STRUCTURING WAYFINDING TASKS WITH IMAGE SCHEMATA

Transcription

1 STRUCTURING WAYFINDING TASKS WITH IMAGE SCHEMATA By Martin M. Rauba A THESIS Submitted in Partia Fufiment of the Requirements for the Degree of Master of Science (in Spatia Information Science and Engineering) The Graduate Schoo University of Maine December, 1997 Advisory Committee: Max J. Egenhofer, Associate Professor of Spatia Information Science and Engineering, Advisor M. Kate Beard-Tisdae, Associate Professor of Spatia Information Science and Engineering Werner Kuhn, Professor of Geoinformatics and Digita Cartography, University of Muenster, Germany Andrew U. Frank, Professor of Geoinformation, Technica University Vienna, Austria 1

2 "! #! By Martin M. Rauba Thesis Advisor: Dr. Max J. Egenhofer An Abstract of the Thesis Presented in Partia Fufiment of the Requirements for the Degree of Master of Science (in Spatia Information Science and Engineering) December, 1997 Wayfinding is a basic activity that peope do throughout their entire ives as they navigate from one pace to another. Many theories of spatia cognition have been deveoped to account for this behavior. But most of the computationa modes focus on knowedge representation (e.g., cognitive maps) and do not consider the process of structuring wayfinding tasks and space. This thesis presents a methodoogy based on image schemata to structure peope s wayfinding tasks. Image schemata are recurring menta patterns (e.g., the CONTAINER or PATH schema) that peope use to understand a spatia situation. They are highy structured and grounded in peope s experience. The area of our attention is airport space which is used as a case study. Many airports are bady designed and passengers are often unfamiiar with the particuars of the situations. We compare two seected airports in regard to the ease of performing a common wayfinding task. In order to do so, the methodoogy of structuring space with image schemata is combined with a proposed wayfinding mode. We show that sequences of image schemata are sufficient to describe wayfinding tasks in spatia environments at an abstract eve. Therefore, they can be used to compare the compexity of wayfinding tasks for different airports. 2

3 The integration of image schemata into the design process of spatia environments such as airports (i.e., the impementation of our method in a computer system) wi hep to identify architectura probems with regard to wayfinding prior to construction. Our structuring methodoogy can be generaized and wi, thereby, contribute to the design of future geographic information systems that are supposed to integrate eements of human spatia understanding. 3

4 TABLE OF CONTENTS Acknowedgments ii List of Figures vii List of Tabes viii Chapter 1: Introduction1 1.1 Wayfinding in Airports Background of the Thesis Goa and Hypothesis Research Design Genera Methodoogy Deveopment Appication of Methodoogy Reevance of the Work Organization of Thesis 7 Chapter 2: Common-Sense Knowedge of Geographic Space and Human Wayfinding Common-Sense Geographic Knowedge and Naive Geography Common-Sense Knowedge and GIS Quaitative Reasoning Naive Geography Human Wayfinding Spatia Knowedge and Cognition Cognitive Maps Human Wayfinding Performance Computer Modes for Wayfinding 20 4

5 Chapter 3: Image Schemata What are Image Schemata? Exampe Image Schemata Reevance of Image Schemata for Spatia Appications Image Schemata Reated to Common-Sense Geographic Knowedge and Human Wayfinding 31 Chapter 4: A Methodoogy to Structure Wayfinding Tasks with Image Schemata Task Description Interviews Extraction of Image Schemata from Interviews Image Schemata Orientationa Image Schemata Image Schemata within Different Contexts Structuring Wayfinding Tasks with Image Schemata Transation to Forma Representation Semi-Forma Representation Forma Representation with Gofer Forma Representation with Proog 53 Chapter 5: Appication of the Methodoogy to Wayfinding in Airports Interview Procedure Airport Spaces Simuation of Task Procedure and Subjects Vienna Internationa Airport Task Description Interviews Extraction of Image Schemata from Interviews Frankfurt Internationa Airport Task Description Interviews Extraction of Image Schemata from Interviews Wayfinding Mode 69 5

6 5.4.1 Choices Cues Comparing Wayfinding at Vienna Internationa Airport versus Frankfurt Internationa Airport Method Anaysis for Vienna Internationa Airport Anaysis for Frankfurt Internationa Airport Proof of Hypothesis and Resuts 78 Chapter 6: Summary, Concusions, and Future Work Summary Concusions Future Work 81 Bibiography 85 Appendix Biography 107 6

7 ACKNOWLEDGMENTS I am very gratefu for having had the opportunity to work with Dr. Max Egenhofer, my thesis advisor. His guidance, advice, ideas, and encouragement made it a peasure for me to do this work. I woud aso ike to thank the other members of my advisory committee, Dr. Werner Kuhn, Dr. Andrew Frank, and Dr. Kate Beard-Tisdae for their comments and suggestions for improvement. I want to thank Dr. Scott Freundschuh, Dr. Stephen Hirte, Dr. David Mark, and Dr. Dan Monteo for commenting on my work and providing me with numerous references. Thanks aso to Juia Finkernage and Heinz Horatschek for showing me around Frankfurt and Vienna Internationa Airports. My friends and coeagues at the Department of Spatia Information Science and Engineering have made my stay a wonderfu experience, both scientificay and personay. Specia thanks go out to Thomas Bittner, i capo Andreas Baser, Dr. Doug Feweing, Katheen Hornsby, and Andrea Rodríguez. Thomas Windhoz has been a great friend and roommate to me. He aways heped me when I needed something, cheered me up with country music, and made a terrific chess coach. Thanks aso to Eric Po and Juio Fernández for taking the time to be my chauffeurs. It woud have been ong waks without them. Thanks to Wateroo & Robinson, Wofgang Ambros, R.E.M., and a the other great musicians who heped to keep me going during the days and nights of working on this thesis. Last but not east I want to thank my famiy, especiay my parents for their ove and support during a of my ife. It has aways been comforting to know that they are 7

