UNIT 73 - GIS AND SPATIAL COGNITION

• Components of the user interface
• Fundamental questions

• 1. Developmental psychology perspective
• 2. Cognitive and environmental psychological perspective

• Images or propositions?
• Hierarchical or non-hierarchical structures?
• Frames of reference

• Causes of errors in spatial reasoning

• Examples
• Fuzziness

• Design of better user interfaces and query languages
• Design of universal GIS systems
• New database models
• Improved data entry techniques
• Expert Systems

UNIT 73 - GIS AND SPATIAL COGNITION

Compiled with assistance from Suchi Gopal, Boston University

A. INTRODUCTION

• the next two units (73 and 74) examine advanced topics:
  • knowledge based techniques
  • spatial cognition

• both are efforts to deal with the complexity of real GIS applications
  • complexity of real-world problems - the number of goals and issues which have to be dealt with in real problem- solving
  • complexity of the knowledge and rules which can be brought to bear on a problem
  • complexity of the man/machine interaction which ultimately determines the effectiveness of GIS

• interest in both areas is high
  • progress is still largely in the research domain

• GIS are tools for supporting human decision-making

• in applications such as car navigation systems, electronic atlases, GIS are tools to help people acquire spatial information, learn about geography
  • e.g. research is under way on the design of a portable GIS to help visually impaired people navigate in complex spaces
  • the information acquired through a GIS is used in this case to make simple route-finding decisions

• the interface between the GIS and the user is a filter which determines how successfully information can be transferred

• physical design - keyboard, mouse, tablet, color or monochrome, screen resolution, sound, speech recognition
  • how can these be combined to maximize transfer of information?
  • a car navigation system might use either a display screen with a map, or spoken instructions (e.g. turn left), or some combination - which is the most effective mode of communication between the GIS and the driver?

• functionality - what set of operations is allowed?

• control technique - commands, picking from menus, pointing at icons?

• to design effective user interfaces we need to know more about how people learn, reason with spatial information

• issues that need to be addressed are:

  1. spatial learning

  • how is spatial knowledge learned or acquired by people?

  2. form of internal spatial representation
  • what is the nature of people's internal representation of space?
  • how is spatial information stored in the brain?
  • can this help us design ways of representing spatial data in GIS that lead to better user interfaces?

  3. effects on spatial reasoning
  • how does this internal spatial representation affect decision-making and behavior
    • e.g. navigation, search for housing
  • how do people's naive-models of space lead to errors in geographic reasoning?
  • how to design GIS user interfaces to minimize these errors?

  4. natural language
  • how does the language people use to communicate (natural language) affect their ability to deal effectively with spatial information?
  • would GIS interfaces be more effective if they used natural language to describe spatial relations?

  5. relevance to GIS
  • how should the results of research on these fundamental questions be used to improve GIS user interfaces?

• this unit looks at each of these major issues

• how do people learn about space and the objects and routes within it?

• two disciplinary perspectives

• study of the qualitative changes in the cognitive and perceptual development of a child

• most influential theory of spatial learning is the developmental stage theory proposed by Piaget
  • describes the stages in a child's development of spatial skills

• 4 stages
  • sensorimotor stage - from birth to about 2 years - locations of all objects with reference to self
  • preoperational stage - 2 to 7 years of age - simple spatial problems are solved - an understanding of spatial relations between objects and self
  • concrete operational stage - 7 to 11 years of age - properties of Euclidean space are understood - more complex spatial problems are solved - e.g. concept of reversibility - n steps in one direction followed by n steps in the opposite direction returns one to the same place
  • formal operational stage - 11 to adulthood - child masters more abstract spatial problems - self and other objects located in an independent frame of reference

• e.g. child begins to understand simple spatial relations - "in front of", "left of" in stage 2 - abstract coordinate systems such as UTM are not understandable until stage 4

• studies the sequence of development of knowledge about a space by an adult

• many alternatives proposed (see references) - following is a consensus:
  • landmark knowledge - ability to recognize certain features, but no knowledge of their locations or relationships between them
  • procedural knowledge - knowledge of certain routes, and the procedures necessary to navigate from one end to the other
  • topological knowledge - knowledge of how the known routes intersect and form a network - ability to combine parts of known routes into new routes
  • metric knowledge - ability to recall metric relations between locations - distances, angles - this level of knowledge is needed to reason about previously untraveled routes and shortcuts

• how do our minds construct mental images of the world which somehow capture its basic properties and structure?

• three major questions:
  • what form of representation? images or propositions?
  • what types of structures are used in the representation of spatial relations? hierarchical or non-hierarchical?
  • what frames of reference are used?

