representation. Artificial Intelligence, 47, 139-
Chen, G., & Kotz, D. (2000). A Survey of Context-Aware Mobile Computing Research
(Dartmouth Computer Science Technical Report TR2000-381). Hanover,
NH: Department of Computer Science, Dartmouth College.
Clark, H. H. (1973). Space, time, semantics, and the child. In T. E. Moore (Ed.),
Cognitive development and the acquisition of language (pp. 28-63). New
York: Academic Press.
Dey, A. K. (1998). Context-Aware Computing: The CyberDesk Project. Paper
presented at the AAAI '98 Spring Symposium, Stanford, CA.
Dix, A., Rodden, T., Davies, N., Trevor, J., Friday, A., & Palfreyman, K. (2000).
Exploiting space and location as a design framework for interactive mobile
systems. ACM Transactions on Computer-Human Interaction, 7(3), 285-321.
Duckham, M., & Kulik, L. (2003). "Simples" paths: Automated route selection for
navigation. In W. Kuhn, M. Worboys & S. Timpf (Eds.), Spatial Information
Theory: Foundations of Geographic Information Science. Conference on
Spatial Information Theory (COSIT) 2003. (pp. 182-199). Berlin: Springer.
Elias, B. (2003). Extracting Landmarks with Data Mining Methods. In W. Kuhn, M.
F. Worboys & S. Timpf (Eds.), Spatial Information Theory (Vol. 2825, pp.
398-412). Berlin: Springer.
Frank, A. U. (2000). Spatial communication with maps: Defining the correctness of
maps using a multi-agent simulation. In C. Freksa, W. Brauer, C. Habel & K.
F. Wender (Eds.), Spatial cognition II: Integrating abstract theories,
empirical studies, formal methods, and practical applications. Berlin:
Freksa, C. (2004). Spatial Cognition - an AI perspective. In R. López de Mantaras &
L. Saitta (Eds.), ECAI 2004. Amsterdam: IOS Press.
Furbach, U., Dirlich, G., & Freksa, C. (1985).
Towards a theory of knowledge
representation systems. In W. Bibel & B. Petkoff (Eds.), Artificial
intelligence methodology, systems, applications (pp. 77-84). Amsterdam:
Gibson, J. J. (1979). The ecological approach to visual perception. Boston, MA:
Houghton Mifflin.
Habel, C. (2003). Incremental generation of multimodal route instructions. Paper
presented at the Natural Language Generation in Spoken and Written
Dialogue, AAAI Spring Symposium 2003, Palo Alto, CA.
Ishii, H., & Ullmer, B. (1997). Tangible Bits: Towards Seamless Interfaces between
People, Bits and Atoms. Paper presented at the Conference on Human
Factors in Computing Systems (CHI '97), Atlanta.
Jackendoff, R. (1997). The architecture of the language faculty. Cambridge, MA:
MIT Press.
Kjeldskov, J., Howard, S., Murphy, J., Carroll, J., Vetere, F., & Graham, C. (2003).
Designing TramMate – a context aware mobile system supporting use of
public transportation. Paper presented at the Designing User Interface 2003
Conference, San Francisco, CA.
Klippel, A., Tappe, T., & Habel, C. (2003). Pictorial Representations of Routes:
Chunking Route Segments during Comprehension. In C. Freksa, W. Brauer,
C. Habel & K. F. Wender (Eds.), Spatial Cognition III. Routes and
Navigation, Human Memory and Learning, Spatial Representation and
Spatial Learning. (pp. 11-33). Berlin: Springer.
Levinson, S. C. (1996). Frames of reference and Molyneux’s question: Cross-
linguistic evidence. In P. Bloom, M. Peterson, L. Nadel & M. Garrett (Eds.),
Language and Space (pp. 109-169). Cambridge, MA: MIT Press.
Levinson, S. C. (2003). Space in Language and Cognition. Cambridge: Cambridge
University Press.
Mark, D. M. (1985). Automated route selection for navigation. IEEE Aerospace and
Electronic Systems Magazine, 1, 2-5.
Mark, D. M., Skupin, A., & Smith, B. (2001). Features, Objects, and Other Things:
Ontological Distinctions in the Geographic Domain. In D. R. Montello (Ed.),
Spatial Information Theory (Vol. 2205, pp. 489-502). Berlin: Springer.
Mark, D. M., & Turk, A. G. (2003). Landscape Categories in Yindjibarndi. In W.
