Celebrating AAAI’s 25th Anniversary Courtesy, Carnegie Mellon University Archives / The Allen Newell and Herbert CarnegieCourtesy, Mellon University Archives Simon Collections Conference Program Twentieth National Conference on Artificial Intelligence (AAAI-05) Seventeenth Conference on Innovative Applications of Artificial Intelligence (IAAI-05)

July 9 – 13, 2005 Westin Convention Center Pittsburgh Pittsburgh, Pennsylvania

Sponsored by the American Association for Artificial Intelligence

Cosponsored by Colognet, DARPA, Michael Genesereth, Google, IBM Research, Intelligent Information Systems Institute, Cornell University, Intel Corporation, Microsoft Research, National Science Foundation, Naval Research Laboratory, ACM/SIGART, & Yahoo! Research Labs AAAI-05 Intelligent Systems AAAI-05 Outstanding Senior Pro- Contents Demonstrations Cochairs gram Committee Member Award Biplav Srivastava, IBM Research Labs Acknowledgments / 2 This award will be presented by AAAI-05 Belinda Thom, Harvey Mudd College Awards / 2–3 program chairs Manuela Veloso and Sub- Conference at a Glance / 5 AAAI-05 Competition Cochairs barao Kambhampati to Gaurav Suk- Doctoral Consortium / 4 hatme, University of Southern California Exhibition / 16 Sheila Tejada, University of New Orleans Paul E. Rybski, Carnegie Mellon University Game Playing Competition / 15 IAAI-05 Deployed General Information / 20 Game Playing Competition Chair IAAI-05 Program / 8–13 Applications Awards Michael Genesereth, Stanford University Intelligent Systems Demos / 16 IAAI-05 awards will be announced by Invited Talks / 6–7 AAAI-05 Sponsorship Chair IAAI-05 chair Neil Jacobstein and Posters / 14–15 Carla Gomes, Cornell University cochair Bruce Porter. See the schedule Registration / 21 & Exhibition / 18–20 A complete listing of the AAAI-05 and for paper titles. Certificates will be pre- Special Events and Programs / 3–4, 15 IAAI-05 program committee members sented during paper sessions. Special Meetings / 4 appears in the conference proceedings. Sponsoring Organizations / 2 Robert S. Engelmore Memorial Technical Program / 8–13 Award and Lecture Tutorial Forum / 6 The Robert S. Engelmore Award is spon- Workshop Program / 4 Sponsoring Organizations sored by IAAI-05 and AI Magazine. It will be presented by Neil Jacobstein, IAAI-05 AAAI gratefully acknowledges the gener- chair, and David Leake, editor-in-chief, ous contributions of the following orga- AI Magazine. This award and lecture was nizations to AAAI-05: established in 2003 to honor Dr. Engel- I Colognet more’s extraordinary service to AAAI, AI I DARPA Magazine, and the AI applications com- Acknowledgments I Michael Genesereth munity, and his contributions to applied I The American Association for Artificial Google I IBM Research AI. The 2005 award will be presented to Intelligence wishes to acknowledge and I Intelligent Information Systems Institute, James Hendler, University of Maryland, thank the following individuals for their Cornell University for two decades of technical and govern- generous contributions of time and en- I Intel Corporation ment leadership in artificial intelligence, ergy to the successful creation and plan- I Microsoft Research as well as pioneering research on agent- I ning of the Twentieth National Confer- National Science Foundation I Naval Research Laboratory based systems and the semantic web. The ence on Artificial Intelligence and the I ACM/SIGART lecture will be held Wednesday, July 13, Seventeenth Conference on Innovative I Yahoo! Research Labs 9:00 AM, in the Cambria room. Applications of Artificial Intelligence. Classic Paper Award AAAI Conference Committee Chair James Hendler, University of Maryland Awards The 2005 AAAI Classic Paper Award will Awards will be presented on Monday, July be presented by AAAI president Ronald AAAI-05 Program Cochairs J. Brachman and AAAI President, and 11, and Tuesday, July 12, from 8:30 – 9:00 Manuela Veloso, Carnegie Mellon University AAAI past president and awards commit- AM, in the Allegheny Ballroom Subbarao Kambhampati, tee chair Tom Mitchell. This award is giv- Arizona State University AAAI-05 Outstanding en to the authors of the most influential IAAI-05 Program Chair and Cochair Paper Award papers from the Fifth National Confer- Neil Jacobstein, Teknowledge Corporation ence on Artificial Intelligence, held in Bruce Porter, University of Texas at Austin The outstanding paper award will be pre- 1986 in Philadelphia, Pennsylvania. The sented by AAAI-05 program chairs 2005 award winners are Steve Hanks and AAAI-05 Workshop Cochairs Manuela Veloso and Subbarao Kamb- Drew McDermott for their paper “De- Adele Howe, Colorado State University hampati to Vincent A. Cicirello, Drexel fault Reasoning, Nonmonotonic Logics, Peter Stone, The University of Texas at Austin University and Stephen F. Smith, and the Frame Problem” and David Carnegie Mellon University for their pa- AAAI-05 Tutorial Forum Cochairs Haussler for his paper “Quantifying the per “The Max K- Armed Bandit: A New Yolanda Gil, University of Southern California Inductive Bias in Concept Learning (ex- Qiang Yang, University of Science Model of Exploration Applied to Search tended abstract).” and Technology Heuristic Selection.” Distinguished Service Award AAAI-05 Student Abstract Cochairs AAAI-05 Outstanding Program The AAAI Distinguished Service Award Maria Fox, University of Strathclyde Committee Award Avi Pfeffer, Harvard University recognizes one individual each year for This award will be presented by AAAI-05 extraordinary service to the AI commu- AAAI/SIGART Doctoral program chairs Manuela Veloso and Sub- nity. The 2005 award winner is Nils Nils- Consortium Chair barao Kambhampati to Joerg Hoffmann, son of Stanford University for a lifetime Kiri Wagstaff, Jet Propulsion Laboratory Max-Planck-Institute for Computer Sci- of service to artificial intelligence, in- ence. cluding seminal scientific contributions

2CONTENTS – ACKNOWLEDGEMENTS – SPONSORS – AWARDS to pattern recognition, heuristic search, planning, and expert systems, Keynote Address pioneering achievements in scientific publishing and skilled exposition in Marvin Minsky, MIT Media Laboratory many influential textbooks, and service Marvin Minsky is the Toshiba Professor of Media Arts and Sci- to the AI and computer science commu- ences, and a professor of electrical engineering and computer nities in many key leadership roles in- science, at the Massachusetts Institute of Technology. His re- cluding president of AAAI. This award search has led to both theoretical and practical advances in ar- will be presented by Ronald J. Brachman, tificial intelligence, cognitive psychology, neural networks, and AAAI president, and Tom Mitchell, AAAI the theory of Turing Machines and recursive functions. He has past president and awards committee made major contributions in the domains of symbolic graphical description, com- chair. putational geometry, knowledge representation, computational semantics, machine perception, symbolic and connectionist learning. In 1959, Minsky and John Mc- General Game Playing Carthy founded what became the MIT Artificial Intelligence Laboratory, and his Competition Award long tenure as its codirector placed his imprint upon the entire field of artificial in- telligence. Minsky is a Fellow of AAAI, and served as AAAI president from 1981- The AAAI General Game Playing Com- 1982. petition is designed to test the abilities of general game playing systems by compar- ing their performance on a variety of games. The competition will consist of two phases: a qualification round and a Presidential Address runoff competition during AAAI. A $10,000 award will be presented to the AAAI President Ronald J. Brachman, winning entrant. AAAI gratefully ac- Corporation for National Research Initiatives knowledges the generous contribution of Michael Genesereth, who has made (AA)AI: More than the Sum of its Parts this award possible. This award will be Recent thinking has it that AI, once a vibrant, broadly-encompass- presented by Ronald J. Brachman, AAAI ing field with a grand shared vision of creating intelligent ma- aresident and Michael Genesereth, com- chines, has devolved into a loosely connected set of distinct spe- petition chair. cialty areas with little communication or mutual interest between them. To the extent that this is true, it is a disappointing development and leads one 2005 AAAI Fellows to wonder about the necessity and value of AAAI. But, Brachman argues, the conse- Recognition Dinner quences are actually far worse: because of the nature of intelligence when embed- Each year, the American Association for ded in the real world, the centrifugal force on the field is likely to thwart the very mission that drives it. Brachman will wander through some thinking on the role of Artificial Intelligence recognizes a small systems integration, the value and challenge of architecture, and some promising number of members who have made sig- developments in large projects that are helping to increase the centripetal force on nificant sustained contributions to the AI. Brachman concludes by discussing why AAAI is more essential than ever. field of artificial intelligence, and who have attained unusual distinction in the profession. AAAI is pleased to announce the four newly elected Fellows for 2005: • Usama M. Fayyad, Yahoo, Inc. ISWC-2004, IUI-2005, KCAP-2003, KDD- AAAI-05 Poster / • Raymond J. Mooney, University of Texas at 2004, KR-2004, SAT 2004, and UAI-2004. Austin Please check technical program schedule Demonstration Session • Andrew W. Moore, Carnegie Mellon University • David E. Smith, NASA Ames Research Center for presentation times. A conference-wide poster and demon- stration session will be held on Tuesday, The 2005 Fellows Recognition Dinner July 12 from 6:30 – 9:30 PM and will fea- will be held Monday, July 11, from 7:30 – Opening Reception ture AAAI-05 Technical Posters, Student 10:00 PM in the Westmoreland Room of Abstract Posters, Doctoral Consortium the Westin Hotel. A reception will begin The AAAI-05 Opening Reception will be Posters, Intelligent Systems Demonstra- at 7:30 PM, followed by dinner at 8:00 PM held Sunday, July 10, 7:00 – 8:00 PM in the tions, Robot Competition Events, and (by invitation only). Allegheny Ballroom. This event will pro- Exhibitors. (For a complete listing of vide the traditional opportunity for at- posters, please refer to pages 14–15.) tendees to socialize in a unique setting The accompanying reception will in- Sister Conference Track prior to the beginning of the first day of clude a light dinner buffet and a no-host This new program will feature represen- technical sessions. A variety of hors bar. Admittance to the reception is free tatives from AI sister conferences who d’oeuvres and a no-host bar will be avail- to AAAI-05 registrants. A $30.00 per per- will summarize the highlights of their able. Admittance to the reception is free son fee ($10.00 for children) will be most recent conferences. Participating to AAAI-05 registrants. A $30.00 per per- charged for spouses and other nontech- conferences include AAMAS 2004, ACL- son fee ($10.00 for children) will be nical conference registrants. 2003, AIIDE-05, CogSci 2004, CP 2004, charged for spouses and other nontech- ICAPS-2005, ICCBR 2003, ICML-2004, nical conference registrants.