8 there for me. Thanks aso to Gwen for becoming a part of my ife and for correcting my Engish. This work has been partiay supported by the NCGIA under NSF grants SBR and SBR Additiona funding was provided by the Austrian government. I am gratefu for this support. To the memory of my grandfathers, Dr. Gustav Hubert and Edmund Rauba, who I am sure woud be very proud of me. 8

9 LIST OF FIGURES Figure 3.1 Large-scae containment 27 Figure 3.2 Sma-scae containment 27 Figure 3.3 The PATH schema in the physica 29 Figure 4.1 The four stages of the methodoogy to structure wayfinding tasks and space with image schemata 35 Figure 5.1 Moving through passport contro at Vienna Internationa Airport (sides 5, 6, 7, and 8) 60 Figure 5.2 Part of the departure ha at Frankfurt Internationa Airport 67 Figure 5.3 Proposed wayfinding mode 70 9

10 LIST OF TABLES Tabe 2.1 Reation between research methodoogies in Naive Geography and categories of human wayfinding research 13 Tabe 3.1 Seective ist of image schemata (Johnson 1987 p.126) 26 Tabe 4.1 Exampe of transcript and image-schematic representation 51 Tabe 5.1 Resuts of a questionnaire distributed amongst 25 frequent fyers. Subjects were asked at what airports they had most difficuties in wayfinding 56 Tabe 5.2 Subjects tested 58 Tabe 5.3 Task and subtasks at Vienna Internationa Airport 59 Tabe 5.4 Transcript and image-schematic representation of side 5 (passport contro) 62 Tabe 5.5 Transcript and image-schematic representation of side 6 (duty-free area after passport contro) 63 Tabe 5.6 Transcript and image-schematic representation of side 7 (duty-free area after passport contro) 64 Tabe 5.7 Transcript and image-schematic representation of side 8 (duty-free area after passport contro) 65 Tabe 5.8 Task and subtasks at Frankfurt Internationa Airport 66 Tabe 5.9 Transcript and image-schematic representation of side 5 (departure ha) 68 Tabe 5.10 Rues that hep to infer cues from the image-schematic representation 72 Tabe 5.11 Paths and cues for Vienna Internationa Airport (viewpoints within probem areas are highighted, dp = decision points) 74 Tabe 5.12 Paths and cues for Frankfurt Internationa Airport (viewpoints within probem areas are highighted, dp = decision points) 77 10

11 1. INTRODUCTION The Earth is not a brute fact to be taken as given, but aways inserted between Man and the Earth is an interpretation, a structure and a perspective on the word, an enightenment which reveas the rea within the rea, a point of departure from which understanding deveops. Eric Darde (1952): L Homme et a Terre Peope do wayfinding throughout their entire ives. They navigate from pace to pace, using common-sense knowedge. Such knowedge is mediated by structures and categories of understanding peope s daiy experiences in the space they ive (Johnson 1987). Wayfinding is a natura ski that peope earn as sma chidren (Piaget and Inheder 1967) and deveop as they grow up. It takes pace in many different situations, such as driving across a country, waking in a city, or moving through a buiding (Guck 1991). In a of these situations peope have one thing in common: they use commonsense knowedge of geographic space. Within the ast years research on human wayfinding has mainy deat with the exporation of cognitive representations and has neither focused on the processes of wayfinding (e.g., the information needs) themseves (Guck 1991) nor on the design of spatia environments. In other words, most of the work has focused on what Norman (1988) cas knowedge in the head (i.e., interna knowedge) instead of knowedge in the word (i.e., externa knowedge). But as Norman argues, peope do not need to have 11

12 a knowedge in the head in order to behave precisey. Knowedge can be distributed party in the head, party in the word, and party in the constraints of the word (Norman 1988 p.54). Norman further states that much of the information peope need to perform a task is in the word and that the human mind is perfecty taiored to make sense of this word. Piaget and Inheder (1967) have ong since argued that spatia behavior and spatia representations are very different. They distinguished between practica space (i.e., acting in space) and conceptua space (i.e., representing space). In order to buid rea-word spaces that are easy to navigate it is necessary to find out about how peope immediatey understand spatia situations, i.e., how they make sense of practica space whie performing a wayfinding task. Our work focuses on properties of environments (i.e., knowedge in the word ) as perceived and cognized by peope and, therefore, deas primariy with the exporation of practica space as defined by Piaget and Inheder. It is important to investigate peope s perceptua and cognitive structures in order to be abe to mode them in future spatia information and design systems. These systems can then be used to simuate rea-word appications, such as wayfinding tasks, in a cognitivey pausibe way, because they integrate human spatia concepts. The importance of human spatia cognition in the area of geographic information systems (GIS) is aso indicated by various research agendas (Mark et a. 1997, UCGIS 1996). 1.1 WAYFINDING IN AIRPORTS This thesis deas with wayfinding in airports a specia case of moving through a buiding. Passengers at an airport have to find their way from check-in to their gate, from their gate to the baggage caim, and between gates. They are often in a hurry and must avoid getting ost. This can be a difficut task, because many airports are poory designed, have poor signage, and are densey crowded. Aso, many passengers are unfamiiar with the particuar space and fast motion, which puts them in stressfu situations. In an emergency case things become even worse. One coud see the consequences just recenty when a fire accident happened at the Düssedorf Airport in Germany. $&%(' )+*-,.,/' ) A(2J8 7Q5I0<7Q'&8%1RS*7QTUC%LVXW,<=?5I*0<5I: *K@.8 *@G8%(' V`8%LaZDL79*@.W 12

