Cognitive Shape Processing

Stanford University, March 22-24, 2010

Program

Please click here for program details and papers

Description

In the recent decades there has been a growing interest in understanding and computationally investigating how spatial information is processed in natural intelligent sys tems. The inter disci plinary field of spatial cognition, besides its basic research related motivation, also aims at improving artificial systems by transferring natural principles to technical systems, e.g. in ro botics, in intelligent instruction systems or in other in telligent interactive systems.

In spatial cognition, numerous aspects of spatial knowledge are investigated, among these spatial reference systems, topological information, route knowledge, knowledge about distances and directions, etc. For all these aspects, specific forms of representation and formalisms for reasoning about them have been devised. Common to most of the formalisms is that they usually deal with spatial knowledge on a high level of abstraction. In some cases the formalisms only consider some knowledge aspects in isolation (e.g., orientation knowledge), and in other cases they only deal with highly simplified spatial objects such as points or basic geometric forms.

In contrast to this abstraction, real-world problems typically deal with diverse types of spatial knowledge at the same time and involve complex objects with meaningful and specific shapes. Understanding mental processing of knowledge about shapes thus seems essential for under standing mental processing of spatial knowledge in real world scenarios. Importantly, addressing shape knowledge also offers the potential of integrating diverse aspects of spatial knowledge processing since all types of spatial knowledge are affected by shape. So, on the one hand, shape is a specific type of spatial knowledge (among others), on the other hand shape processing involves the most challenging aspect of spatial knowledge processing since it cannot be dealt with in an abstract manner and it affects all other forms of spatial knowledge.

With the term Cognitive Shape Processing we refer to all forms of knowledge processing in volving shape information that are related to, inspired by, or derived from principles found in natural cognitive systems. We thus exclude purely technical approaches to shape processing, however we strongly encourage considering cognitive principles as potential solutions for technical approaches. Unfortunately, and contrary to many other visuo-spatial aspects of cognition, cognitive shape processing has not yet received the appropriate level of attention in the scientific community. Considering the potentials of understanding and employing the principles of cognitive shape processing for both basic and applied research, this clearly calls for a joint endeavor in AI and the cognitive sciences to sufficiently address its most fundamental questions.

Goals of the Symposium

The goal of the symposium is to bring together researchers from artificial intelligence and the cognitive sciences to promote the understanding – from a cognitive point of view – of how shape information can be represented, retrieved, (re-)constructed, and integrated with other types of spatial information. We consider this symposium as a kick-off event that is meant to provide the grounds for identifying the most important research questions, for proposing a structuring of the scientific field, and for collecting first perspectives for potential directions of research. The long-term goal is to reach a thorough under standing of how all types of spatial knowledge interact with shape information and of how such interaction actually creates the unique flexibility and in tegrative processes which we can observe in cognitive shape processing. This understanding is meant to directly inform approaches in cognitive modeling and approaches in AI of processing visuo-spatial information.

Sample questions of interest in cognitive shape processing are:

How is shape knowledge represented in and retrieved from long-term mental storage, and how can it be represented in and retrieved from technical storage and knowledge bases?
Is shape knowledge compositional (i.e., constructed from elementary shapes) or are specific shapes uniquely represented?
Is shape knowledge contour-based or area-based? Or neither?
What are useful measures for shape properties (e.g. metrics for shape, shape com plexity, curviness)
How do prototypical (categorial) shapes relate to specific shapes?
How does partial shape matching work, i.e. when only parts of a specific shape are known or visible? What is its representational and procedural basis?
How can varying levels of granularity be modeled in shape representation and processing?
Given that both visual and spatial aspects are involved in spatial knowledge processing, how does shape information interact with these modes?
Is shape information dealt with in 2D, 2½D, 3D, … and how do the dimensions scale up/down?
What is the role of attention-related processes in cognitive shape processing? How does shape knowledge guide (visual) attention?
What is the relation between control processes in visual perception and knowledge about shapes?
How can brain-imaging studies contribute to the understanding of cognitive shape processing? What stories do eye movements and other behavioral data tell?
How do different modes of shape perception interact, e.g. visual and haptic shape per ception?

The symposium will be scheduled to provide extensive discussion time and group inter actions. In addition to a series of regular presentations with significant question-and-answer time, we intend to reserve about one third of the overall time for parallel, small-group, topic-oriented breakout discussions, with subsequent reporting-back to the plenum. Depending on the number of appropriate submissions, we may also arrange for an extended poster session in order to allow for additional discussions among participants.

Submission Information

Please email submissions of 4-6 pages (preferably in AAAI format as PDF) to barkowsky [at] sfbtr8.uni-bremen.de. Submissions can be position statements, work in progress, or completed work. For general information regarding the AAAI Spring Symposium Series please see http://www.aaai.org/Symposia/Spring/spring-symposia.php.

Deadlines:

	Submission of contributions:	October 2, 2009
	Notification of acceptance:	November 6, 2009
	Camera-ready copies of contributions:	January 16, 2010

Chair

Sven Bertel, University of Illinois at Urbana-Champaign, USA, bertel [at] illinois.edu, [web]

Organizing Committee

Thomas Barkowsky, Universität Bremen, Germany, barkowsky [at] informatik.uni-bremen.de

Sven Bertel, University of Illinois at Urbana-Champaign, USA, bertel [at] illinois.edu

Christoph Hölscher, University of Freiburg, Germany, hoelsch [at] cognition.uni-freiburg.de

Thomas F. Shipley, Temple University, USA, tshipley [at] temple.edu

Program Committee

B. Chandrasekaran, Ohio State University

Ellen Yi-Luen Do, Georgia Tech

Ron Ferguson, Atlanta, GA

Kenneth D. Forbus, Northwestern University

Christian Freksa, University of Bremen

Isabel Gauthier, Vanderbilt University

Gabriela Goldschmidt, Technion, Haifa

Mark D. Gross, Carnegie Mellon University

Mary Hegarty, UC Santa Barbara

Stephen C. Hirtle, University of Pittsburgh

Madeleine Keehner, University of Dundee

Philip J. Kellman, UC Los Angeles

Jan J. Koenderink, Universiteit Utrecht

Richard Lowe, Curtin University of Technology

Fred Mast, University of Bern

Ennio Mingolla, Boston University

Luis A. Pineda Cortés, Universidad Nacional Autónoma de México

Kerstin Schill, University of Bremen

Michael Tarr, Brown University

James T. Todd, Ohio State University