KEYNOTE – PRESIDENTIAL ADDRESS – SISTER CONFERENCES – OPENING RECEPTION – POSTERS 3 Workshop Program Attendance at the workshops W4: Human Comprehensible W7: Integrating Planning into W12: Multiagent Learning is limited. All workshop partic- Machine Learning Scheduling Organizer: Eduardo Alonso. ipants must register for the Organizer: Dan Oblinger. Fayette, Organizer: Mark Boddy. Washing- Somerset East, second level, AAAI-05 technical program. second level, 9:00 AM – 4:05 PM ton, second level, 8:30 AM – 5:30 PM 8:45 AM – 5:00 PM W5: Inference for Textual W8: Learning in W13: Question Answering in Saturday, July 9 Question Answering Computer Vision Restricted Domains W1: Contexts and Ontologies: Organizer: Sanda M. Harabagiu, Organizer: Bir Bhanu. Organizer: Diego Molla Aliod. Armstrong, second level, Theory, Practice and Applica- Somerset East, second level, 8:30 Executive Boardroom, 26th floor, AM – 5:30 PM 8:30 AM – 5:30 PM 8:30 AM – 5:30 PM tions Organizers: Pavel Shvaiko and Deb- W14: Spoken Language W9: Link Analysis Wednesday, July 13 orah McGuinness, Washington, Understanding second level, 8:30 AM – 7:00 PM Organizers: Dunja Mladenic, Na- Organizer: Gokhan Tur. Armstrong, tasha Milic-Frayling, and Marko W10: Mobile Robotic W3: Exploring Planning and second level, 8:45 AM – 5:15 PM Grobelink. Westmoreland East, Competition and Scheduling for Web Services, second level, 8:30 AM – 5:30 PM Exhibition Workshop Grid and Autonomic Sunday, July 10 W11: Modular Construction of Chair: Jeff Forbes. Washington, second level, 12:00 – 5:30 PM Computing W2: Educational Data Mining Human-Like Intelligence Organizers: Biplav Srivastava and Organizer: Joseph E. Beck. Fayette, Organizer: Kristinn R. Thorisson. Jim Blythe. Somerset West, second second level, 8:30 AM – 5:30 PM Somerset West, second level, level, 8:30 AM – 5:30 PM 9:00 AM – 5:45 PM

AAAI/SIGART AAAI Conference Committee ences, reactions, thoughts and questions. The Conference Committee Meeting will Pictures from the conference will be up- Doctoral Consortium loaded to linked photo blog. Other stu- be held Wednesday, July 13, 7:45 – 8:45 dents attending AAAI are welcome to The Tenth AAAI/SIGART Doctoral Con- AM, Fayette Room, Westin Hotel. sortium program will be held on Satur- participate by adding their own observa- tions via comments attached to posts and day, July 9, 8:30 AM – 5:00 PM, and Sunday AAAI Publications Committee July 10, 8:45 AM – 6:00 PM in Allegheny I photographs. The blog url is http://aaai The Publications Committee meeting 05blog.blogspot.com/. on the third level of the Westin. The Doc- will be held Tuesday, July 12, 12:30 – 1:50 toral Consortium provides an opportuni- PM, Harvest Lounge, Westin Hotel. ty for a group of Ph.D. students to discuss Fourth Americas and explore their research interests and AI Magazine Editorial Board career objectives in an interdisciplinary School on Agents and The AI Magazine Editorial Board Meeting workshop together with a panel of estab- will be held Monday, July 11, 12:30 – 1:50 Multiagent Systems lished researchers. The sixteen students PM, Harvest Lounge, Westin Hotel. The Fourth Americas School on Agents accepted to participate in this program and Multiagent Systems will be held July will also participate in the Poster Session Executive Council Meeting 7 – 10 at Carnegie Mellon University and on Tuesday evening. All interested AAAI- the Westin Hotel. This program is aimed 05 student registrants are invited to ob- The AAAI Executive Council Meeting will be held Sunday, July 10, 9:00 AM – at orienting new graduate students in the serve the presentations and participate 5:00 PM, Butler Room, Westin Hotel. field of agents and multiagent systems. in discussions at the workshop. AAAI and Continental breakfast will be available at The program will consists of two days of ACM/SIGART gratefully acknowledge 8:30 AM. lectures by internationally recognized re- grants from the National Science Foun- searchers in agents and multiagent sys- dation and Microsoft that provide partial Sister Conference Speaker tems, followed by two days of AAAI-05 tu- funding for this event. Luncheon torials. Americas Program Chairs are The Sister Conference Speaker Lunch- Pragnesh Jay Modi and Paul Scerri, Meetings eon will be held Tuesday, July 12, 12:30 – Carnegie Mellon University. The event is sponsored by the International Founda- 1:50 PM, Armstrong Room, Westin Hotel. The AAAI business meeting is open to all tion for Multiagent Systems (IFMAS), the AAAI members. All other meetings are American Association for Artificial Intel- by invitation only. AAAI-05 Student Blog ligence (AAAI), and the Robotics Insti- tute (CMU/RI). AAAI Business Meeting AAAI05blog is a student run blog that The AAAI Annual Business Meeting will will describe and document AAAI-05 and be held Wednesday, July 13, 12:30 – 1:00 IAAI-05 from a student’s perspective. A PM, Somerset Room, Westin Hotel. small group of student bloggers attend- ing the conferences will post daily items describing their observations, experi-

4DOCTORAL CONSORTIUM – AMERICAS SCHOOL – STUDENT BLOG – MEETINGS – WORKSHOPS Morning AFTERNOON EVENING

Saturday, July 9 Registration Registration Tutorial Forum Tutorial Forum Workshops Workshops AAAI/SIGART DC AAAI/SIGART DC

Sunday, July 10 Registration Registration Opening Reception Tutorial Forum Tutorial Forum Workshops Workshops AAAI/SIGART DC AAAI/SIGART DC

Monday, July 11 Registration Registration AAAI Fellows Dinner AAAI-05 Keynote Address Invited Talks AAAI-05 & IAAI-05 AAAI-05 & IAAI-05 Sister Conference Highlights Sister Conference Highlights Games Competition Games Competition Exhibits & Exhibits & Robots

Tuesday, July 12 Registration Registration Posters & IS Demos Presidential Address Invited Talks and Reception IAAI–05 Invited Talk AAAI-05 & IAAI-05 AAAI-05 & IAAI –05 Sister Conference Highlights Sister Conference Highlights Exhibits & Robots Exhibits & Robots Robot Finals

Wednesday, July 13 Registration AAAI-05 / IAAI-05 AAAI-05 / IAAI-05 Invited Talks Invited Talks IAAI-05 RSE Lecture AAAI Business Meeting Exhibits Robot Workshop

CONFERENCE AT A GLANCE 5 Tutorial Forum AAAI-05 technical registrants may attend Session II: Saturday, July 9 UA4: Sensor Networks: New Challenges up to four consecutive tutorials and re- and Opportunities for AI 2:00 – 6:00 PM ceive one copy of the comprehensive Carlos Guestrin. Cambria, second level SP1: Market Clearing Algorithms AAAI-05 Tutorial Forum Notes for a small UA5: Where Do Heuristics Come From? additional registration fee. Tutorial atten- Tuomas Sandholm. Westmoreland Central, second level (Using Abstraction to Speed Up Search) dees may redeem their tutorial syllabi tick- Robert C. Holte. Westmoreland Central, second ets at the proceedings distribution area. SP2: Pyro: A Tool for Teaching level All tutorials will be held in the Westin. Robotics and AI Douglas Blank & Holly Yanco. Westmoreland Session IV: Sunday, July 10 Session I: Saturday, July 9 East, second level 2:00 – 6:00 PM 9:00 AM – 1:00 PM SP3: Exploiting Structure and Randomization for Large-Scale UP1: Bioinformatics & Machine Learn- SA1: Advances in Constraint Reasoning ing — An Introduction and Our Success Word Sense Disambiguation Stories Carla Gomes, Martin Sachenbacher, & Rada Mihalcea & Ted Pedersen. Westmoreland Brian Williams. Cambria, second level Jinyan Li & Limsoon Wong. Westmoreland East, second level Central, second level SA2: Building Intelligent Agents Using Session III: Sunday, July 10 UP2: Intelligent User Interfaces: SOAR (full day) 9:00 AM – 1:00 PM An Introduction John E. Laird. Westmoreland West, Mark T. Maybury. Allegheny II, third level second level UA1: Empirical Methods for Artificial Intelligence UP3: Systematic Bounding Techniques SA3: Downsizing Data for High Perfor- Paul Cohen. Allegheny II, third level for Combinatorial Optimization mance in Learning — Introduction to Sharlee Climer & Weixiong Zhang. Cambria, Feature Selection Methods UA2: Robotics for Beginners: Using second level Huan Liu & Robert Stine. Cambria, Robot Kits to Teach Agents and AI second level (full day) UP4: Text Analytics: Theory and Practice Ronen Feldman. Allegheny III, third level SA4: Multiagent Planning: A Survey of Elizabeth Sklar & Simon Parsons. Research and Applications Westmoreland West, second level Bradley J. Clement & Keith Decker. UA3: The Semantic Web Westmoreland Central, second level Deborah L. McGuinness &Mike Dean. Allegheny III, third level