13 b )+*KC8 7Q5I0<7Q'MTUC%+CAL*TU*%L'&B0<%+2J8 )+*K2J8 79*KCI'YC5I*I>-8' 79*B'Y8 0<%+E]2J0<7 5ICEFEF*%+AL*79Ed>-CEd5I: )+8 Ed5ICEFEF*%+AL*79Ed)LDL7Q798 79*B'Y: V`8%L'Y0?' )+*IH.DL7Q%+8%+A(C7Q798_+C:9)+C: : W.g ' >-CEd)+8 Ah)+: V`@.8 2J2J8 BD+:'&2J0<7+' )+*K2J8 79*K@.*5IC7Q' *K8%L'Y0?' )+*KEF*BDL79*K@.*5IC7Q' DL79*I)+C: : W 3i)LD+AL*IHI0.@<V`0.24TU*%12 790<Tj' )+*K2J8 79*K@.*5IC7Q' TU*%L'M' 798 *@?'Y0?79*EFBD+*I5ICEFEF*%+AL*79E]0<DL'&0.24' )+*KB: 0.EF*@G@.C%+AL*7 k 0<%+*IDL%L'Y8 :M: C'Y*I%+8 Ah)L'YW (Standard 1996) Bad architectura design (e.g., ow passageways) was aso bamed for the deady outcome of this catastrophe. Had the airport been constructed with the intention of offering navigationa knowedge (e.g., finding the nearest emergency exit in this case) through its design aone, then it woud have been easier for passengers to find their way out of the buiding. 1.2 BACKGROUND OF THE THESIS In order to make wayfinding easier for passengers at an airport it is necessary to design airport space in such a way that it faciitates peope s structuring processes of tasks and space. Buidings can ony be designed in a user-friendy way if one takes into account how peope understand and structure space. Athough most of the iterature has been focusing on knowedge representation, some research has aso been done on the process of structuring space itsef, e.g., the architect Christopher Aexander deveoped a pattern anguage consisting of 253 patterns. These patterns are based on the experientia nature of things and hep peope to structure their environment (Aexander et a. 1977). Johnson (1987) proposed that peope use so-caed image schemata to understand the word in which they ive. Image schemata are recurring menta patterns that hep peope to structure space in order to know what to do with it. These patterns are highy structured themseves and grounded in peope s experience. Image schemata fit into the category of so-caed aternative conceptuaizations or cognitive modes of space modes that are buit upon peope s experiences with their environment. It has been argued that such conceptuaizations have to be integrated into future GIS in order to match peope s thinking more cosey and, therefore, faciitate peope s interaction with these systems (Mark 1989). The iterature offers many different 13

14 cognitive categorizations of space (Freundschuh and Egenhofer 1997). Couceis and Gae (1986) proposed a forma framework based on agebraic structures of groups to distinguish six kinds of spaces: (1) pure Eucidean space, (2) physica space, (3) sensorimotor space, (4) perceptua space, (5) cognitive space, and (6) symboic space. The gap between perceptua space, i.e., objects are apprehended through the senses at one pace and one time, and cognitive space, i.e., sensory images of objects are inked to eements of cognition, such as beiefs and knowedge, might just be a definitiona one because there seems to be a strong connection between the two. As Lee (1973) pointed out, percepts are not free of concepts, and concepts are not free of percepts. In order to ink perceptua and cognitive space some bridges need to be buit. A theoretica framework that forms one possibe bridge was estabished by Tamy (1996). He adopted the notion of ception which incudes the processing of sensory stimuation, menta imagery, and continuousy experienced thought and affect. Image schemata coud be part of such a framework because they are cognitive concepts that aso occur in the perceptua domain. Peope sense these patterns visuay, as we as they think about them in an abstract way. 1.3 GOAL AND HYPOTHESIS The goa of this thesis is the deveopment of a methodoogy to structure wayfinding tasks and space with eements of human perception and cognition. This methodoogy is then used in combination with a proposed wayfinding mode to measure the compexity of a particuar space in regard to a certain wayfinding task. The hypothesis is twofod: First, representing wayfinding tasks at airports through image schemata is an appropriate method to determine the critica eements (i.e., the choices and cues) of a wayfinding mode. Second, these eements account for the compexity of the wayfinding tasks as rated by traveers. 1.4 RESEARCH DESIGN The fied of our research is human wayfinding in genera and human spatia concepts (i.e., perceptua and cognitive structures of space) in particuar. We do not investigate 14

