AAAI 2011 Fall Symposium Descriptions
Advances in Cognitive Systems
Over the past 20 years, artificial intelligence has made substantial progress on many fronts, but only at the cost of fragmenting into subfields that focus on some facets to the exclusion of others. However, there remains a need for research that pursues the discipline’s original goal of developing intelligent systems that demonstrate the full range of human cognitive abilities. Many AI researchers remain committed in this goal, but they have no common place to gather and report their results.
Over the past 20 years, artificial intelligence has made substantial progress on many fronts, but only at the cost of fragmenting into subfields that focus on some facets to the exclusion of others. However, there remains a need for research that pursues the discipline’s original goal of developing intelligent systems that demonstrate the full range of human cognitive abilities. Many AI researchers remain committed in this goal, but they have no common place to gather and report their results.
The AAAI Fall Symposium on Advances in Cognitive Systems will provide such a venue. The meeting will bring together researchers with interests in human-level intelligence, complex cognition, integrated intelligent systems, cognitive architectures, commonsense reasoning, and related topics. The event harks a return to the initial goals of artificial intelligence and cognitive science, which aimed to explain intelligent behavior in computational terms and to reproduce the entire range of human cognition in computational artifacts.
Topics covered at the meeting will include the representation and organization of complex mental structures, their use in multistep reasoning and problem solving, and their acquisition from experience and instruction. Talks, posters, and discussions will address computational approaches to conceptual inference and reasoning, memory storage and retrieval, high-level execution and control, problem solving and heuristic search, language processing, social cognition and interaction, structural learning and knowledge capture, and metacognition. The common denominator is that research on these issues contributes toward creating and understanding complete intelligent systems.
The entire AI community stands to benefit from a meeting that draws attention to the field’s original aspirations, but we especially encourage participation by researchers who have been trained only in one of the more focused subareas. The symposium hopes to expose a new generation of AI researchers to the challenges that we must address to build cognitive systems with humanlike abilities.
Organizing Committee
Paul Bello (Office of Naval Research), Nick Cassimatis (Rensselaer Polytechnic Institute), Kenneth Forbus (Northwestern University), John Laird (University of Michigan, Ann Arbor), Pat Langley (Arizona State University), Sergei Nirenburg (University of Maryland, Baltimore County)
For More Information
For more information, please see the supplementary symposium website.
Building Representations of Common Ground with Intelligent Agents
Much of the success of natural language interaction is caused by the participants’ mutual understanding of the circumstances surrounding the communication. These circumstances range from reminiscing about a shared experience, such as a birthday party, to coordinating firefighting efforts amongst a team using joint beliefs about mutual capabilities. This mutual understanding of perceived context is termed common ground, and is made up of all of the background and shared information that will lead to the eventual success of the communication. Some measure of common ground is used in most, if not all, successful interactions between human actors. For humans to have a convincing and beneficial experience interacting with intelligent agents, the agents must have mechanisms that support the fundamentals of common ground. Otherwise, the consequence may be unsuccessful and incomplete interactions.
This symposium aims to bring together researchers from robotics, artificial intelligence, human-computer interaction, computational linguistics, and cognitive modeling to share their diverse perspectives on common ground and its component factors. This investigation of how existing work explicitly or implicitly uses aspects of common ground will bring us closer to a practical approach for a complete common ground in intelligent, computational agents. Work to be presented may consist of currently implemented applications that contain common ground components, theoretical formalisms of these components, proposed frameworks, and work in progress.
Topics
Components of common ground include, but are not limited to, background knowledge, world knowledge, presuppositions, mutual knowledge, mutual beliefs, beliefs about beliefs, and conversational maxims.
Examples of how these components may have been incorporated into current work include building of shared histories between actors; representations of beliefs or beliefs about beliefs; identifying and representing suppositions and presuppositions; reasoning about other actors’ beliefs; creating untruthful or deceptive common ground; formalizations of any common ground component; agents or models that use common ground components in interaction; discourse or dialogue models which use some aspect of common ground; extraction of perlocutionary actions or effects from utterances; inferring common ground; and using common ground to ground referents.