AAAI-05 / IAAI-05 Invited Talks AAAI-05 invited talks will be held in Allegheny I and Westmoreland. IAAI-05 invited talks will be held in Cambria, unless noted otherwise. Monday, July 11 tralized knowledge systems have been mechanisms. In this talk Baral will pre- demonstrated for years, but now, thanks sent recent advances in this front, in par- AAAI-05 Keynote Address to the Web and Moore’s law, they appear ticular in the development of large sup- Marvin Minsky, MIT Media Laboratory ready for prime time. Tenenbaum will in- port structures around some of the for- Allegheny Ballroom, 9:00 – 10:00 AM troduce the architectural concepts for in- malisms and their use in application do- crementally growing an Internet-scale mains such as space shuttles, molecular IAAI-05 Invited Talk knowledge system, and describe early interaction in cells, and textual question AI Meets Web 2.0: Building commercial deployments in manufactur- answering. the Web of Tomorrow Today ing and healthcare. Jay Martin Tenenbaum, CommerceNet AAAI-05 Invited Talk Cambria, second level, 1:50 – 2:50 PM AAAI-05 Invited Talk Multiagent Learning in Games Imagine an Internet-scale From Knowledge to Intelligence— Amy Greenwald, Brown University knowledge system where Building Blocks and Applications Westmoreland, second level. 5:30 – 6:30 PM people and intelligent Chitta Baral, Arizona State University What is the outcome of agents can collaborate on Allegheny I, third level. 5:30 – 6:30 PM multiagent learning in solving complex problems Acquiring knowledge and games? Convergence can in business, engineering, reasoning with knowledge be slippery: no-regret learn- science, medicine, and other endeavors. is central to manifestation ing in repeated games con- Its resources include semantically tagged of intelligence. Thus from verges only to a set of equi- Web sites, wikis and blogs, as well as so- the early days of AI there libria, while value iteration in Markov cial networks, vertical search engines and has been a quest for devel- games can converge to a cycle. Green- a vast array of Web services from business oping knowledge representation for- wald will describe these findings and pre- processes to AI planners and domain malisms and corresponding reasoning sent the theory of stochastic stability as a models. Research prototypes of decen-

6TUTORIAL FORUM – INVITED TALKS means of characterizing the dynamics of AAAI-05 Invited Talk Robert S. Engelmore Memorial Award and Lecture multiagent learning in games. May All Your Plans Succeed! Knowledge as Power: A View from the Semantic Web Tuesday, July 12 Dana S. Nau, University of Maryland Westmoreland, second level, 1:50 – 2:50 PM James A. Hendler, University of Maryland AAAI Presidential Address Automated planning Cambria second level, 9:00 – 10:00 AM (AA)AI: More than the Sum of its Parts technology has be- come mature enough The emerging semantic web focuses on Ronald J. Brachman, Corporation for to be useful in appli- bringing KR-like capabilities to Web ap- National Research Initiatives cations that range plications in a Web-friendly Allegheny Ballroom. 9:00 – 10:00 AM from game-playing to way. The ability to put control of space vehicles. In this talk, knowledge on the Web, IAAI-05 Invited Talk Nau will discuss where automated-plan- share it, and reuse it ning research has been, where it is likely through standard Web From AI Winter to AI Spring: Can a New to go, some directions to aspire to, and mechanisms provides new Theory of Neocortex Lead to Truly In- some major challenges. and interesting challenges to Artificial telligent Machines? Intelligence. In this talk, Hendler ex- Jeff Hawkins, Founder, Numenta, Inc. Wednesday, July 13 plores the similarities and differences be- Allegheny III, third level. 10:20 – 11:20 AM tween the Semantic Web and traditional In his recent book On Intel- AAAI-05 Invited Talk AI knowledge representation systems, and sees if he can validate the analogy ligence, Hawkins proposed Representation Policy Iteration: “the semantic web is to KR as the Web is that the neocortex can be A Unified Framework for Learning to hypertext.” understood as a hierarchi- Behavior and Representation cal sequence memory. Sridhar Mahadevan, University of Since the book was written, IAAI-05 Invited Talk Massachusetts, Amherst the theory has been formalized as a mod- Real World Applications of Genetic Allegheny I, third level, 9:00 – 10:00 AM ified belief propagation network. Proto- Programming: Circuits, Optics, type implementations can solve previous- Mahadevan discusses a Dynamic System Control ly intractable vision recognition prob- longstanding intellectual Martin A. Keane, Genetic Programming lems. However the theory is not a theory puzzle in AI: How can Inc. of “vision” but a theory of cortex and agents bootstrap the learn- Cambria, second level, 1:50 – 2:50 PM therefore is applicable to many difficult ing of novel representa- AI problems. Hawkins will describe the tions from experience, Genetic programming has delivered a basics of the theory, demonstrate a work- freeing the human designer from having progression of qualitatively ing prototype, and discuss its potential to specify this knowledge? He describes a more substantial results in impact on the AI community. In Febru- novel class of adaptive planning algo- synchrony with five ap- ary Hawkins formed a new company, Nu- rithms—representation policy itera- proximately order-of-mag- menta, Inc. to promote and develop this tion—which can simultaneously learn nitude increases in the ex- technology. representations and value functions. penditure of computer time. In this talk Keane will discuss the AAAI-05 Invited Talk AAAI-05 Invited Talk most recent results: real world results in the use of genetic programming in the How Can AI and Robotics Help Us Faceted Metadata in design of electronic circuits, optical sys- Understand Social Animal Behavior? Search Interferences tems and controllers for dynamic systems Marti Hearst, University of California, Tucker Balch, Georgia Institute including examples in each field of the Berkeley of Technology functional replication of recently patent- Westmoreland, second level, 9:00 – 10:00 AM Allegheny I, third level, 1:50 – 2:50 PM ed inventions. Animal behavior, especially In the feverish debate social insect behavior, is a about how to improve well-known, rich source of search, the flexible use of inspiration for intelligent metadata has been win- system design. In this talk ning advocates. Hearst will Balch asks the inverse ques- describe the advantages tion: “How can AI contribute to the un- and pitfalls of using faceted metadata for derstanding of social animal behavior?” integrated browsing and search of large He will present several successful exam- information collections. This opens up ples, including work at his lab and else- an exciting opportunity for AI tech- where. niques to truly help improve search, since methods are needed for automated creation of categories and their labels, and assignment of items to categories, for a wide range of media types including text, audio, images, and video.

INVITED TALKS 7 Allegheny I Westmoreland Butler

Allegheny Ballroom 8:30 – 10:00 AM: AAAI-05 Welcome and Opening Remarks / Paper Award Presentations (Manuela Veloso and Subbarao Kambhampati, AAAI-05 Program Cochairs) IAAI-05 Welcome / Deployed Application Award Announcements (Neil Jacobstein and Bruce Porter, Program Cochairs) AAAI Special Award Presentations (Tom Mitchell, Awards Committee Chair and Ron Brachman, AAAI President) 9:00 – 10:00 AM: AAAI-05 Keynote Address (Marvin Minsky, MIT Media Laboratory )

Machine Learning 1 Planning 1 Multiagent Systems 1 Chair: Marie desJardins Chair: Mark Boddy Chair: Makoto Yokoo Discriminative Model Selection for Belief Net State Agnostic Planning Graphs and the Supporting Collaborative Activity AM Structures Application to Belief-Space Planning Meirav Hadad, Gilad Armon-Kest, Gal A. Kaminka, Yuhong Guo and Russ Greiner William Cushing and Daniel Bryce and Sarit Kraus nFOIL: Integrating Naïve Bayes and FOIL Genome Rearrangement and Planning OAR: A Formal Framework for Niels Landwehr, Kristian Kersting, and Luc De Raedt Esra Erdem and Elisabeth Tillier Multi-Agent Negotiation Online Query Relaxation via Bayesian Causal Conformant Planning for Domains with Jiaying Shen, Ingo Weber, and Victor Lesser Structures Discovery Constraints—A New Approach Distributing Coalitional Value Calculations Ion Muslea and Thomas J. Lee Tran Cao Son, Phan Huy Tu, Michael Gelfond, Among Cooperative Agents and A. Ricardo Morales Talal Rahwan and Nicholas R. Jennings 10:20 – 11:20

Constraint Satisfaction 1 Natural Language Processing 1 Multi-Robot Systems Chair: Weixiong Zhang Chair: Raymond Mooney Chair: Ashley Stroupe Generalized NoGoods in CSPs Scaling Up Word Sense Disambiguation via Parallel Coordination and Adaptation in Impromptu Teams PM George Katsirelos and Fahiem Bacchus Texts Michael Bowling and Peter McCracken Constrained Decision Diagrams Yee Seng Chan and Hwee Tou Ng Robust and Self-Repairing Formation Control for Kenil C. K. Cheng and Roland H. C. Yap Exploiting Subjectivity Classification to Improve Swarms of Mobile Agents A Fast Arc Consistency Algorithm for n-ary Information Extraction Jimming Cheng, Winston Cheng, and Radhika Nagpal Constraints Ellen Riloff, Janyce Wiebe, and William Phillips Heterogeneous Multirobot Coordination Olivier Lhomme and Jean-Charles Régin Capturing Expression Using Linguistic Information with Spatial and Temporal Constraints Özlem Uzuner and Boris Katz Mary Koes, Illah Nourbakhsh, and Katia Sycara 11:30 – 12:30