15 representationa aspects, such as cognitive maps, but focus on properties of spatia environments as immediatey perceived and cognized by peope. Another important subfied that is of concern in this thesis is human spatia reasoning 1. Based on previous work in the areas of psychoogy, cognitive science, artificia inteigence, urban panning, architecture, and geography, we investigate the fied of human wayfinding. We ook at the performance iterature (i.e., empirica studies on how peope find their ways) as we as the competence iterature (i.e., cognitive wayfinding modes). The atter incudes simuations of spatia cognitive processes using computationa modes. In addition we dea with research on common-sense knowedge and reasoning, and Naive Geography (Egenhofer and Mark 1995). These areas are cosey reated to the methodoogy we deveop ater on. The research in this thesis is divided into two major parts: GENERAL METHODOLOGY DEVELOPMENT The first part of our research focuses on the deveopment of a genera methodoogy to structure wayfinding tasks and space with eements of human perception and cognition. We introduce the term image schema as the fundamenta eement of our methodoogy and expain why it is of importance for human common-sense reasoning and wayfinding. Based on image schemata we estabish the genera methodoogy that consists of four sequentia stages: (1) a task sequence is formuated; (2) during interviews peope describe their spatia experiences whie performing a wayfinding task in the appication space; (3) these interviews are anayzed and image schemata extracted; and (4) the extracted image schemata are used to structure the wayfinding task APPLICATION OF METHODOLOGY In the second research part we appy the genera methodoogy to wayfinding in airports in order to demonstrate its usefuness and appicabiity. We use the four steps of the methodoogy in combination with a proposed wayfinding mode to compare the 1 Frank (1992) defines spatia reasoning as any reasoning process that reates to objects in space and makes use of their ocation, position, shape, etc. 15

16 compexity of two different airports with regard to peope performing a common wayfinding task. The two airports (i.e., Vienna Internationa Airport in Austria and Frankfurt Internationa Airport in Germany) were seected based on the resuts of an informa questionnaire where we asked frequent fyers at what airports it was easiest or most difficut for them to find the pace they wanted to go. The resuts prove our hypothesis because they show that sequences of image schemata are sufficient to describe wayfinding tasks in airports at an abstract eve and can be used to compare the compexity of wayfinding tasks for different airports. 1.5 RELEVANCE OF THE WORK Our methodoogy wi hep to design airport space in such a way that it faciitates wayfinding, because passengers can mainy rey on common-sense rather than expert geographic knowedge. The expected benefits are increased passenger satisfaction passengers wi save time when doing certain tasks, such as finding the right gate, emergency exit, or duty free store; airport safety in an emergency case peope wi find the emergency exits much faster; and airine profitabiity airines wi save money which they currenty oose due to passengers and, therefore, airpanes being ate. An eventua impementation of the methodoogy wi ead to spatia information and design systems that can be used to test airport space or other pubic buidings in the design phase for compexity of particuar wayfinding tasks peope have to perform. The structuring methodoogy wi aso highight reevant concepts that are to be part of a comprehensive theory of Naive Geography (Egenhofer and Mark 1995). It wi, therefore, contribute to the design of future GIS that are supposed to support commonsense reasoning. 1.6 ORGANIZATION OF THESIS Chapter 2 of this thesis reviews the iterature on the two areas of common-sense knowedge and human wayfinding. In particuar, it addresses empirica studies of how 16

17 peope find their ways in various arge-scae spaces such as urban environments, subway systems, and arge buidings, and computationa wayfinding modes. Chapter 3 introduces image schemata as the main component of our methodoogy. We expain their meaning, show exampes, and describe their importance for peope s structuring of wayfinding tasks. Finay, we show how image schemata reate to common-sense geographic knowedge and human wayfinding. In Chapter 4 we present a methodoogy to structure wayfinding tasks and space according to these eements of peope s perception and cognition. We describe the four stages of the methodoogy (1) task description, (2) interviewing, (3) extracting image schemata from the interviews, and (4) structuring wayfinding tasks and space with the extracted image schemata. In Chapter 5 the methodoogy is combined with a proposed wayfinding mode to compare the compexity of a common wayfinding task in two different airports. A summary, concusions, and directions for future work are presented in Chapter 6 of the thesis. 17

18 2. COMMON-SENSE KNOWLEDGE OF GEOGRAPHIC SPACE AND HUMAN WAYFINDING Common sense is not so common. Votaire Finding one s way in an airport reies on a variety of eements. Peope have to make intuitive and quick decisions whie at the same time they must avoid getting ost. In this chapter we first review the state-of-the-art of two areas that dea with such topics, i.e., research on common-sense knowedge and human wayfinding. We then ook at empirica studies of how peope find their ways in different arge-scae spaces and review computationa wayfinding modes. 2.1 COMMON-SENSE GEOGRAPHIC KNOWLEDGE AND NAIVE GEOGRAPHY COMMON-SENSE KNOWLEDGE AND GIS Starting with peope s first experiences with the environment they are estabishing knowedge about the word in which they ive. Peope need such basic knowedge for their everyday activities, such as waking, eating, shopping, and earning, and ca it common-sense knowedge. It comprises many different domains that have compex interactions. Understanding a situation often invoves concepts of quantity, time, space, physics, pans, goas, needs, and communication (Davis 1990). In this thesis we focus on the domain of space, independent of any cutura and individua differences. Kuipers (1978) defines common-sense (geographic) knowedge as foows: Common-sense knowedge of space is knowedge about the physica environment that is acquired and used, generay without concentrated effort, 18