Organizing Committee
Sam Blisard (Naval Research Laboratory), Will Bridewell (Stanford Center for Biomedical Informatics Research), Wende Frost (Naval Research Laboratory), Arthi Murugesan (Naval Research Laboratory), Candace Sidner (Worcester Polytechnic Institute), Alan Wagner (Georgia Tech Research Institute)
For More Information
For more information, see the supplementary symposium web site.
Complex Adaptive Systems: Energy, Information, and Intelligence
Complex adaptive systems (CAS) and related technologies have proven to be a powerful framework for understanding system-level phenomena across the physical, natural, and social sciences. We characterize a general CAS model as having a significant number of autonomous agents that use one or more levels of feedback; exhibit emergent properties and self-organization; or produce nonlinear dynamic behavior.
This symposium’s theme addresses fundamental issues for understanding complex phenomena: energy, information, and intelligence. This theme builds upon the previous years’ focus of threshold effects (2009) and resilience, robustness, and evolvability (2010).
Energy in a CAS is often more than merely physical energy; it is anything that drives and constrains the system. Agents must cooperate and/or compete for limited resources, whether these resources are “energy,” “power,” “food,” “money,” or some other system resource. The success or failure of various agent strategies depends on their effectiveness in acquiring and utilizing these resources.
Information represents any form of verbal, nonverbal or even nonhuman communication. Information represents what the agents know or learn about their local environment. Papers relating to the theme of information may, for example, cover signal patterns and effects, signal processing, or information storage (memory). Flows of information itself may also be the focus of CAS research, such as in models of political dissent, social contagion, or the dynamical flows across networks.
Intelligence encompasses how agents react to any information that they acquire from the environment, as well as the system-level properties that emerge from these actions and reactions through patterns of correlated feedbacks. Thus, intelligence may refer to the agents themselves or to the system as a whole. Such intelligence can exist at almost any level of complexity, from simple examples of swarm intelligence to complex human cognition.
Format
Our symposium will have invited talks from leaders in the field, as well as paper presentations on both completed and speculative work. Due to the nature and the novelty of the theme, it is essential to allow ample time for both open-ended and targeted discussions; as such, we will hold panel discussions, roundtable talks, and a poster session, to allow for a spirited interaction among participants.
Organizing and Program Committee
Mirsad Hadzikadic, chair (University of North Carolina, Charlotte), Ted Carmichael, cohair (University of North Carolina, Charlotte), Mark Altaweel (University of Chicago), Tony Beavers (University of Evansville), Aaron Bramson (University of Toronto), Matthieu Branlat (Ohio State), Patrick Grim (SUNY Stony Brook), Liz Johnson (University of North Carolina, Charlotte), Ardeshir Kianercy (University of Southern California), Kiran Lakkaraju (Sandia National Laboratory), Megan Olsen (University of Massachusetts, Amherst), Jonathan Ozik (Argonne National Laboratory), Mark Pizzato (University of North Carolina, Charlotte), Bill Rand (University of Maryland), Bob Reynolds (Wayne State University), Molly Rorick (Yale University), John Stamper (Carnegie Mellon University), Forrest Stonedahl (Northwestern University), Paul Youngman (University of North Carolina, Charlotte), Tina Yu (Memorial University of Newfoundland).
For More Information
For more information about this symposium, and for submission guidelines and links, please visit the supplementary website, or e-mail cochair Ted Carmichael at tedsaid@gmail.com.