Vision/Challenge Temporal Reasoning Knowledge Acquisition and Information Retrieval Chair: Michael Littman Chair: Karen Myers Chair: Eyal Amir Tool Use for Autonomous Agents Temporal Dynamic Controllability Revisited Searching for Common Sense: Populating Cyc™ Robert St. Amant and Alexander B. Wood Paul Morris and Nicola Muscettola PM from the Web A Computational Model of the Cerebral Cortex Anytime, Complete Algorithm for Finding Cynthia Matuszek, Michael Witbrock, Robert C. Kahlert, John Thomas Dean Utilitarian Optimal Solutions to STPPs Cabral, David Schneider, Purvesh Shah, and Doug Lenat Samuel Meets Amarel: Automating Value Function Bart Peintner and Martha E. Pollack Impact of Linguistic Analysis on the Semantic Graph Approximation Using Global State Space Analysis Exploiting the Structure of Hierarchical Plans in Coverage and Learning of Document Extracts Sridhar Mahadevan Temporal Constraint Propagation Jure Leskovec, Natasa Milic-Frayling, Marko Grobelnik Neil Yorke-Smith An Analysis of Knowledge Collected from 1:50 – 2:50 Volunteer Contributors Timothy Chklovski and Yolanda Gil

Information Integration Planning 2 Auctions and Market-Based Systems 1 Chair: Huan Liu Chair: Reid Simmons Chair: Bernardine Dias Constraint-Based Entity Matching Planning for Stream Processing Systems Profit Sharing Auction Warren Shen, Xin Li, and AnHai Doan Anton Riabov and Zhen Liu Sandip Sen, Teddy Candale, and Susnata Basak PM A Constraint Satisfaction Approach to Geospatial Coordinating Agile Systems through the Approximating Revenue-Maximizing Reasoning Model-based Execution of Temporal Plans Combinatorial Auctions Martin Michalowski and Craig A. Knoblock Thomas Léauté and Brian C. Williams Anton Likhodedov and Tuomas Sandholm Solving Everyday Physical Reasoning Problems by Prottle: A Probabilistic Temporal Planner A New Strategy-Proof Greedy-Allocation Analogy Using Sketches Iain Little, Douglas Aberdeen, and Sylvie Thiébaux Combinatorial Auction Protocol and Its Matthew Klenk, Kenneth D. Forbus, Emmett Tomai, Extension to Open Ascending Auction Protocol 3:00 – 4:00 Hyeonkyeong Kim, and Brian Kyckelhahn Takayuki Ito, Makoto Yokoo, Atsushi Iwasaki, and Shigeo Matsubara

Machine Learning 2 Constraint Satisfaction 2 Multiagent Learning Chair: John Langford Chair: Toby Walsh Chair: Michael Littman Towards Learning Stochastic Logic Programs from Finding Diverse and Similar Solutions in Efficient No-Regret Multiagent Learning Bikramjit Banerjee and Jing Peng PM Proof-Banks Constraint Programming Luc De Raedt, Kristian Kersting, and Sunna Torge E. Hebrard, B. Hnich, B. O’Sullivan, and T. Walsh Agent-Organized Networks for Multi-Agent The Regularized EM Algorithm SAT-Based Versus CSP-Based Constraint Production and Exchange Haifeng Li, Keshu Zhang, and Tao Jiang Weighting for Satisfiability Matthew E. Gaston and Marie desJardins Discovering Domain-Specific Composite Kernels Duc Nghia Pham, John Thornton, Abdul Sattar, and Optimal Efficient Learning Equilibrium: Thomas Briggs and Tim Oates Abdelraouf Ishtaiwi Imperfect Monitoring in Symmetric Games Generating Hard Satisfiable Formulas by Ronen I. Brafman and Moshe Tennenholtz 4:20 – 5:20 Hiding Solutions Deceptively Haixia Jia, Cristopher Moore, and Doug Strain

Invited Talk Session Invited Talk Session PM From Knowledge to Intelligence — Building Blocks Multiagent Learning in Games and Applications Amy Greenwald Chitta Baral

7:30 – 10:00 PM Fellows Dinner (by invitation only) 5:30 – 6:30

8CONFERENCE SCHEDULE—MONDAY, JULY 11 Fayette Somerset Cambria (IAAI)

Coffee breaks will be held at 10:00 – 10:20 AM and 4:00 – 4:20 PM. The lunch break will be held from 12:30 – 1:50 PM.

Nonmonotonic and Common-Sense Reasoning Sister Conference Highlights Emerging Application Chair: Mirek Truszczynski Chair: Adele Howe A Multi-Agent Simulator for Teaching Police Allocation An Axiomatic Account of Formal Argumentation KDD 1: Johannes Gehrke Vasco Furtado and Eurico Vasconcelos Martin Caminada and Leila Amgoud KDD-2004: The Tenth ACM SIGKDD International Emerging Application Cumulative Effects of Concurrent Actions on Conference on Knowledge Discovery and Data Mining A Knowledge-Based Approach to Network Security: Ap- Numeric-Valued Fluents ICAPS 1: Karen Myers plying Cyc in the Domain of Network Risk Assessment Esra Erdem and Alfredo Gabaldon ICAPS-2005: The International Conference on Blake Shepard, Cynthia Matuszek, C. Bruce Fraser, William Wecht- enhiser, David Crabbe, Zelal Güngördü, John Jantos, Todd Hughes, Practical First-Order Argumentation Automated Planning and Scheduling Larry Lefkowitz, Michael Witbrock, Doug Lenat, and Erik Larson Philippe Besnard and Anthony Hunter UAI 1: Joe Halpern UAI-2004: The Twentieth Conference on Uncertainty in Artificial Intelligence

Automated Reasoning Sister Conference Highlights Deployed Application Chair: Miroslav Velev Chair: Adele Howe Loads–n-Limits and Release-n-Sequence: Compact Propositional Encodings of KDD 2: Johannes Gehrke The “Brains” behind WEPS First-Order Theories KDD-2004: The Tenth ACM SIGKDD International Paul S. Cerkez Deepak Ramachandran and Eyal Amir Conference on Knowledge Discovery and Data Mining Emerging Application Propositional Fragments for Knowledge Compilation ICAPS 2: Karen Myers Managing the Life Cycle of Plans and Quantified Boolean Formulae ICAPS-2005: The International Conference on Biplav Srivastava, Jussi Vanhatalo, and Jana Koehler Sylvie Coste-Marquis, Daniel Le Berre, Florian Letombe, Automated Planning and Scheduling and Pierre Marquis AIIDE 1: John Laird Recommender Systems: Attack Types and Strategies AIIDE-05: First Conference on Artificial Intelligence Michael P. O’Mahony, Neil J. Hurley, and Guénolé C. M. Silvestre and Interactive Digital Entertainment

Human-Computer Interaction Sister Conference Highlights IAAI-05 Invited Talk Chair: Donald Patterson Chair: Hector Munoz-Avila AI Meets Web 2.0: Building The Web of Tomorrow Mathematical Domain Reasoning Tasks in Natural CogSci 1: Ken Forbus Today Language Tutorial Dialog on Proofs CogSci 2004: The Twenty-Sixth Annual Meeting Jay Martin Tenenbaum Christoph Benzmüller and Quoc Bao Vo of the Cognitive Science Society A Decision Theoretic Model for Stress Recognition ICCBR 1: Kevin Ashley and User Assistance ICCBR 2003: The Fifth International Conference Wenhui Liao, Weihong Zhang, Zhiwei Zhu, and Qiang Ji on Case-Based Reasoning An Analysis of Procedure Learning by Instruction KCAP 1: John Gennari Jim Blythe KCAP-2003: The Second International Conference on Knowledge Capture

Vision, Music, Speech Sister Conference Highlights Emerging Application Chair: Belinda Thom Chair: David Aha Markov Decision Processes for Control of a Sensor Learning Static Object Segmentation from Motion CogSci 2: Ken Forbus Network-based Health Monitoring System Segmentation CogSci 2004: The Twenty-Sixth Annual Meeting Anand Panangadan, Syed Muhammad Ali, and Ashit Talukder Michael G. Ross and Leslie Pack Kaelbling of the Cognitive Science Society Emerging Application Modeling Form for On-line Following of Musical ICCBR 2: Kevin Ashley Activity Recognition from Accelerometer Data Performances ICCBR 2003: The Fifth International Conference Nishkam Ravi, Nikhil Dandekar, Preetham Mysore, and Michael L. Littman Bryan Pardo and William Birmingham on Case-Based Reasoning Spotting Subsequences Matching an HMM Using the KCAP 2: John Gennari Average Observation Probability Criteria with KCAP-2003: The Second International Conference Application to Keyword Spotting on Knowledge Capture Marius Calin Silaghi

Markov Decision Processes 1 Sister Conference Highlights Chair: Sven Koenig CP 1: Toby Walsh Planning and Execution with Phase Transitions CP 2004: The Tenth International Conference on Håkan L. S. Younes Principles and Practice of Constraint Programming Efficient Maximization in Solving POMDPs ISWC 1: Sheila McIlraith Zhengzhu Feng and Shlomo Zilberstein ISWC-2004: The Third International Semantic Error Bounds for Approximate Value Iteration Web Conference Rémi Munos

CONFERENCE SCHEDULE—MONDAY, JULY 11 9 Allegheny I Westmoreland Butler

Allegheny Ballroom 8:30 – 9:00 AM: AAAI-05 Program Committee Award Presentations (Manuela Veloso and Subbarao Kambhampati, Program Cochairs) General Game Playing Competition Award Presentation (Ronald J. Brachman, AAAI President & Michael Genesereth, Competition Chair) 9:00 – 10:00 AM: AAAI Presidential Address (Ronald J. Brachman)