Kuhn, M. F. Worboys & S. Timpf (Eds.), Spatial Information Theory (Vol.
2825, pp. 28-45). Berlin: Springer.
Mayrhofer, R., Radi, H., & Ferscha, A. (2003). Recognizing and Predicting Context
by Learning from User Behavior. Paper presented at the International
Conference On Advances in Mobile Multimedia (MoMM2003), Linz,
Montello, D. R. (1993). Scale and multiple psychologies of space. In A. U. Frank & I.
Campari (Eds.), Spatial information theory: A theoretical basis for GIS. (pp.
312-321). Berlin: Springer.
Nicklas, D., Großmann, M., Schwarz, T., Volz, S., & Mitschang, B. (2001). A Model-
Based Open Architecture for Mobile, Spatially-Aware Applications. In C. S.
Jensen, M. Schneider, B. Seeger & V. J. Tsotras (Eds.), Advances in Spatial
and Temporal Databases (Vol. 2121, pp. 117-135). Berlin: Springer.
Norman, D. A. (1980).
The Psychology of Everyday Things. New York: Basic Books.
Palmer, S. E. (1978). Fundamental aspects of cognitive representation. In E. Rosch &
B. B. Lloyd (Eds.), Cognition and categorization (pp. 259-303). Hillsdale,
NJ: Lawrence Erlbaum.
Raubal, M., & Worboys, M. (1999). A formal model of the process of wayfinding in
built environments. In C. Freksa & D. M. Mark (Eds.), Spatial information
theory. Cognitive and computational foundations of geographic information
science. (pp. 381-399). Berlin: Springer.
Read, S., & Budiarto, L. (2003). Human scales: Understanding places of centering
and de-centering. Paper presented at the Fourth International Symposium on
Space Syntax, London, UK.
Richardson, D. C., & Spivey, M. J. (2000). Representation, space and Hollywood
Squares: Looking at things that aren't there anymore. Cognition, 76, 269-295.
Richter, K.-F., & Klippel, A. (2005). A model for context-specific route directions. In
C. Freksa, M. Knauff & B. Krieg-Brueckner (Eds.), Spatial Cognition IV.
Reasoning, Action, and Interaction: International Conference Spatial
Cognition 2004 (pp. 58-78). Berlin: Springer.
Scaife, M., & Rogers, Y. (1996). External cognition: how do graphical representations
work? International Journal of Human-Computer Studies, 45, 185-213.
Schilit, B., Adams, N., & Want, R. (1994). Context-Aware Computing Applications.
Paper presented at the IEEE Workshop on Mobile Computing Systems and
Applications, Santa Cruz, CA.
Schiller, J., & Voisard, A. (2004). Location-Based Services. San Francisco: Elsevier.
Scott-Young, S., & Kealy, A. (2002). An Intelligent Navigation Solution for Land
Mobile Location Based Services. Journal of Navigation, 55, 225-240.
Sloman, A. (1985). Why we need many knowledge representation formalisms. In M.
Bramer (Ed.), Research and development in expert systems, Proceedings
BCS Expert Systems Conf. 1984. Cambridge, MA: Cambridge University
Want, R., & Schilit, B. (2001). Expanding the Horizons of Location-Aware
Computing. IEEE Computer Journal, 34(8), 31-34.
Weiser, M. (1991). The Computer for the Twenty-First Century. Scientific
American(9), 94-104.
Wilson, M. (2002). Six view on embodied cognition. Psychonomic Bulletin and
Review, 9, 625-636.
Winter, S. (2003). Route Adaptive Selection of Salient Features. In W. Kuhn, M. F.
Worboys & S. Timpf (Eds.), Spatial Information Theory (Vol. 2825, pp.
320-334). Berlin: Springer.
Winter, S., Raubal, M., & Nothegger, C. (2005). Focalizing Measures of Salience for
Wayfinding. In L. Meng, A. Zipf & T. Reichenbacher (Eds.), Map-based
Mobile Services - Theories, Methods and Implementations (pp. 127-142).
Berlin: Springer Geosciences.
... Furthermore, sensor-based localization methods typically provide a user's location on a 2-D map (i.e., allocentric information). It requires extra mental effort to interpret the map and associate it with egocentric perception of the environment [6], [7], largely owing to the different mental mechanisms for processing allocentric and egocentric information [8]. ...