19 to find and foow routes from one pace to another, and to store and use the reative position of paces. The current generation of GIS supports common-sense knowedge of geographic space ony insufficienty. Cacuations in these systems are based on Cartesian coordinate space (i.e., pane Eucidean geometry) and the standard concepts of space are not aways appropriate and force the user to transform tasks into often-inappropriate form (Frank 1992). Peope have to dea with incompete information and are abe to fi in the missing gaps due to common-sense knowedge. It wi be important for future GIS to incude common-sense knowedge and reasoning concepts that peope actuay use, such as rues based on common sense, hierarchica schemata, and intuition. As pointed out by Egenhofer and Mark (1995), today s GIS ack modes that integrate different kinds of spatia concepts in a cognitivey sound and pausibe way. Lifschitz (1995) finds a pace for the theory of common-sense reasoning in the history of ogic: It provides an axiomatic basis for reasoning about the word inhabited by agents ike us by agents who have beiefs and goas, who perform actions in order to reach these goas and, by doing so, change the state of the word. He aso highights the use of defauts, one of the main features of common-sense reasoning. Defauts are vaues or propositions that are supposed to be true uness there is information that says otherwise QUALITATIVE REASONING Instead of doing exact cacuations, peope most often appy quaitative methods of spatia reasoning (Frank 1996, Cohn 1995, Frank 1992, Freksa 1992) that rey on magnitudes and reative, instead of absoute, vaues. When peope perceive space through different channes they arrive at various kinds of information that are usuay quaitative in nature. Peope rarey move through the environment using ruers or tape measures. When visuay viewing a scene the resut is a retina image that is of quantitative nature, but the knowedge peope retrieve from this image is quaitative (Freksa 1991). Freksa argues that such knowedge is exacty what peope need for the process of spatia reasoning and mentions three advantages: (1) expressive power of 19

20 quaitative constraints based on their interaction (e.g., concept of transitivity), (2) independence from specific vaues and scae, and (3) invariance under transformations. As an exampe he introduces the aquarium metaphor where observers can ocate fish by quaitative means, athough they have to dea with incompete, imprecise, and subjective knowedge. An important feature that is used during this process is caed conceptua neighborhood of reations: if reative positions of objects change graduay, the change between neighboring spatia reations is just stepwise (Freksa 1992, Egenhofer and A- Taha 1992). Peope usuay use topoogica instead of metrica information. Topoogica properties of objects stay invariant under such transformations as transations, rotations, or scaings. By using abstract geometrica anaysis Piaget and Inheder (1967) demonstrated that fundamenta spatia concepts are topoogica, but not Eucidean at a. They showed that chidren start to conceptuaize space by buiding up and using eementary topoogica reationships, such as proximity, separation, order, and encosure NAIVE GEOGRAPHY Naive Geography is a current fied of study that deas with common-sense geographic words. It estabishes the ink between knowedge that peope have about their surrounding geographic space and the deveopment of forma modes that integrate such knowedge. Egenhofer and Mark (1995) define Naive Geography as the study of the body of knowedge that peope have about the surrounding geographic word. This definition is based on Hayes s (1985) definition of Naive Physics, a fied that investigates peope s knowedge of the everyday physica word, such as peope s intuitive ideas about faing rocks or evaporating iquids. Peope use concepts of Naive Geography for spatia reasoning in their everyday ives; therefore, research in this area wi hep us to understand how peope think and how they find their way in the geographic word. This information is essentia in the design process of particuar spaces such as airport space. It has to be the foremost goa of the designer to create microwords in which peope can easiy move around. Passengers in an airport, for exampe, shoud be abe to find their ways without a big effort, reying excusivey on common-sense knowedge. In the case of an emergency situation peope 20

21 must make intuitive judgments, because they do not have the time to interpret compicated emergency signs. Egenhofer and Mark suggest two different research methodoogies as part of the framework for deveoping Naive Geography. These are directy reated to the two categories of human wayfinding research (Tabe 2.1). Naive Geography The deveopment of formaisms of naive geographic modes for particuar tasks. The testing and anayzing of forma modes. Human Wayfinding Simuations of spatia cognitive processes using computationa modes (Section 2.3.2). Empirica resuts of how peope find their ways (Section 2.3.1). Tabe 2.1: Reation between research methodoogies in Naive Geography and categories of human wayfinding research. 2.2 HUMAN WAYFINDING Human wayfinding research investigates the processes that take pace when peope orient themseves and navigate through space. Theories try to expain how peope find their ways in the physica word, what they need to find them, how they communicate directions, and how peope s verba and visua abiities infuence wayfinding. Lynch (1960 p.3) defines wayfinding as based on a consistent use and organization of definite sensory cues from the externa environment. Wayfinding takes pace in many different situations in which peope find themseves, such as driving across a country, waking in a city, or moving through a buiding (Guck 1991). The utimate goa of human wayfinding is to find the way from one pace to another. The space in which human wayfinding usuay 2 takes pace is caed arge-scae space (Kuipers 1978). Because 2 Human wayfinding can aso take pace in virtua spaces, such as virtua geographies for the Word Wide Web (Dieberger and Boter 1995) or maps. This thesis focuses on wayfinding in arge-scae space. 21