Reinforcement Learning 1 Search Constraint Satisfaction 3 Chair: Michael Rosenstein Chair: Shlomo Zilberstein Chair: Miroslav Velev Giving Advice about Preferred Actions to External-Memory Pattern Databases Using SymChaff: A Structure-Aware Satisfiability Solver AM Reinforcement Learners Via Knowledge-Based Structured Duplicate Detection Ashish Sabharwal Kernel Regression Rong Zhou and Eric A. Hansen DC-SSAT: A Divide-and-Conquer Approach to Richard Maclin, Jude Shavlik, Lisa Torrey, Trevor Walker, AAAI-05 Outstanding Paper: The Max K- Armed Solving Stochastic Satisfiability Problems Efficiently and Edward Wild Bandit: A New Model of Exploration Applied to Stephen M. Majercik and Byron Boots Improving Action Selection in MDP’s Search Heuristic Selection Weighted Super Solutions for Constraint Programs via Knowledge Transfer Vincent A. Cicirello and Stephen F. Smith Alan Holland and Barry O’Sullivan Alexander A. Sherstov and Peter Stone Large-Scale Parallel Breadth-First Search Reasoning about Intended Actions Richard E. Korf and Peter Schultze 10:20 – 11:20 Chitta Baral and Michael Gelfond

Machine Learning 3 Temporal Reasoning 2 Auctions and Market-Based Systems 2 Chair: Pedro Domingos Chair: Eyal Amir Chair: Bernadine Dias Spectral Clustering of Biological Sequence Data Extending Continuous Time Bayesian Networks Solving the Auction-Based Task Allocation Problem PM William Pentney and Marina Meila Karthik Gopalratnam, Henry Kautz, and Daniel S. Weld in an Open Environment Data-Driven MCMC for Learning and Inference Augmenting Disjunctive Temporal Problems with David Sarne and Sarit Kraus in Switching Linear Dynamic Systems Finite-Domain Constraints Mechanism Design for Single-Value Domains Sang Min Oh, James M. Rehg, Tucker Balch, Michael D. Moffitt, Bart Peintner, and Martha E. Pollack Moshe Babaioff, Ron Lavi, and Elan Pavlov and Frank Dellaert Functional Specification of Probabilistic Process Combinatorial Auctions with k-wise Analogical Learning of Visual/Conceptual Models Dependent Valuations Relationships in Sketches Avi Pfeffer Vincent Conitzer, Tuomas Sandholm, and Paolo Santi Kenneth D. Forbus, Jeffrey Usher, and Emmett Tomai 11:30 – 12:30

Invited Talk Session Invited Talk Session How Can AI and Robotics Help Us Understand May All Your Plans Succeed! PM Social Animal Behavior? Dana S. Nau, University of Maryland Tucker Balch, Georgia Institute of Technology 1:50 – 2:50

Knowledge Representation Activity Recognition Game Theory Chair: Michael Gelfond Chair: Donald Patterson Chair: Eugene Nudelman Only-Knowing: Taking It Beyond Autoepistemic Unsupervised Activity Recognition Using Approximate Strategic Reasoning through Hierarchi-

PM Reasoning Automatically Mined Common Sense cal Reduction of Large Symmetric Games Gerhard Lakemeyer and Hector J. Levesque Danny Wyatt, Matthai Philipose, and Tanzeem Choudhury Michael P. Wellman, Daniel M. Reeves, Kevin M. Lochner, Issues in Reasoning about Interaction Networks Large-Scale Localization from Shih-Fen Cheng, and Rahul Suri in Cells: Necessity of Event Ordering Knowledge Wireless Signal Strength Coalitional Games in Open Anonymous Environ- Nam Tran, Chitta Baral, and Carron Shankland Julie Letchner, Dieter Fox, and Anthony LaMarca ments Prioritized Component Systems Multiple Agent Event Detection and M. Yokoo, V. Conitzer, T. Sandholm, N. Ohta, and A. Iwasaki Gerhard Brewka, Ilkka Niemelä, and Miroslaw Truszczynski Representation in Videos Mixed-Integer Programming Methods for 3:00 – 4:00 Asaad Hakeem and Mubarak Shah Finding Nash Equilibria Tuomas Sandholm, Andrew Gilpin, and Vincent Conitzer

Machine Learning 4 Case-Based Reasoning Chair: Drew Bagnell Competence Driven Case-Base Mining Unsupervised and Semi-Supervised Multi-Class Rong Pan, Qiang Yang, Jeffrey Junfeng Pan, and Lei Li

PM Support Vector Machines Complexity-Guided Case Discovery for Linli Xu and Dale Schuurmans Case Based Reasoning Semi-Supervised Sequence Modeling with Stewart Massie, Susan Craw, and Nirmalie Wiratunga Syntactic Topic Models Interactive Knowledge Validation and Wei Li and Andrew McCallum Query Refinement in CBR A Hybrid Generative/Discriminative Approach to Monica H. Ou, Geoff A. W. West, Mihai Lazarescu, Semi-Supervised Classifier Design and Chris Clay 4:20 – 5:20 Akinori Fujino, Naonori Ueda, and Kazumi Saito

Natural Language Processing 2 Game Playing and Multiagent Systems Chair: Raymond Mooney Chair: Richard Korf Clustering and Classifying Person Names by Origin Effective Short-Term Opponent Exploitation in Sim- PM Fei Huang, Stephan Vogel, and Alex Waibel plified Poker A Probabilistic Classification Approach for Lexical Bret Hoehn, Finnegan Southey, Robert C. Holte, and Valeriy Textual Entailment Bulitko Oren Glickman, Ido Dagan, and Moshe Koppel Search versus Knowledge for Solving Life and Death 6:30 – 9:30 PM—Allegheny Ballroom Automatic Text Summarization of Newswire: Lessons Problems in Go AAAI-05 Poster/Demo Session Akihiro Kishimoto and Martin Müller Please see page 14 – 15 for complete list of posters. Learned from the Document 5:30 – 6:30 Understanding Conference The Semantics of Potential Intentions Ani Nenkova Xiaocong Fan and John Yen

10 CONFERENCE SCHEDULE—TUESDAY, JULY 12 Fayette Somerset Cambria (IAAI)

Coffee breaks will be held at 10:00-10:20 AM and 4:00 – 4:20 PM. The lunch break will be held from 12:30 – 1:50 PM.

Logic Programming Sister Conference Highlights IAAI-05 Invited Talk Chair: Michael Gelfond Chair: Makoto Yokoo AI Winter into Spring: Can a New Theory of Properties of Programs with Monotone and Convex KR 1: Chris Welty Neocortex Lead to Truly Intelligent Machines? Constraints KR-2004: The Ninth International Conference on the Jeff Hawkins, Redwood Neuroscience Institute Lengning Liu and Miroslaw Truszczynski Principles of Knowledge Representation and Reasoning This talk will be held in Allegheny III Using SAT and Logic Programming to Design ICML 1: Russ Greiner Polynomial-Time Algorithms for Planning in ICML-2004: The Twenty-First International Conference Non-Deterministic Domains on Machine Learning Chitta Baral, Thomas Eiter, and Jicheng Zhao AAMAS 1: Milind Tambe A Theory of Forgetting in Logic Programming AAMAS 2004: The Third International Joint Kewen Wang, Abdul Sattar, and Kaile Su Conference on Autonomous Agents and Multi Agent Systems

Robotics Sister Conference Highlights Emerging Application Chair: Brian Scassellati Chair: Weixiong Zhang Automatically Acquiring Domain Knowledge for Learning CPG Sensory Feedback with Policy Gradient KR 2: Chris Welty Adaptive Game AI Using Evolutionary Learning for Biped Locomotion for a Full-Body Humanoid KR-2004: The Ninth International Conference on the Marc J. V. Ponsen, Héctor Muñoz-Avila, Pieter Spronck, and David W. Aha Gen Endo, Jun Morimoto, Takamitsu Matsubara, Jun Nakanishi, Principles of Knowledge Representation and Reasoning and Gordon Cheng ICML 2: TBD Emerging Application A Relational Representation for Procedural Task ICML-2004: The Twenty-First International Conference The DEFACTO System: Training Tool for Incident Knowledge on Machine Learning Commanders Stephen Hart, Roderic Grupen, and David Jensen Nathan Schurr, Janusz Marecki, J. P. Lewis, Milind Tambe, and Paul Scerri Recovery Planning for Ambiguous Cases in Perceptual Anchoring M. Broxvall, S. Cordeschi, L. Karlsson, and A. Saffiotti

Emerging Application Development of a Hybrid Knowledge-Based System for Multiobjective Optimization of Power Distribution System Operations Robert J. Sárfi and A. M. G. Solo Emerging Application A Learning Architecture for Automating the Intelligent Environment G. Michael Youngblood, Diane J. Cook, and Lawrence B. Holder

Markov Decision Processes 2 Sister Conference Highlights Deployed Application Chair: Haakan Younes Chair: Yolanda Gil Scheduling Engineering Works for the MTR Geometric Variance Reduction in Markov Chains. ACL 1: Dan Roth/Hwee Tou Ng Corporation in Hong Kong Application to Value Function and Gradient Estimation ACL-2003: The Forty-first Annual Meeting of the Andy Hon Wai Chun, Dennis Wai Ming Yeung, Garbbie Pui Shan Lam, Daniel Lai, Richard Keefe, Jerome Lam, and Rémi Munos Association for Computational Linguistics Helena Chan Risk-Sensitive Planning with One-Switch Utility IUI-2005: John Riedl Emerging Application Functions: Value Iteration 2005 International Conference on Intelligent The Deep Space Network Scheduling Problem Yaxin Liu and Sven Koenig User Interfaces Bradley J. Clement and Mark D. Johnston Networked Distributed POMDPs SAT 1: David Mitchell R. Patrascu, C. Boutilier, R. Das, J. Kephart, G. Tesauro, SAT 2004: The Seventh International Conference on and W. Walsh Theory and Applications of Satisfiability Testing