Multiagent Coordination under Uncertainty
In domains ranging from earth observing sensor webs to collaborating ambulances or fire fighters during disaster rescue or software personal assistants scheduling meetings to “coordinators” assisting in executing military missions or exploration of underwater terrains using autonomous underwater vehicles (AUVs) to handling large scale humanitarian logistics, multiple intelligent agents need to coordinate in the presence of uncertainty to achieve team goals. While there have been many interesting models and frameworks presented for addressing and solving various subaspects of the problem, the algorithms employed for solving the models have had to trade off between richness in representation and scalability. In fact, the key hindrances to achieving scalable and quality guaranteed coordination between agents are as follows:
Uncertainty: This can arise due to probabilistic effects of actions taken by agents, inability of agents (either individually or collectively) to completely observe the world, partial availability or unavailability of communication, partially known or unknown motivations of other agents and so on.
Coordination horizon: The total number of decision epochs for which agents need to coordinate affects the solution methodology. Depending on the number of decision epochs left and the history of agent decisions, the coordination decision at the current decision epoch can change. Such dependence on history makes the coordination computationally expensive and one of the key reasons for reduced scalability.
Objective: Due to the lack of certainty about the world, the goal can be one of: (1) Maximize expected value of all the agents involved; (2) Maximize the probability of achieving the goal; (3) Minimize or prevent certain bad states of the world; and many others.
Researchers from various communities, such as artificial intelligence, operations research, game theory and logistics (supply chain) communities, have been actively pursuing the various aspects (mentioned above) of the problem. The goal of this symposium is to bring together these researchers to discover and/or better understand the commonalities in the hope of making fundamental breakthroughs in the field.
Organizing Committee
Pradeep Varakantham (Singapore Management University), Janusz Marecki (IBM T. J. Watson), William Yeoh (University of Massachusetts, Amherst), Paul Scerri (Carnegie Mellon University)
For More Information
For more information, see the supplementary website.
Open Government Knowledge: AI Opportunities and Challenges
The 2011 AAAI Fall Symposium on Open Government Knowledge: AI Opportunities and Challenges (OGK2011) is focused on issues of publishing public government data as reusable knowledge on the web.
Websites like data.gov, research.gov and USASpending.gov aim to improve government transparency, increase accountability, and encourage public participation by publishing public government data online. Although this data has been used for some intriguing applications, it is difficult for citizens to understand and use. This symposium will explore how AI technologies such as the semantic web, information extraction, statistical analysis and machine learning can be used to make the knowledge embedded in the data more explicit, accessible and reusable. The symposium’s location of Washington, DC will facilitate the participation of U.S. federal government agency members and enable interchange between researchers and practitioners. We also expect attendance of international open government data players from countries such as the United Kingdom and Australia.
Relevant topics include the automatic and semiautomatic creation of linked data resources, ontologies for government data, entity linking and coreference detection between linked data resources, adding temporal qualifications to government data, creating mashups with open government data, linked open government data analysis, metadata for provenance, certainty and trust, policies for information sharing, privacy and use, social networks and government data, machine learning applied to government data, data visualization techniques, and applications.
This symposium will include a mixture of invited talks, paper presentations, panels, system demonstrations, a poster session, and discussions. We plan to have several invited speakers drawn from government, academia and industry. We will run panels on the emerging challenges and best practices, including how to enhance transparency and interoperability within an agency and across different agencies/countries, and how to promote a nationwide health information network that effectively integrates government-curated public records and citizens’ personal health data.
Organizing Committee
Li Ding (RPI), Tim Finin (UMBC), Lalana Kagal (MIT), and Deborah McGuinness (RPI). Program committee members and additional information are listed on the supplementary symposium site.
For More Information
For more information about the symposium, send email inquiries to ogk11-info@googlegroups.com or visit the supplementary website.
Question Generation
Asking questions is a fundamental cognitive process that underlies higher-level cognitive abilities such as comprehension and reasoning. Research on question generation (QG) has a long history in artificial intelligence, psychology, education, and natural language processing, and has far-reaching applications in intelligent technologies, such as dialogue systems, question answering systems, web search, intelligent tutoring systems, automated assessment systems, inquiry-based environments, adaptive intelligent agents and gamebased learning environments.
Interest in question generation has been steadily growing over the past few years. The 2011 Symposium on Question Generation follows three previous workshops (held in 2008, 2009, and 2010), which resulted in the first Question Generation Shared Task and Evaluation Campaign. Detailed information on these previous events are available at the supplementary website.
