Statement of Interest

1. My interests with respect to "Cognitive Models of Dynamic Phenomena and Their Representations"

Representation of spatial knowledge and the human capacity to communicate this knowledge to other persons has been my main research interest during the last 15 years. In my work, which is interdisciplinary integrating methods and results from linguistics, artificial intelligence, logics and psychology, there are two mayor lines of research topics: Firstly, hybrid-systems, in which propositional and depictorial, i.e. analog, representations interact and are used by hybrid-processes, and secondly, the way humans use language to describe space and the spatial aspects of their environment. Currently, I do research with two projects funded by the German Science Foundation, which is directed to the topics described above. (1) In ajoint project with Carola Eschenbach, who also sends a proposal toparticipate, we develop formal characterizations of spatial concepts. (2) In a second project I investigate the conceptualization processes, which are part of the production of natural language descriptions of events (descriptions of dynamic maps). Both areas of research have directed my interest to dynamic phenomena of spatial representations. On the one hand, there is evidence from psychology that analog representations are dynamic, and that the dynamics of these representations leads to specific effects in processing, especially in reasoning. On the other hand, the majority of work in language production has been done on verbal descriptions of static constellations; on the basis of this research it is possible to investigate producing descriptions of dynamic domains.

2. My contribution to the meeting:

Formal models of dynamic mental representations (hybrid architecture integrating propositional and depictorial representations); Empirical results from conceptualization experiments in the framework of language production tasks. I am interested that these contributions lead to new cooperations, for example, with respect to communication via pictorial representations (using techniques of visualization) or to reasoning with dynamic representations.[My participation at the Varenius meeting on "Scale and Detail in the Cognition of Geographic Information" has been successful in starting some fruitful discussions and cooperations.]

1. Formats of representation: Pictorial and propositional representations

During the last decades representations which serve as analogical counterparts to entities and configurations they model have become to the focus of interest in Artificial Intelligence and Cognitive Science. On the empirical side, research on imagery and mental rotation has given overwhelming evidence for analog representations of spatial con-figurations, i.e. representations with intrinsic spatial properties(cf. Kosslyn 1980, 1994; Palmer 1978). Such non-propositional representations can be induced by perceptual or linguistic inputs and are used successfully in reasoning tasks. From the knowledge representation and processing perspective of AI, processes of diagrammatical reasoning, which extend the inferential performance of reasoning systems, have demonstrated their importance for the development of systems with a wide spectrum of tasks, e.g., multi-media communication systems, visualization systems in software design or molecularscene analysis. Both Cognitive Science and Artificial Intelligence approaches to diagrammatic representations deal with internal as well as with external representations: on the one hand, there are investigations on spatial mental representations, e.g. mental images (Kosslyn, 1994),mental maps and knowledge about the environment (Montello, 1992; Hirtle & Heidorn 1993) or spatial mental models (Johnson-Laird, 1983) and on knowledge representations with spatial properties, e.g. the computational imagery approach of Glasgow and Papadias (1992), on the other hand, there is research on external spatial and diagrammatic means for processing knowledge, especially on geographic maps, (cf. Bertin, 1981; Monmonier, 1996). The benefits of analog representations - whether they are internal or external, whether they are called pictorial, spatial or diagrammatic - are based on their property of having the same inherent constraints as the domain they model.

2. Representing dynamic phenomenaanalogously

Research on analog representations as mentioned above has a strong bias to the spatial domain. Although processes using analog representations, for example "rotation of mental images" or "zooming and scanning of mental images", have been in the centre of AI and Cognitive Science since nearly 20 years, investigations on analog representations of change and dynamics are only in their infancy (but, a major exception is the "dynamic mental representation" paradigm by Freyd and colleagues, see, e.g., Freyd 1987). External pictorial or analog representations, as maps, pictures, or diagrams, are static. To use them for representing change, movement or other dynamic phenomena some methods have been developed, for example, annotation of entities in maps by temporal information (e.g., dates of historical events or natural disasters), usage of different types of arrows in maps in diagrams, or the technique of representing the development of states in the real world by sequences of representations, e.g. a series of maps or diagrams, in the minimal case, a before-after pair of diagrams. In the technical context of multimedia and the internet it is possible to present information about spatial domains, about geographic space in particular, not only statically but also dynamically. Dynamic visualization will be especially relevant to represent knowledge about change and processes. To understand the processes and representations involved in the use of dynamic visualizations by a user, it is necessary to distinguish at least - the following levels of representations and representational entities (Following Palmer (1978), I assign the represented world to a distinguished representational level):