22 objects are arger than peope and can, therefore, not be moved, peope have to navigate through arge-scae space in order to earn about it. Exampes for arge-scae spaces are andscapes, cities, and houses. Compementary to arge-scae space is sma-scae space whose objects are smaer than peope (e.g., things on a desktop). Sma-scae space objects are the domain of Eucidean geometry and they are usuay manipuabe (Itteson 1973, Downs and Stea 1977) SPATIAL KNOWLEDGE AND COGNITION Peope need to have spatia knowedge and various cognitive abiities in order to succeed in wayfinding (e.g., reading a map or foowing a path). Human spatia knowedge of geographic space is assumed to consist of three eves: (1) andmark knowedge comprises saient points of reference in the environment, (2) route knowedge puts andmarks into a sequence (e.g., navigation paths), and (3) survey or configurationa knowedge aows peope to ocate andmarks and routes within a genera frame of reference (i.e., incorporating Eucidean measurements) (Siege and White 1975). The cognitive abiities depend on the task at hand. Finding one s way in a street network (Timpf et a. 1992, Car 1996) uses a different set of cognitive abiities than navigating from one room to another in a house. Peope are usuay good in appying their individua skis to the task at hand: if their spatia skis are weak, they use verba skis to navigate, and vice versa (Vanetti and Aen 1988) COGNITIVE MAPS To successfuy perform wayfinding, peope need cues within their environment (i.e., knowedge in the word) or representations of spatia knowedge about their environment. One usefu metaphor suggests that peope have a cognitive map in their heads (Kuipers 1982) a menta representation that corresponds to peope s perceptions of the rea word athough other metaphors, such as cognitive coage and spatia menta mode (Tversky 1993), or cognitive atas (Hirte 1997) have aso been proposed. Despite the fact that these representations are caed spatia, it is important to notice that our memory has to integrate spatia information with non-spatia information (Gäring et a. 1984). Considering the process of acquiring spatia knowedge of an environment, the cognitive map deveops from a menta andmark map to a menta route map and shoud 22

23 eventuay resut in a menta survey map. The ast stage is cosest to a cartographic map, though it sti contains inaccuracies and distortions 3. Davis (1990) points out two main differences between cognitive and cartographic maps: a cognitive map may consist of different knowedge structures and it has to integrate incompete, imprecise, and subjective knowedge. This is usuay not the case for cartographic representations that represent information ony pictoriay and are reasonaby accurate and compete. Peope construct and deveop their cognitive maps based on the recording of information through perception, natura anguage, and inferences. Compex environmenta structures can ead to sower deveopment of cognitive maps and aso to representationa inaccuracies. Considering the structure of cognitive maps, there exist two casses of theories: hierarchica and non-hierarchica. Recent studies suggest that a hierarchica mode is more appropriate for cognitive maps than a non-hierarchica (Hirte and Heidorn 1993). It incudes distinct patterns of encoding spatia information at oca (e.g., Eucidean knowedge) and goba eves (e.g., topoogica knowedge). Hierarchies can either be based on expicit evidence, such as physica boundaries, or impicit evidence, such as semantic and functiona custers (McNamara et a. 1989). Hierarchica structures are even imposed in spaces without any inherent structure. One consequence of hierarchies in cognitive maps is that they may have an infuence on wayfinding performance (i.e., bias in spatia judgments such as distance estimates) (Hirte and Jonides 1985). This effect was aso confirmed in a study by Goedge et a. (1985) where a hierarchica representation of route eements was found to account for different errors depending on the choice point and the compexity of the segments. Researchers from various discipines have thoroughy investigated the roe cognitive maps pay in spatia behavior, spatia probem soving, acquisition, and earning (Kitchin 1994). Much ess, however, has been found out about how peope immediatey understand different spatia situations whie performing a wayfinding task. Guck (1991) points out this ack of information by arguing that previous work on wayfinding concentrated on the description of the cognitive map and negected affective and ogistica concerns in most of the cases. As an aternative approach Guck suggests 3 According to Lynch (1960), these errors in cognitive maps are most often metrica and rarey topoogica. 23

24 to expore the information needs. He further envisions a typoogy of wayfinding scenarios and proposes the use of the sense-making investigation method. Thus sense-making is a creative human process of understanding the word at a particuar point in time and space imited by our physioogica capacities, our present, past and future (Guck 1991 p.129). Such understanding can be seen as a snapshot of common-sense knowedge of space and time and is, therefore, important for the process of common-sense spatia reasoning. The idea behind the sense-making method is to ook at the wayfinding process itsef instead of ooking at the representation (i.e., the cognitive map). 2.3 HUMAN WAYFINDING PERFORMANCE Human wayfinding research can be divided into two categories (Guck 1991): performance and competence (Section 2.4). The iterature on performance discusses empirica resuts of how peope find their ways. Investigations are based on coecting individua s perceptions of distances, anges, or ocations. An exampe for a typica experiment is the pairwise judgment of distance between points. Such experiments hep in describing features of the cognitive map. Kevin Lynch s (1960) The Image of the City is the first documentation of human wayfinding research in the iterature. In a study he asked peope of three US cities (i.e., Boston, Los Angees, and Jersey City) how they viewed their city what they iked and disiked. His goa was to deveop a method for the evauation of city form based on the concept of imageabiity 4, and to offer principes for city design. As part of the interviews peope had to perform menta trips across their cities, describing the sequence of things and andmarks they woud see aong the way. Eventuay they were asked whether it was easy to find their ways in the city or not. Based on his investigations Lynch divided the contents (i.e., the physica forms) of the city images into five casses: (1) paths, (2) edges (i.e., boundaries), (3) districts (i.e., regions), (4) nodes, and (5) 4 Imageabiity: that quaity in a physica object which gives it a high probabiity of evoking a strong image in any given observer. (Lynch 1960 p.9) 24