This symposium aims to foster theoretical and applied research on computational and cognitive aspects of question generation, featuring involvement of participants from diverse disciplines including, but not limited to, natural language processing, artificial intelligence, linguistics, psychology, and education. The symposium will feature thematic sessions, panel discussions, break-out sessions, and interactive poster and demo sessions. There will be two keynote speeches. Symposium participants will engage in multidisciplinary discussions on the following topics:
- Cognitive models of question generation
- Question taxonomies
- Empirical approaches to question generation
- Question generation tasks and subtasks
- Impact of NLP technologies on question generation tasks
- Target concept selection for question generation
- Context-sensitive question type selection or ranking
- Descriptions of implemented systems or components
- Applications of question generation (intelligent tutoring systems, dialogue systems, and others)
- Generation from different inputs — knowledge bases, ontologies, text, queries
- Future of QGSTECs
Our goal for this year’s meeting is to create an environment for an active exchange of ideas among researchers working on the question generation problem from different perspectives, and to move forward as a community in designing focused question generation tasks.
Organizing Committee
Arthur Graesser (University of Memphis, USA), James Lester (North Carolina State University, USA), Jack Mostow (Carnegie Mellon University, USA), Rashmi Prasad (University of Pennsylvania, USA), Svetlana Stoyanchev (The Open University, UK)
For More Information
For more information, see the supplementary website.
Robot-Human Teamwork in Dynamic Adverse Environment
The goal of this symposium is to discuss collaboration between humans and robots in situations which require all to continuously adapt to developing events. Examples are adverse circumstances encountered during emergency responses in various types of search and rescue missions, or security or military deployments. Adaptation needs to reflect changes both in what is to be done by human and robot (the tasks), and in how they are supposed to do so (the roles). It needs to explicitly couple the task-work with the social aspects of team-work for the human-robot team to maintain cohesive, effective operations. What makes this difficult is that humans are performing under stress. And that robots, so far, are far from being well-tuned to the human factor in human-robot collaboration.
Topics and Intended Audience
The symposium aims to bring together scientists from different research areas with a common interest in the human factor in human-robot and multirobot collaboration in dynamic adverse environments, to answer an emerging issue in cognitive modelling for human-robot teams. Topics that are core to the symposium are the following:
- Cognitive modeling of human and robot workload in stressing situations
- Doctrines and strategies for adaptive multimodal team communication
- Robust decision making for instant or reciprocal urgencies
- Attention and cognitive perception of environments and events
- Joint planning, mixed initiative planning and joint collaboration schedules
- Situation awareness and robot-operator communication
- Performance metrics of joints operations
- Risk and ethical considerations of human-robot collaboration
Format
There are 4 invited speakers, Ron Arkin (Georgia Tech), Jeff Bradshaw (Florida Institute for Human and Machine Cognition), Maarten Sierhuis (PARC’s Knowledge) and Satoshi Tadokoro, director of the International Rescue Systems Institute, who is the leading rescue robots researcher in Japan.
The symposium will combine a variety of activities intended to facilitate interaction among participants from different communities and discussion of key challenges in bridging research in human factors, human-robot interaction, collaborative planning, and teamwork modelling. These activities will include, 13 individual technical presentations, two panels on risks and ethic in teaming, and how could we model human-robot teams in dynamic adverse environments? Two breakout discussion sessions focused on developing a roadmap to facilitate crosscutting research in human-robot teaming, on multimodal interfaces and on how to get prepared for emergencies.
Organizing Committee
Jeff Bradshaw, Alexander Ferrein, Vaclav Hlavac, Gerhard Lakemeier, Geert Jan Kruijff, Leora Morgenstern, Mark Neerincx, Fiora Pirri, Roland Siegwart, Hartmut Surman, Tomoichi Takahashi, Roland Siegwart, Mary-Anne Williams
For more information about the symposium see the supplementary symposium web site.