2.1 Analog dynamic representationsand the temporal structure of representational levels

Whether a dynamic representation functions adequately as an analog representation depends on the agreement of properties and constraints between the temporal structures of the two levels of representation in question, i.e. of level (1) to level (2) mappings, or of level (2) to level (3) mappings (on analog representations, see Palmer 1978). Some aspects of temporal structure highly relevant for constituting adequate analog representations are:

• Topological structure of the temporal domain: discrete, dense or continuous (see, e.g., Habel 1994): Whereas continuous representations seem to be necessary for an adequate dynamic representation of processes and events, e.g. motion, there are also changes, for which only the initial state and the final state are relevant. In these cases, discrete dynamic representations seems to be more appropriate.
• Scale and granularity constraints in mapping between representational levels: The mapping between the temporal structures of two levels of representation has to consider temporal scale. This can be performed by a linear mapping, or one with non-linear properties. Especially, the change from normal speed to slow motion or time-lapse motion, has non-linear transitions.
• Direction of time: Whereas the forward direction of time is the standard way to present a course of events or a development of states of affairs, backward presentation can be very helpful in explaining causal histories (going from the current state of a system to its precessors).

2.2 Layers orrepresentations

A person perceiving and comprehending a dynamic map builds up internal representations (conceptualizations) which refer to both, graphical entities and real world entities. Since the latter correspond to the things a person who designs a dynamic map intends to represent we call them intended entities. Thus, the internal representations of the comprehender of a dynamic map can be divided into four representational layers (see Tappe & Habel 1998):

1. the layer of graphical objects(e.g. lines, squares, rectangles, ...),
2. the layer of intended objects(e.g. a campus, the xy-street, a subway station...),
3. the layer of visualizationevents (e.g. a line appears, a square is highlighted, ...) and
4. the layer of intended events(e.g. driving on a street, crossing a mountain range, ...).

3. Conceptualization of Dynamic Sketch Maps - An empirical case study

To investigate the processes of conceptualization during comprehension of dynamic maps we performed experiments in a language production task (Habel 1997, Tappe & Habel 1998). As stimuli we used sketch maps of the route from the computer science department to the main campus of the University of Hamburg. The sketch maps were drawn on a electronic A3 sketch-pad and the drawing process was documented in a computer program that was especially designed for this purpose. In the experiments, the subjects were presented a replay of the drawing events on a computer screen. They were instructed to watch carefully what happened and to describe it. The verbalizers saw pixels appear in a sequential manner—one after the other—on the previously blank computer screen; the graphical objects became visible in the same chronological order that they were previously produced in. The speech data were recorded with a digital recorder; the transcripts were proofread by a second person. Our analyses of the dynamic map presentations are twofold: Firstly, we formalize the graphical entities, i.e. objects and events, in the framework of a ‘sketch grammar’. Such as grammar reflects part of a person’s competence that allows him or her to draw sketch maps and to interpret maps drawn by others. Secondly, we analyze elicited verbal descriptions with formal linguistic and psycholinguistic methods. In particular, the analysis of the language production process leads to insights in the conceptualization processes performed during comprehension of the dynamic map, especially the sub-processes of segmentation, grouping, structuring, and linearization. The experimental design allows to investigate the comprehension of different types of dynamic maps as well as a variety of factors which have an effect on the conceptualization process. Especially, it is possible to make systematic variations of the some of these parameters, for example: (1) In an on-line condition the speakers are instructed to start speaking as soon as the graphical objects started to appear and to describe simultaneously what happens on the screen. In an off-line condition the subjects first perceive the whole genesis of the respective sketch, then the computer screen goes completely blank and afterwards they describe what they had seen. (2) Variation of speed of presentation and difference between continuous presentation vs. presentation of discrete, contentful chunks of information. (3) Presentation of isolated dynamic maps and those presented in interaction with further information:(i) information given before the dynamic map presentation started vs. simultaneously given information, (ii.) visually vs. auditory presented information.