25 andmarks. These eements were described as the buiding bocks in the process of making firm, differentiated structures at the urban scae and have been the basis for ater wayfinding research. It has been estabished that peope earn about their environment incrementay (eves of human spatia knowedge, Section 2.2.1). First, they derive knowedge about andmarks. The fact that judgments concerning andmarks are faster, indicates that peope tend to give them a specia status. During the next step andmarks are connected through routes. Routes have a directiona basis and are subsequenty integrated into a whoe network. Organization of routes in such a network is primariy topoogica. At the fina stage of the earning process peope arrive at knowedge of efficient inks between ocations (i.e., survey knowedge). Such knowedge comprises information about distances and orientation. Gäring et a. (1983) assume that wayfinding is not possibe uness orientation is maintained (e.g., the orientation of you-are-here-maps significanty infuences the abiity of peope to successfuy compete wayfinding tasks). They aso state that distance estimates have a directiona basis and generay tend to be ess accurate than direction estimates. The accumuation of survey knowedge as the fina stage of a spatia earning process is not undisputed. In a study about the deveopment of cognitive mapping abiities of student nurses in a hospita, Moeser (1988) found that menta representations of survey maps do not deveop automaticay in a geographic spaces. The author bamed this fact on the compexity and bad design (e.g., each foor designed differenty, no main corridors, no rectanguar rooms) of the buiding and further suggested that architects in addition to functionaity of a buiding shoud aso consider peope s imitations in deveoping menta representations of their surroundings. Weisman (1981) identified four casses of environmenta variabes that infuence wayfinding performance within buit environments: (1) visua access, (2) the degree of architectura differentiation, (3) the use of signs and room numbers to provide identification or directiona information, and (4) pan configuration. His resuts were confirmed by other researchers. In Gäring et a. s (1983) study of orientation in a arge university department visua access was regarded as an important factor, because wayfinding performance of subjects with restricted sight improved ess over time. The impact of orientation toos ike foor pans was aso investigated. The performance of 25

26 subjects with restricted sight using foor pans improved as fast as that of subjects with no restricted sight, foor pans, therefore, counteracting the negative effect. In another study Gäring et a. (1986) proposed to cassify the environment by examining the degree of differentiation, the degree of visua access, and the compexity of spatia ayout. The infuence of foor pan compexity on both cognitive mapping and wayfinding performance, and the existence of an interaction between foor pan compexity and the quaity of signage was demonstrated in two studies by O Nei (1991a, 1991b). His resuts showed that an increase in foor pan compexity eads to a decrease in wayfinding performance. The presence of signage was an important factor but coud not compensate for foor pan compexity. A difference between the use of textua signage and graphica signage was aso found: textua signage produced greater accuracy (i.e., ess errors) whereas graphica signage enhanced the rate of trave. Seide s (1982) study at the Daas/Fort Worth Airport confirmed that the spatia structure of the physica environment has a strong infuence on peope s wayfinding behavior. For passengers arriving at the gate with direct visua access to the baggage caim wayfinding was easier. In addition to Weisman s four casses of environmenta variabes, peope s famiiarity with the environment aso has a big impact on wayfinding performance: frequency of prior use had a big faciitating effect in university buidings (Gäring et a. 1983) as we as in airports (Seide 1982). Research on peope s wayfinding performance has been particuary hepfu for estabishing practica guideines on how to design pubic buidings in order to faciitate wayfinding. Architects seem to have come to the concusion that faciitating peope s wayfinding needs more than putting up signs, because most of the time signage cannot overcome architectura faiures (Arthur and Passini 1992). Therefore, wayfinding principes have to be considered during the design process both for the overa spatia structure and for the formgiving features. Some guideines (Arthur and Passini 1992, 1990) despite focusing on the design and pacement of signage highy stress the importance of architectura features. In Evauation and Design Guide to Wayfinding, Arthur and Passini (1990 p.a-1) introduce the term environmenta communication (i.e., transfer of orientation, wayfinding (direction), and other information within the buit environment, by means of signs and other communications 26

27 devices or architectura features to enabe peope to reach destinations ), arguing that the buit environment and its parts shoud function as a communication device. Such communication shoud begin at the outside of a buiding: the outside form is very instructive to the user, because it usuay gives an impression of the buiding s interna spatia organization. Arthur and Passini mention two major aspects regarding the understanding of buidings: (1) a spatia aspect that refers to the tota dimensions of the buiding (e.g., was encose space and eements such as an interior atrium break it up) and (2) a sequentia one that considers a buiding in terms of its destination routes. Destination routes shoud eventuay ead to so-caed destination zones. These are groupings of simiar destinations within buidings into ceary identifiabe zones (Arthur and Passini 1992). In order to faciitate wayfinding to such destination zones the circuation system 5 shoud be of a form peope can easiy understand. It is further suggested that fewer decision points on any route and redundancy in wayfinding information are aso faciitating effects. Based on an investigation about the nature of orientationa probems users faced whie traveing on the New York City subway, Bronzaft et a. (1976) suggested severa improvements to be made in future design: consistent coding of information, appying structura detais in a systematic and coherent fashion, presenting structure and operations of the system in different forms (i.e., to aow the user getting detaied information when needing it, but aso seeing the entire system), and integrating coors as a type of coding for informationa aids. 2.4 COMPUTER MODELS FOR WAYFINDING In addition to empirica studies of performance (Section 2.3), cognitive wayfinding modes have been investigated in what is referred to as competence iterature. It incudes simuations of spatia cognitive processes using computationa modes. Hirte and Heidorn (1993) emphasize the importance of distinguishing between computationa modes of human cognitive processes and computationa systems that perform the same task without paying much attention to human aspects. Cognitivey based computer 5 Circuation system: the overa horizonta and vertica pedestrian paths of a setting; circuation systems can be organized on a inear, centra, composite, or network basis (Arthur and Passini 1992 p.223) 27