Multiagent Systems 2 Sister Conference Highlights Chair: Sven Koenig Chair: Yolanda Gil A Synthesis of Distributed Constraint ACL 2: Dan Roth/Hwee Tou Ng Optimization and POMDPs ACL-2003: The Forty-first Annual Meeting of Ranjit Nair, Pradeep Varakantham, Milind Tambe, the Association for Computational Linguistics and Makoto Yokoo IUI 2: John Riedl Controversial Users Demand Local Trust Metrics: IUI-2005: 2005 International Conference An Experimental Study on Epinions.com Community on Intelligent User Interfaces Paolo Massa and Paolo Avesani SAT 2: David Mitchell Cooperative Exploration in the Electronic Marketplace SAT 2004: The Seventh International Conference on David Sarne and Sarit Kraus Theory and Applications of Satisfiability Testing

Planning 3 Chair: Adele Howe Learning Measures of Progress for Planning Domains SungWook Yoon, Alan Fern, and Robert Givan Learning Planning Rules in Noisy Stochastic Worlds Luke S. Zettlemoyer, Hanna M. Pasula, and Leslie Pack Kaelbling Lazy Approximation for Solving Continuous Finite-Horizon MDPs Lihong Li and Michael L. Littman

CONFERENCE SCHEDULE—TUESDAY, JULY 12 11 Allegheny I Westmoreland Butler

Invited Talk Session Invited Talk Session Representation Policy Iteration: A Unified Frame- Faceted Metadata in Search Interfaces work for Learning Behavior and Representation Marti Hearst, University of California Berkeley School of Infor-

AM Sridhar Mahadevan, University of Massachusetts, Amherst mation Management and Systems 9:00 – 10:00

Reinforcement Learning 2 Natural Language Processing 3 Discovery and Learning Chair: Michael Littman Chair: Hwee Tou Ng Chair: Terry Zimmerman Online Resource Allocation Using Decompositional Cross-Lingual Bootstrapping of Semantic Lexicons: Representing Conditional Independence AM Reinforcement Learning The Case of FrameNet Using Decision Trees Gerald Tesauro Sebastian Padó and Mirella Lapata Jiang Su and Harry Zhang Planning in Models that Combine Memory with Robust Textual Inference Via Learning and Inducing Hierarchical Process Models in Dynamic Predictive Representations of State Abductive Reasoning Domains Michael R. James and Satinder Singh Rajat Raina, Andrew Y. Ng, and Christopher D. Manning L. Todorovski, W. Bridewell, O. Shiran, and P. Langley Value Functions for RL-Based Behavior Transfer: Learning to Transform Natural to Formal Languages Incremental Estimation of Discrete Hidden Markov A Comparative Study Rohit J. Kate, Yuk Wah Wong, and Raymond J. Mooney Models Based on a New Backward Procedure Matthew E. Taylor, Peter Stone, and Yaxin Liu German Florez-Larrahondo, Susan Bridges, and Eric A. 10:20 – 11:20 Hansen

Constraint Satisfaction 4 Search and Planning Bayesian Networks Chair: Mark Boddy Chair: Haakan Younes Chair: Reid Simmons Performing Bayesian Inference by Weighted Model Simultaneous Heuristic Search for Identifying Direct Causal Effects in Linear Models PM Counting Conjunctive Subgoals Jin Tian Tian Sang, Paul Beame, and Henry Kautz Lin Zhu and Robert Givan A Multifrontal QR Factorization Approach to Dis- Superstabilizing, Fault-Containing A Domain-Independent System for Case-Based Task tributed Inference Applied to Multirobot Localiza- Distributed Combinatorial Optimization Decomposition without Domain Theories tion and Mapping Adrian Petcu, Boi Faltings Ke Xu and Hector Muñoz-Avila Frank Dellaert, Alexander Kipp, and Peter Krauthausen Exploiting Temporal Flexibility to Cost-Algebraic Heuristic Search Distribution-Free Learning of Bayesian Network Obtain High Quality Schedules Stefan Edelkamp, Shahid Jabbar, and Alberto Lluch Lafuente Structure in Continuous Domains Nicola Policella, Xiaofang Wang, Stephen F. Smith, Dimitris Margaritis 11:30 – 12:30 and Angelo Oddi

Machine Learning 5 Logics Diagnosis Chair: Pedro Domingos Chair: Mirek Truszczynski Chair: Johan de Kleer Robust Supervised Learning Strong and Uniform Equivalence in Answer-Set On Compiling System Models for J. Andrew Bagnell PM Programming: Characterizations and Complexity Faster and More Scalable Diagnosis Finite Sample Error Bound for Parzen Windows Results for the Non-Ground Case Jinbo Huang and Adnan Darwiche Peng Zhang, Jing Peng, and Norbert Riedel Thomas Eiter, Michael Fink, Hans Tompits, and Stefan Woltran Model-Based Monitoring and Diagnosis of Speeding Up Learning in Real-time Search via Integrating Description Logics and Action Systems with Software-Extended Behavior Automatic State Abstraction Formalisms: First Results Tsoline Mikaelian, Brian C. Williams, Vadim Bulitko, Nathan Sturtevant, and Maryia Kazakevich Franz Baader, Carsten Lutz, Maja Milicic, Ulrike Sattler, and and Martin Sachenbacher Frank Wolter Diagnosis as Approximate Belief State Enumeration 1:50 – 2:50 Diagnosing Terminologies for Probabilistic Concurrent Constraint Automata Stefan Schlobach Oliver B. Martin, Brian C. Williams, and Michel D. Ingham PM 3:00 – 4:00

A coffee break will be held at 10:00 – 10:20 AM. The lunch break will be held from 12:30 – 1:50 PM.

12 CONFERENCE SCHEDULE—WEDNESDAY, JULY 13 Somerset Cambria (IAAI)

Robert S. Engelmore Memorial Award and Lecture Knowledge as Power: A View from the Semantic Web James A. Hendler, University of Maryland

Preferences Deployed Application Chair: Marie desJardins Automating the Underwriting of Insurance Applications Constraint-Based Preferential Optimization Kareem S. Aggour and William Cheetham Steve Prestwich, Francesca Rossi, Kristen Brent Venable, and Toby Deployed Application Walsh Knowledge-based Interactive Selling of Optimal Recommendation Sets: Covering Uncertainty Financial Services with FSAdvisor over User Preferences Alexander Felfernig and Alfred Kiener Robert Price and Paul R. Messinger Anyone but Him: The Complexity of Precluding an Alternative Edith Hemaspaandra, Lane A. Hemaspaandra, and Jörg Rothe

Data Mining Deployed Application Chair: Yiming Yang NESTA: NASA Engineering Shuttle Telemetry Agent A Maximum Likelihood Framework for Integrating Glenn S. Semmel, Steven R. Davis, Kurt W. Leucht, Dan A. Rowe, Taxonomies Kevin E. Smith, and Ladislau Bölöni Suju Rajan, Kunal Punera, and Joydeep Ghosh Deployed Application Enhanced Direct Linear Discriminant Analysis for CORMS AI: Decision Support System for Monitoring Feature Extraction on High Dimensional Data US Maritime Environment A. K. Qin, S. Y. M. Shi, P. N. Suganthan, and Marco Loog Haleh Vafaie, Ph.D. and CDR Carl Cecere, USPHS Using Modified Lasso Regression to Learn Large Undirected Graphs in a Probabilistic Framework Fan Li and Yiming Yang

Information Retrieval and IAAI-05 Invited Talk Human Computer Interaction Real-World Applications of Genetic Programming Chair: Rong Jin Martin Keane Selection and Ranking of Propositional Formulas for Large-Scale Service Directories Ion Constantinescu, Walter Binder, and Boi Faltings A Learning-Based Term-Weighting Approach for Information Retrieval GuangCan Liu, Yong Yu, and Xing Zhu On the Evaluation of Dynamic Critiquing: A Large-Scale User Study Kevin McCarthy, Lorraine McGinty, Barry Smyth, and James Reilly

Emerging Application Boosting Sex Identification Performance Shumeet Baluja and Henry Rowley Deployed Application TEXTAL™: Automated Crystallographic Protein Structure Determination Kreshna Gopal, Tod Romo, Erik Mckee, Kevin Childs, Lalji Kanbi, Reetal Pai, Jacob Smith, James Sacchettini, and Thomas Ioerger

A coffee break will be held at 10:00 – 10:20 AM. The lunch break will be held from 12:30 – 1:50 PM.