References to papers by the author:

Habel, Ch. (1994). Discreteness, Finiteness, and the Structure of Topological Spaces. In C. Eschenbach, Ch. Habel & B. Smith (Eds.): Topological Foundations of Cognitive Science. Papers from the Workshop at the FISI-CS, Buffalo, NY. July 9--10, 1994. Hamburg: Graduiertenkolleg Kognitionswissenschaft Hamburg. Report 37,81-90.

Habel, Ch. (1997). Discours etreprésentations spatiales dans la description de plans. In M Denis (ed.), Langage et cognition spatiale. (pp. 103-126). Paris: Masson.

Tappe, H. & Ch. Habel (1998):Verbalization of Dynamic Sketch Maps: Layers of Representation and their Interaction. [extended version of one page abstract / poster at Cognitive Science Conference; Madision WI, August, 1.-4., 1998.] http://www.informatik.uni-hamburg.de/WSV/sprachproduktion/CogSci98.ps

Christopher Habel – Curriculum Vitae

Scientific Education and Training

• 1985 Habilitation: "Artificial Intelligence and Cognitive Linguistics" (Dept. of Computer Science, Technical University of Berlin),
• 1979 Doctor of Philosophy: Dept. of Linguistics, University of Osnabrück
• 1968-75 Studying Mathematics and Theoretical Linguistics, Diploma (M.Sc.) in Mathematics, University of Göttingen

• since 1990 Founding Director of the Doctoral Program for Cognitive Science at the University of Hamburg (funded by the German Science Foundation)
• since 1986 Full Professor (C4) in Computer Science (Artificial Intelligence) and Adjunct Professor of Linguistics at the University of Hamburg

• since 1997 "Processes of conceptualization in language production: A computational model based on the empirical investigation of event descriptions" , DFG research initiative "Language Production".
• since 1996 "Formal specification of spatial concepts and structures with definitions and axiomatic characterizations"(Joint project with Dr. Carola Eschenbach), DFG research initiative "Spatial Cognition".
• since 1994 "The function of non-linguistic representations - especially spatial and visual models - in text comprehension."(Joint project with Prof. Dr. Stephanie Kelter)
• 1986 - 91 "Spatial expressions of German", Subproject of the IBM funded research initiative LILOG (Linguistic and logic foundations of knowledge processing)

• 1995 Cofounder of the DFG research initiative "Spatial Cognition" (jointly with W. Brauer, München; Ch. Freksa, Hamburg; K. Wender, Trier)

Habel, Ch. (1990). Propositional and depictorial representations of spatial knowledge: The case of path concepts. In R. Studer (ed.): Natural language and logic. (pp. 94-117).Lecture Notes in Artificial Intelligence. Berlin: Springer.

Habel, Ch. (1997). Discours etreprésentations spatiales dans la description de plans. In M Denis (ed.), Langage et cognition spatiale. (pp. 103-126). Paris: Masson.

Tappe, H. & Ch. Habel (1998):Verbalization of Dynamic Sketch Maps: Layers of Representation and their Interaction. [extended version of one page abstract / poster at Cognitive Science Conference; Madision WI, August, 1.-4., 1998.] http://www.informatik.uni-hamburg.de/WSV/sprachproduktion/CogSc98.ps

Habel, Christopher (1998). Piktorielle Repräsentationen als unterbestimmte räumliche Modelle. (Pictorial representations as underdetermined spatial models)Kognitionswissenschaft,7(2) [in German].

Eschenbach, C., Ch. Habel, L. Kulik(submitted to publication). Representing simple trajectories as oriented curves.