... In the near future, wearable cameras and computing devices may enable new technologies for online personal services based on first-person-view (FPV) input [13], [14]. In this paper, we propose a wearable virtual usher (WVU) for indoor navigation using a wearable camera driven by cognitive models of wayfinding [6], [15]. The basic idea is to provide aided wayfinding [15], whereby the user follows a set of verbal instructions from the wearable system with less cognitive load. ...
... For each sequence, we apply MIM and MSR to each frame image first. Then, we obtain four sequences of location labels (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18) for each test route. We use the leave-one-out cross validation to train and test the HMM models for each route. ...
Inspired by progresses in cognitive science, artificial intelligence, computer vision, and mobile computing technologies, we propose and implement a wearable virtual usher for cognitive indoor navigation based on egocentric visual perception. A novel computational framework of cognitive wayfinding in an indoor environment is proposed, which contains a context model, a route model, and a process model. A hierarchical structure is proposed to represent the cognitive context knowledge of indoor scenes. Given a start position and a destination, a Bayesian network model is proposed to represent the navigation route derived from the context model. A novel dynamic Bayesian network (DBN) model is proposed to accommodate the dynamic process of navigation based on real-time first-person-view visual input, which involves multiple asynchronous temporal dependencies. To adapt to large variations in travel time through trip segments, we propose an online adaptation algorithm for the DBN model, leading to a self-adaptive DBN. A prototype system is built and tested for technical performance and user experience. The quantitative evaluation shows that our method achieves over 13% improvement in accuracy as compared to baseline approaches based on hidden Markov model. In the user study, our system guides the participants to their destinations, emulating a human usher in multiple aspects.
... Chen and Kotz (2000) clarified context-aware means that applications are enabled to anticipate users' needs and to act in advance. For instance, a location-aware mobile navigation system automatically adapts to the changing location of the mobile user and specifies route directions about that location without further user interaction (Freksa et al., 2007). ...
Full-text available
Communication between people conveniently uses qualitative spatial terms, as shown by the high frequency of vague spatial prepositions such as ‘near’ in natural language corpora. The automatic interpretation of these terms, however, suffers from the challenges of capturing the conversational context in order to interpret such prepositions. This research presents an experimental approach to solicit impressions of near to identify distance measures that best approximate it (nuanced by the type of referent, and contrast sets). The presented model computes topological distances to sets of possible answers allowing a ranking of what is near in a context-aware manner. Context is introduced through contrast sets. The research compares the performance of topological distance, network distance, Euclidean distance, Manhattan distance, number of intersections, number of turns, and cumulative direction change. The aim of this comparison is to test whether a metric distance or topological distance is closer to human cognition, challenging the well-known paradigm of ’topology first, metric second’. The comparison results from our experiments show that topological distance appears to be closer to human perception of nearness than other distance measures only at larger scales, while a metric distance (Euclidean distance or Manhattan distance) is closer to how people perceive nearness in smaller scales. People with different sense-of-direction show no obvious differences in their inclination of the seven distance measures with regard to nearness. This research caters for the interpretation of ‘near’ with granular and local context, and provides a cognitively inspired method to answer near-queries automatically. The findings apply to urban environments and may need further verification in less structured environments.
... We distinguish three classes of context variables in interpretation of place description: the environment, the human who generates a place description, and the place description as a linguistic phrase. This classification adapts a characterisation of context in performing map-based tasks by Freksa et al. (2007) to a text-based task. Our linguistic context implied by the place description corresponds to their map context, yet we separate objective environmental factors from all cognitive factors. ...
... Brezillon (1999) worked on the context definition for more than 25 years and he concluded that since context does not have a clear definition, it is better to design a knowledge model that takes into account contextual information and helps in making decisions. A similar conclusion was drawn by Freksa et al. (2007) who proposed an approach to model context from a cognitive perspective. We use Freksa's model to define the relationship between contexts and spatiotemporal patterns. ...
... Spatial awareness is defined as the relationship between and synthesis of information garnered from the spatial environment (in this case, a field site), a cognitive agent (in this case, a robot), and a cartographic map (in this case, maps that the support system provides) 29 . In real-time remote robotic science operations, supporting both individual and team awareness of spatial information is critical, because it is directly related to questions of space, place, and distance. ...