• images preserve the visual properties of objects and relations between them
  • the "map in the head"

• propositions provide abstract representation of both verbal and visual information
  • e.g. images of street maps, memory of street names, verbal directions

• one form can be generated from the other if we assume the mind is capable of simple processing
  • compare the GIS's ability to compute vector from raster
  • impossible to determine if one form is more accurate than the other as a model of the way people store spatial information

• hierarchical structures represent spatial information in a nested fashion
  • local and global are different levels of a tree
  • compare hierarchical data structures, e.g. quadtrees

• non-hierarchical structures have no clear differentiation of levels

1. egocentric frame moves with the individual - objects are always represented in their relationship to the individual

2. environmental frame uses a local point as reference, moves when the individual moves from one local area to another

3. global frame is constant irrespective of the location of the individual

• internal representation can be identified by the pattern of errors it produces
  • errors in direction, distance, orientation, judgment of spatial relations have been studied

• lack of explicit representation in memory
  • not all information is perceived or remembered

• use of incorrect procedures in storing or retrieving information
  • e.g. errors because of incorrect rotation of information to or from internal alignment
    • suppose the "mental map" has North at the top
    • errors can be made in reasoning about which way to turn when approaching a junction from the North

• natural language used to describe spatial relations may be vague or context-dependent
  • e.g. "is north of" does not indicate how far or how exactly north

• decay of information

• processing constraints

• limits to size of memory

• storage and type of representation
  • e.g. Reno, NV is actually to the west of San Diego, CA, however, because CA is largely west of NV and the mind stores a hierarchical relationship between states and cities, we expect Reno to be east of San Diego

• natural language appears to affect the way we think and reason about space

• basic components of spatial information - objects, relations between objects, motion - are roughly equivalent to nouns, locative expressions and verbs in natural language

• however correspondence is not exact
  • natural language reflects the human view of the world, is more complex than abstract mathematical structures
  • it may be very difficult to represent the complex human view of the world within a digital system

• use of prepositions to convey spatial relations is subject to complex, hidden rules
  • "in", "on", "between", "across", "near" convey complex meanings
  • e.g. we say "the car is near the house" but not "the house is near the car" - why?

  • e.g. "across the lake" suggests different spatial relationship than "along the lake"
  • e.g. in North America we live "in" a city but "on" a street

• the structure of names has hidden meanings
  • e.g. whether the word "lake" occurs first or second in a name is determined to some extent by its size - "Lake Erie" vs. "Trout Lake" - but "Great Bear Lake" is very large

• nouns can be chosen to convey spatial relations
  • e.g. "timber" has no spatial meaning by itself, but "stand of timber" suggests a small area occupied by trees - "forest" suggests a large area of trees

• translation of prepositions from one language to another poses enormous problems
  • a multilingual natural language interface for a GIS would have to deal with these

• the spatial relationships defined by natural language are fuzzy and context-dependent
  • e.g. meaning of "near" an object depends on the size of the object and is imprecise
  • a natural language GIS interface would have to know the range of distances conveyed by "near"

• research in the area of spatial cognition can have several benefits for GIS development, including:

• given the problems of determining the meaning of natural language, are natural language interfaces worth pursuing?
  • yes, because some applications must use natural language, e.g. GIS for the visually impaired
  • yes, because other forms of interface may be impractical, e.g. car navigation aids must not distract the driver's visual attention to the road
  • yes, because some applications require more than one mode of interaction to maximize effectiveness, e.g. voice can be used in digitizing to augment input from cursor

• such systems should be compatible with cognitive models of the way we perceive and structure space
  • thus would avoid costly problem of transferring GIS technology between different countries and languages

• understanding how spatial information is represented internally may provide novel designs for database models

• permit representation is transformed from natural language into GIS database and vice versa

• natural language is the simplest way of collecting information about the world, but difficult to formalize into precise structures in a digital environment

• knowledge of how spatial information is stored and processed will provide fertile input to the design of intelligent expert systems for spatial information

Herskovits, A., 1987. Spatial Prepositions in English. Cambridge University Press. Interesting book on the use and meaning of spatial prepositions.

Kuipers, B., 1978. "Modeling spatial knowledge," Cognitive Science 2:129-53. One of the most influential papers on the classes of spatial knowledge.

Piaget, J. and B. Inhelder, 1967. The Child's Conception of Space. The classic developmental theory.

Talmy, L., 1983. "How language structures space," in H. Pick and L. Acredolo, editors, Spatial Orientation: Theory, Research and Application, Plenum Press, New York. Argues that language affects the ways in which we think about spatial relationships.

EXAM AND DISCUSSION QUESTIONS

1. Summarize the arguments for believing that an understanding of processes of spatial learning and reasoning is essential if we are to design better GISs, particularly better user interfaces.

2. What would be the desirable functions and other characteristics of a portable GIS for the visually impaired?

3. A paper by Openshaw and Mounsey ("Geographic Information Systems and the BBC Domesday Interactive Videodisk," International Journal of Geographical Information Systems 1:173-180, 1987) describes the design of the BBC Domesday Project, a form of electronic atlas using optical disk technology. What features of the conventional atlas does this system implement? In what ways does it go beyond the capabilities of the conventional atlas? How might principles of human spatial learning and reasoning be combined with the capabilities of GIS to significantly improve the usefulness of the atlas concept? (Note: a number of other atlas-like digital products are available and might be used as similar bases for discussion.)

4. A simple way to illustrate the problems of spatial relations in natural language is to take a formal representation of some spatial data - e.g. a small part of a topographic map or a city street map. One person is asked to describe the contents of the map using only natural language to another person, who must then try to reconstruct the map. Both are aware of the rules governing the map's contents, e.g. contour interval. The participants could be asked to summarize the results, including the role of non- verbal communication, e.g. facial expressions and gestures.

Back to Geography 470 Home Page

Back to GIS & Cartography Course Information Home Page