CONFERENCE SCHEDULE—WEDNESDAY, JULY 13 13 Generation of Hard Non-Clausal Machine Perception AAAI-05 Poster Session Random Satisfiability Problems Cross-Modal Clustering Juan A. Navarro and Andrei Voronkov The poster session will be held Tuesday, Michael H. Coen Game Theory and Economic Models Function-Based Classification from 3D July 12, in the Allegheny Ballroom, from Data via Generic and Symbolic Models A Generalized Strategy Eliminability Criterion and Michael Pechuk, Octavian Soldea, and Ehud Rivlin 6:30 – 9:30 PM. Computational Methods for Applying It Vincent Conitzer and Tuomas Sandholm Semantic Scene Concept Learning by an Autonomous Agent AAAI-05 Technical Paper Posters Fast and Compact: On a Simple Class of Weiyu Zhu Activity and Plan Recognition Congestion Games Samuel Ieong, Robert McGrew, Eugene Nudelman, Yoav Multiple-Goal Recognition from Low-Level Signals Markov Decision Processes and Uncertainty Shoham, and Qixiang Sun Xiaoyong Chai and Qiang Yang A Particle Filtering Based Approach to A Variational Learning Algorithm for the Abstract Human-Computer Interaction Approximating Interactive POMDPs Prashant Doshi and Piotr J. Gmytrasiewicz Hidden Markov Model Real-Time Classification of Electromyographic Jeffrey Johns and Sridhar Mahadevan Signals for Robotic Control Natural Language Processing Activity Recognition through Beau Crawford, Kai Miller, Pradeep Shenoy, and Speech Recognition and Rajesh Rao Goal-Based Segmentation An Inference Model for Semantic Jie Yin, Dou Shen, Qiang Yang, and Ze-Nian Li Goal-Directed Site-Independent Recommendations Entailment in Natural Language from Passive Observations Agents / Multiagent Systems Rodrigo de Salvo Braz, Roxana Girju, Vasin Pun- Tingshao Zhu, Russ Greiner, Gerald Häubl, Kevin Jewell, yakanok, Dan Roth, and Mark Sammons Team Member Reallocation via Tree Pruning and Robert Price Noa Agmon, Gal A. Kaminka, and Sarit Kraus Unsupervised Multilingual Word Sense Disambiguation via an Interlingua Solving DisCSPs with Penalty Driven Search Knowledge Representation and Reasoning Kornél Markó, Stefan Schulz, and Udo Hahn Muhammed Basharu, Ines Arana, and Hatem Ahriz Hybrid Possibilistic Networks Supervised Ranking for Pronoun An Extended Protocol for Multiple-Issue Salem Benferhat and Salma Smaoui Resolution: Some Recent Improvements Concurrent Negotiation DL-Lite: Tractable Description Logics Vincent Ng Jiangbo Dang and Michael N. Huhns for Ontologies Towards Model-Based Diagnosis of Coordination Diego Calvanese, Giuseppe De Giacomo, Domenico Lembo, Word Sense Disambiguation with Failures Maurizio Lenzerini, and Riccardo Rosati Semi-Supervised Learning Thanh Phong Pham, Hwee Tou Ng, and Wee Sun Lee Meir Kalech and Gal A. Kaminka Merging Argumentation Systems Flexible Teamwork in Behavior-Based Robots Sylvie Coste-Marquis, Caroline Devred, Sébastien Dependency Parsing with Dynamic Gal A. Kaminka and Inna Frenkel Konieczny, Marie-Christine Lagasquie-Schiex, and Pierre Bayesian Network Virginia Savova and Leonid Peshkin Modeling Human Behavior for Virtual Marquis Training Systems DD-PREF: A Language for Expressing Planning and Scheduling Yohei Murakami, Yuki Sugimoto, and Toru Ishida Preferences over Sets Quasi-Monotonic Segmentation of An Ecological Approach to Agent Population Marie desJardins and Kiri L. Wagstaff State Variable Behavior for Reactive Control Management Tractable Reasoning in First-Order Knowledge Will Fitzgerald, Daniel Lemire, and Martin Brooks Maxim D. Peysakhov, Robert N. Lass, William C. Regli, Bases with Disjunctive Information Validating Plans in the Context of Processes and Moshe Kam Yongmei Liu and Héctor J. Levesque and Exogenous Events Observation-based Model for BDI-Agents Knowledge Integration for Description Logics Maria Fox, Richard Howey, and Derek Long Kaile Su, Abdul Sattar, Kewen Wang, Xiangyu Luo, Thomas Meyer, Kevin Lee, and Richard Booth Fast Planning in Domains with Derived Predicates: Guido Governatori, and Vineet Padmanabhan Analysis of Strategic Knowledge in Back An Approach Based on Rule-Action Graphs Stable Service Placement on Dynamic Peer-to-Peer of the Envelope Reasoning and Local Search Networks: A Heuristic for the Distributed k-Center Praveen K. Paritosh and Kenneth D. Forbus Alfonso Gerevini, Alessandro Saetti, Ivan Serina, and Problem Generalized Link Properties for Paolo Toninelli Evan A. Sultanik and William C. Regli Expressive-Connections of Description Logics New Admissible Heuristics for Auctions and Market-Based Systems Bijan Parsia and Bernardo Cuenca Grau Domain-Independent Planning Patrik Haslum, Blai Bonet, and Héctor Geffner Expressive Negotiation in Settings with Externalities Logic Programming Vincent Conitzer and Tuomas Sandholm Using Domain-Configurable Search Control for A Unified Framework for Representing Probabilistic Planning Automated Reasoning Logic Program Updates Ugur Kuter and Dana Nau Yan Zhang and Norman Foo The Achilles’ Heel of QBF Sensor Selection for Active Information Fusion Carlos Ansotegui, Carla P. Gomes, and Bart Selman Machine Learning Yongmian Zhang and Qiang Ji Combining Stochastic and Greedy Search in Weighted One-Against-All Robotics Hybrid Estimation Alina Beygelzimer, John Langford, and Bianca Zadrozny Lars Blackmore, Stanislav Funiak, and Brian Williams Reactive Planning in a A Comparison of Novel and State-of-the-Art Motivated Behavioral Architecture Axiom Schemata as Metalevel Axioms: Polynomial Bayesian Network Learning Algorithms Éric Beaudry, Yannick Brosseau, Carle Côté, Clément Model Theory Laura E. Brown, Ioannis Tsamardinos, Raïevsky, Dominic Létourneau, Froduald Kabanza, Timothy L. Hinrichs and Michael R. Genesereth and Constantin F. Aliferis and François Michaud A Discourse Planning Approach to Cinematic Cam- Reducing Labeling Effort for Structured A Distributed Approach to era Control for Narratives in Virtual Environments Prediction Tasks Passive Localization for Sensor Networks Arnav Jhala and R. Michael Young Aron Culotta and Andrew McCallum Rahul Biswas and Sebastian Thrun Dependency-Directed Reconsideration Belief Base Transforming between Propositions and Features: Tactic-Based Motion Modeling Optimization for Truth Maintenance Systems Bridging the Gap and Multi-Sensor Tracking Frances L. Johnson and Stuart C. Shapiro Daniel T. Halstead and Kenneth D. Forbus Yang Gu Constraint Satisfaction and Satisfiability Non-Stationary Policy Learning in Controlling Tiny Multi-Scale Robots Old Resolution Meets Modern SLS 2-Player Zero Sum Games for Nerve Repair A. Anbulagan, Duc Nghia Pham, John Slaney, Steven Jensen, Daniel Boley, Maria Gini, and Paul Tad Hogg and David W. Sretavan and Abdul Sattar Schrater Consciousness: Drinking from the CSP Properties for Quantified Constraints: Redescription Mining: Structure Theory Firehose of Experience Definitions and Complexity and Algorithms Benjamin Kuipers Lucas Bordeaux, Marco Cadoli, and Toni Mancini Laxmi Parida and Naren Ramakrishnan Semantic Place Classification of Indoor Neighborhood Interchangeability and Dynamic Discriminative Training of Markov Logic Networks Environments with Mobile Robots Using Boosting Bundling for Non-Binary Finite CSPs Parag Singla and Pedro Domingos Axel Rottmann, Óscar Martínez Mozos, Cyrill Stachniss, Anagh Lal, Berthe Y. Choueiry, and Eugene C. Freuder Software Testing by Active Learning for and Wolfram Burgard Quick Shaving Commercial Games Learning to Prevent Failure States Olivier Lhomme Gang Xiao, Finnegan Southey, Robert C. Holte, and for a Dynamically Balancing Robot A Framework for Representing and Solving NP Dana Wilkinson Jeremy Searock and Brett Browning Search Problems Hidden Naive Bayes Autonomous Color Learning on a David G. Mitchell and Eugenia Ternovska Harry Zhang, Liangxiao Jiang, and Jiang Su Mohan Sridharan and Peter Stone