After several decades of rather sporadic use in the scientific literature, the concept of hidden geographies is still usually based on provisional definitions that support the specific geographical hiddenness of the topic presented in a publication. This chapter focuses on hidden geographies, with the aim of providing a usable, not necessarily definitive understanding and definition of the concept. After a conceptual-semantic view at hidden geographies, the meanings of the concept and the term are presented, based on the analysis of literature, which provides a colourful variety of connotations and names of the concept in practise. In the discussion, some of the contexts underlying the concept under study are highlighted, as well as questions regarding its understanding and use, such as understanding the blurred line between hidden and revealed geography, and the roles of geography and geoinformatics in revealing or hiding geographies. Finally, a general definition and some specific definitions are proposed, linked to four layers of understanding of the concept: undiscovered, uncognised, unpublished and deliberately hidden geographies.
Unlike the large body of research on investigating interactions among cities using survey data, the social media-based city interaction study has received much less exploration. Based on geographical studies of social media content in China, we develop a few indices quantifying various levels of geographical awareness among cities. (1) We find that the geographical awareness proxy by the social media-based indices can measure interactions among cities. Specifically, the geographical awareness among cities follows gravitational law and is highly correlated with mobility flows. (2) The spatial in-awareness index (SIAI) is an appropriate index indicating a city's ranking in the urban hierarchy (3) the spatial out-awareness rate (SOAR) can indicate the interactions from a focal city to other cities. Our findings also show that SOAR can predict the number of people infected during a pandemic in a city system. Once the origin city or hotspots of the outbreak and the number of infected persons within those cities are known, we can use the social media-based SOAR index to predict number of cases for other else cities in the urban system. With this information, governments can properly and efficiently deliver medical equipment and staff to cities where large populations are infected.
This chapter concludes the book. It briefly summarizes what we have discussed in the previous six chapters and then looks ahead. In particular, we contemplate what it takes for a geospatial system to be intelligent , and what we still miss at the moment in order to build such systems. Overall, we believe that we have provided an appreciation and better understanding of both the challenges and potential of landmarks in intelligent geospatial systems.
Full-text available
Summary Current mobile devices like mobile phones or personal digital assistants have become more and more powerful; they already offer features that only few users are able to exploit to their whole extent. With a number of upcoming mobile multimedia applications, ease of use becomes one of the most important aspects. One way to improve usability is to make devices aware of the user's context, allowing them to adapt to the user instead of forcing the user to adapt to the device. Our work is taking this approach one step further by not only reacting to the current context, but also predicting future context, hence making the devices proactive. Mobile devices are generally suited well for this task because they are typically close to the user even when not actively in use. This allows such devices to monitor the user context and act accordingly, like automatically muting ring or signal tones when the user is in a meeting or selecting audio, video or text communication depending on the user's current occupation. This article presents an architecture that allows mobile devices to continuously recognize current and anticipate future user context. The major challenges are that context recognition and prediction should be embedded in mobile devices with limited resources, that learning and adaptation should happen on-line without explicit training phases and that user intervention should be kept to a minimum with non-obtrusive user interaction. To accomplish this, the presented architecture consists of four major parts: feature extraction, classification, labeling and prediction. The available sensors provide a multi -dimensional, highly heterogeneous input vector as input to the classification step, realized by data clustering. Labeling associates recognized context classes with meaningful names specified by the user, and prediction allows
Full-text available
The CyberDesk project is aimed at providing a software architecture that dynamically integrates software modules. This integration is driven by a user's context, where context includes the user's physical, social, emotional, and mental (focus-of-attention) environments. While a user's context changes in all settings, it tends to change most frequently in a mobile setting. We have used the CyberDesk system in a desktop setting and are currently using it to build an intelligent home environment.
Location-based services (LBS) are a new concept integrating a user's geographic location with the general notion of services, such as dialing an emergency number from a cell phone or using a navigation system in a car. Incorporating both mobile communication and spatial data, these applications represent a novel challenge both conceptually and technically. The purpose of this book is to describe, in an accessible fashion, the various concepts underlying mobile location-based services. These range from general application-related ideas to technical aspects. Each chapter starts with a high level of abstraction and drills down to the technical details. Contributors examine each application from all necessary perspectives, namely, requirements, services, data, and scalability. An illustrative example begins early in the book and runs throughout, serving as a reference.
1.0 What this is all about 2.0 Cross-modal transfer of frame of reference: evidence from Tenejapan Tzeltal 2.1 Tzeltal absolute linguistic frame of reference 2.2. Use of absolute frame of reference in non-verbal tasks