28 modes generay simuate a wayfinder that can sove route-panning tasks with the hep of a cognitive-map-ike representation. This map consists of earned andmarks and paths, and buids the foundation for navigation. The focus of these modes is to find out how spatia knowedge is stored and used, and what cognitive processes operate upon it. The TOUR mode is considered the starting point for a computationa theory of wayfinding (Kuipers 1978). It is a mode of spatia knowedge whose spatia concepts are based mainy on observations by Lynch (1960) and Piaget and Inheder (1967). With the TOUR mode Kuipers simuates earning and probem soving whie traveing in a arge-scae urban environment. Besides deaing with states of partia knowedge his main focus of attention is the cognitive map which he defines as the physicay unobservabe structure of information that represents spatia knowedge. Kuipers takes three different metaphors (i.e., map in the head, map ike a network, and map ike a cataog of routes ) for the cognitive map and combines them into one common framework. Knowedge in this cognitive map is divided into five categories: (1) routes (i.e., sequences of actions), (2) topoogica street network, (3) reative position of two paces (i.e., vector within coordinate frame), (4) dividing boundaries, and (5) containing regions. This knowedge is represented through environmenta descriptions, current positions, and inference rues that manipuate them (i.e., if a certain set of conditions is true, then the rues trigger some action). Different kinds of knowedge are stored in these representations and new information is assimiated. Because TOUR copes with incompete spatia knowedge of the environment, it earns about it by assimiation of observations into the given structure. To describe orientation in the eary TOUR mode, eight headings at 45 degree intervas are used. But Kuipers points out that peope do not actuay use such numerica vaues when orienting themseves in the environment. A subsequent appication to this computationa mode of the human cognitive map utiizes an approach to robot earning based on a hierarchy of types (i.e., sensorimotor interaction, procedura behaviors, topoogica mapping, and metric mapping) of knowedge of the robot s senses, actions, and spatia environment (Kuipers and Levitt 1988). With this semantic hierarchy approach a computationa mode is buit that expresses the compexity and moduar structure of a nontrivia domain of human knowedge. It supports the position that a compex body of knowedge is not acquired by 28

29 a singe representation and earning agorithm but by a highy structured mechanism consisting of severa distinct and interacting representations and earning agorithms. Severa other cognitivey based computer modes (e.g., TRAVELLER (Leiser and Zibershatz 1989), SPAM (McDermott and Davis 1984), ELMER (McCaa et a. 1982)) have been deveoped to simuate earning and probem soving in spatia networks. A mode of spatia earning that integrates concepts from both cognitive psychoogy and artificia inteigence was created by Gopa et a. (1989): NAVIGATOR represents basic components of human information processing 6, such as fitering and seecting (i.e., important andmarks), and forgetting. In the mode, two views of a suburban environment an objective and a subjective (i.e., cognitive) one are compemented by cognitive processes reating to spatia earning and navigation. The cognitive map is modeed through a hierarchica network consisting of nodes, inks, subnodes, and subinks 7. The goa of this computer mode was to investigate how the process of extracting and using environmenta information is conducted by the architecture of human information processing. The focus of these computer modes ies primariy in the creation and exporation of the cognitive map: how it is structured, what transitions occur during the earning process, etc. However, they do not te the whoe story of how peope find their ways. Goedge (1992) argues that most of the computer modes do not simuate the behavior of human wayfinders, because they fai to integrate asymmetric distances and directions 8. Furthermore, these modes do not incude individua wayfinding criteria, such as minimizing trave time and/or distance, minimizing effort and/or stress, or minimizing the chance of getting ost by taking onger but more famiiar routes. Peope aso earn differenty: instead of exporing a spatia network sequentiay they choose sectors based on even a sma piece of information (e.g., hypothesizing that the 6 The information processing framework is a dominant paradigm in cognitive psychoogy. Its goa is to anayze the structures and menta processes invoved in the performance of a cognitive task. 7 This approach is aso caed neuroogicay based information processing. 8 It has been shown that peope perceive distances differenty depending on whether they are to or from a andmark. 29

30 destination is to the north). With their own approach Goedge et a. (1985) tried to overcome some of the imitations of other computationa wayfinding modes. Their mode of route earning is based on four issues: (1) acquisition and representation based on episodic experience and subsequent generaization, (2) different types of knowedge and forms of representation, (3) systematic inaccuracies and distortions in the cognitive representation, and (4) behaviora errors associated with inaccurate and hierarchicay organized knowedge. The conceptua mode itsef consists of the set of actions of the individua, the set of structures encoding knowedge about the task environment, the set of cognitive processes (i.e., perceiving, storing, retrieving, and reorganizing environmenta knowedge) operating on the knowedge structures, and the set of contro processes determining the interaction of the decision-maker with the environment. One resut of the empirica anaysis conducted to test the mode showed that knowedge about saient wayfinding points aong a route consists of four types of nodes. The origin and destination nodes estabish the task environment. Cue and feature knowedge was found to be highest there. On the next hierarchica eve there were second and third order nodes identifying key choice points for actions (e.g., a direction change). Next, there were ower order nodes needed to carify the ocation of choice points (e.g., signaing a direction change). Finay, there were misceaneous cues that coud aso be nonpermanent features but had sufficient impact to guarantee reca for a specific task. Athough with the incusion of behaviora and representationa errors this mode seems to be an improvement over previous ones, Goedge (1992) ater argued that more research on human understanding and use of space has to be done. He aso mentioned the possibiity of spatia knowedge not being we described by existing theories or modes of earning and understanding. 30