14 TECHNICAL PAPER POSTERS Mobile Robot Mapping and Localization Continuous Speech Recognition Using Modified in Non-Static Environments Stack Decoding Algorithm Cyrill Stachniss and Wolfram Burgard David Lee Game Playing Improving Simultaneous Mapping and Localization Qualitative Dimensions in Question Answering: in 3D Using Global Constraints Extending the Definitional QA Task Competition Rudolph Triebel and Wolfram Burgard Lucian Vlad Lita, Andrew Hazen Schlaikjer, WeiChang Bitbots: Simple Robots Solving Complex Tasks Hong, and Eric Nyberg The Game Playing Competition will be Anna Yershova, Benjamín Tovar, Robert Ghrist, and A Learning Support Method in held in the Washington Room on Sun- Steven M. LaValle Qualitative Simulation-Based Economic Education day from 6:00 – 9:00 PM and Monday Tokuro Matsuo, Takayuki Ito, and Toramatsu Shintani Search Evolving AI Opponents in a from 10:00 AM – 6:00 PM. An Algorithm Better than AO*? First-Person-Shooter Video Game General game players are computer Blai Bonet and Héctor Geffner C. Adam Overholtzer and Simon D. Levy systems able to accept formal descrip- Backbones and Backdoors in Satisfiability A Framework for Bayesian Network Mapping tions of arbitrary games and able to play Philip Kilby, John Slaney, Sylvie Thiébaux, Rong Pan and Yun Peng and Toby Walsh Minimizing Environmental Swings with a those games effectively without human Domain-Dependent Parameter Selection of Recurrent Neural Network Control System intervention. General game playing sys- Search-based Algorithms Compatible with Sam Skrivan, Jianna Zhang, and Debra Jusak User Performance Criteria tems are characterized by their use of Autonomous Learning of Tool Affordances Biplav Srivastava and Anupam Mediratta general cognitive information-process- by a Robot Partial Pathfinding Using Map Abstraction Alexander Stoytchev ing technologies (such as knowledge and Refinement Heuristics for Agent Routing and Itinerary representation, reasoning, learning, Nathan Sturtevant and Michael Buro Optimization on Dynamic Networks and rational behavior). Unlike special- Evan A. Sultanik Semantic Web, Information Retrieval, ized game playing systems (such as and Extraction Approximate Inference of Bayesian Networks WebCrow: A Web-Based System for through Edge Deletion Deep Blue), they do not rely on algo- Crossword Solving Julie Thornton rithms designed in advance for specific Marco Ernandes, Giovanni Angelini, and Marco Gori Towards Truthful Mechanisms for games. Query Translation Disambiguation Binary Demand Games: A General Framework as Graph Partitioning Weizhao Wang and Xiang-Yang Li Yi Liu and Rong Jin The Competition Doctoral Consortium Posters A Graph Theoretical Foundation The AAAI competition is designed to for Integrating RDF Ontologies Leveraging Language into Learning Octavian Udrea, Yu Deng, Edna Ruckhaus, Jacob Beal test the abilities of general game play- and V.S. Subrahmanian Dissertation in Progress: An Empirical Analysis of ing systems by comparing their perfor- the Costs and Benefits of Naturalness in mance on a variety of games. The com- AAAI-05 Student Abstract Posters Spoken Dialog Systems petition consists of two phases: a quali- Machine Learning and Its Application at Ellen Campana Nooksack Falls Hydroelectric Station Learning Source Descriptions for Web Services fication round and a runoff competi- Scott Alexander and Jianna Zhang Mark James Carman tion. Helicopter Routing for Maintaining Remote Computational Aspects of Mechanism Design In the qualification round, entrants Sites in Alaska Using a Genetic Algorithm Vincent Conitzer play several different types of games, in- Nicholas Armstrong-Crews and Kenrick Mock On Boosting Semantic Web Data Access Autonomous Subgoal Discovery and Hierarchical Li Ding cluding single player games (such as Abstraction for Reinforcement Learning Using Dynamic Regime Identification and Prediction maze search) and multiplayer games Monte Carlo Method Based on Observed Behavior in Electronic (such as tic-tac-toe or some variant of Mehran Asadi and Manfred Huber Marketplaces chess), including games with both com- Mixed-Initiative Approach to Collaboration Wolfgang Ketter in the Mathematical Domain Adaptive Modeling and Planning for petitors and cooperators. In some cases, Nadya Belov and Joshua Shaffer Reactive Agents the game is exhaustively searchable (as On Predicting User Intent Mykel J. Kochenderfer in tic-tac-toe); in other cases, this is not Nadya Belov Self-Emergence of Structures in Gene possible (as in chess). Players have to DR-Prolog:A System for Reasoning with Expression Programming Rules and Ontologies on the Semantic Web Xin Li handle all of these possibilities. En- Antonis Bikakis and Grigoris Antoniou Concurrent Hierarchical Reinforcement Learning trants are evaluated on the basis of con- Genre Classification of Web Documents Bhaskara Marthi sistent legal play and ability to attain Elizabeth Sugar Boese and Adele Howe Discourse Factors in Multi-Document winning positions; the best advance to Rule Refinement by Domain Experts Summarization in Complex Knowledge Bases Ani Nenkova the second round. Cristina Boicu, Gheorghe Tecuci, and Mihai Boicu Structure Learning for Statistical Relational Models In the runoff round, the best of the Use of Expert Knowledge for Decision Jennifer Neville qualifiers are pitted against each other Tree Pruning Towards Competence in Autonomous Agents Jingfeng Cai and John Durkin Özgür Simsek in a series of games of increasingly com- Learning Support Vector Machines from Rover Science Autonomy: Probabilistic plexity. The entrant to win the most Distributed Data Sources Planning for Science-Aware Exploration games in this round will be the winner Cornelia Caragea, Doina Caragea, and Vasant Honavar Trey Smith of the overall competition. Boosting Semantic Web Data Access Using Swoogle Natural Language Generation for Text-to-Text Li Ding and Tim Finin Applications Using an Information-Slim Note that, prior to the competition, Towards Exploiting Duality in Approximate Representation players are told nothing about the Linear Programming for MDPs Radu Soricut games to be played. The rules of all Dmitri Dolgov and Edmund Durfee Planning for Geospatial Data Integration games are transmitted to the players Manufacturing Processes Recognition of Snehal Thakkar Machined Mechanical Parts using SVMs Improving Reinforcement Learning Function electronically at the beginning of each Cheuk Yiu Ip and William C. Regli Approximators via Neuroevolution game. A general game playing system An Automated Distributed Meeting Scheduler Shimon Whiteson must be able to read the rules for each for FCVW Plug-in game, receive runtime information Hsiang-Hwa Koo and Elhadi Shakshuki Description Logic-Ground Knowledge Integration from the game manager, and inform and Management the manager of its moves. Joseph Kopena

STUDENT ABSTRACT POSTERS – GAME PLAYING COMPETITION 15 Drexel University researchers, with coop- Exhibit Program eration from local public protector orga- nizations, are developing a Philadelphia Area Urban Wireless Network Testbed— The Exhibit Program will be held in the Morgan Kaufmann, a live testbed for research in distributed Second level rotunda from Monday, July An Imprint of Elsevier AI in resource-constrained, dynamic 11 to Wednesday, July 13. 360 Park Avenue South communications environments. In the New York, NY 10010 Exhibit Hours Tel: 212-633-3656 Demonstrations Program, we will demo Monday, July 11 9:00 AM – 6:00 PM Fax: 212-633-3112 the AI technologies developed as part of Tuesday, July 12 9:00 AM – 6:00 PM Morgan Kaufmann, an imprint of Elsevi- this effort. Wednesday, July 13 9:00 AM – 12:00 PM er, began publishing its prestigious list of Building Applications Using End to End AAAI Press AI publications more than 20 years ago. Composition of Web Services 445 Burgess Drive New and recently published titles include Vikas Agarwal, Girish Chafle, Koustuv Das- Menlo Park, CA 94025-3442 Davies: Machine Vision, 3/e; Brachman gupta, Neeran Karnik, Arun Kumar, Ashish Tel: 650-328-3123 and Levesque: Knowledge Representation Kundu, Anupam Mediratta, Sumit Mittal Fax: 650-321-4457 and Reasoning; Hoos and Stützle: Stochas- and Biplav Srivastava, IBM India Research Email: [email protected] Laboratory, , India Online Catalog: tic Local Search Foundations and Applica- We present the first integrated work in www.aaai.org/Publications/Catalog/ tions; and Ghallab, Nau, and Traverso: composing web services end to end from Automated Planning: Theory and Practice. Google specification to deployment for building 1600 Ampitheater Parkway RoadNarrows LLC large scale applications. The control flow Mountain View, CA 1151 Eagle Dr. #140 of the plan (workflow) is generated using www.google.com Loveland, CO 80537 planning techniques based on specifica- Tel: 970 593-0370 The MIT Press tions that will change slowly (like the We all hear that intelligent robots will be 55 Hayward Street specification of the composite service) commonplace in our everyday lives. They Cambridge, MA 02142 while the data flow of the workflow is will be helpers, mentors, and rescuers. So http://mitpress.mit.edu generated based on factors that are quite how do we get to this future? Everyone The MIT Press publishes books and jour- dynamic (such as quality of service). nals in intelligent robotics, autonomous here knows the key is advances in AI, DiamondHelp: A Collaborative Task agents and complex adaptive and intelli- computer science, and technology. How Guidance Framework for gent systems. We also publish and dis- can RoadNarrows help? Well to start, stu- Complex Devices tribute books by AAAI Press. Please visit dents and researchers need good tools. Charles Rich, Candy Sidner, Neal Lesh, An- our table, browse our titles and purchase As a distributor for K-Team and Cyber- botics for over two years, RoadNarrows drew Garland, Shane Booth, and Markus publications at 20% discount. Chimani, Mitsubishi Electric Research Labora- Featured titles: Bekey / Autonomous has been learning what researchers and tories educators would like: inexpensive robots, Robots; Choset / Principles of Robot Motion; DiamondHelp is a reusable Java frame- open software interfaces, standardized Thrun / Probabilistic Robotics; Dorigo / work for building collaborative task guid- hardware modules, and solid documenta- Ant Colony Optimization ance systems for complex devices, such tion. This is the future we see for Road- as digitally enabled home appliances. Di- Narrows. We welcome your comments. amondHelp combines a generic conver- sational interface, adapted from online chat programs, with an application-spe- cific direct manipulation interface. Dia- AAAI-05 Intelligent research. This year’s demonstrations cov- mondHelp provides “a things to say” Systems Demonstrations er an impressive range of domains: natu- mechanism for use without spoken lan- ral language understanding, robotics, guage understanding; it also supports ex- The Intelligent Systems Demonstrations game-playing, cognitive aides, semantic tensions to take advantage of speech program will be held in the Allegheny web, bioinformatics, automated deduc- technology. DiamondHelp’s software ar- Ballroom on Tuesday, July 12, from 6:30 tion, musical indexing, intelligent system chitecture factors all application-specific – 9:30 PM design aides, and much more. System content into two modular plug-ins, one Continuing advances in AI research builders will be on hand to present their of which includes Collagen and a task are making it possible to develop intelli- work—all that is needed to make this model. evening a big success is your active ex- gent systems in a wide range of applica- Evolution of an Empathetic ploration of these interactive systems! tion areas. The AAAI-05 Intelligent Sys- Digital Entity: Phase One tems Demonstrations program showcases The AI Technologies of the Philadelphia Margaret Manella Kozak, Dominican Univer- state-of-the-art AI implementations and Area Urban Wireless Network Testbed sity, River Forest, Illinois provides AI researchers with an opportu- Gustave Anderson, Andrew Burnheimer, This presentation reveals the first phase nity to show applications of their re- Vincent Cicirello, David Dorsey, Chris of a multidisciplinary approach designed search in action. The program is intend- Dugan, Iris Howley, Moshe Kam, Joseph to create an evolving digital entity capa- ed to highlight innovative contributions Kopena, Rob Lass, Kris Malfettone, Andy ble of human empathy. The digital enti- Mroczkowski, Gaurav Naik, Max Peysakhov, to the science of AI with an emphasis on Brian Pyles, William Regli, Evan Sultanik, ty, ZOE (zero one entity), acquires em- the benefits to be gained from develop- James Thiel, Kyle Usbeck, Dan Venutolo, pathetic understanding through a pro- ing and using implemented systems in AI and Marc Winners, Drexel University gressive series of experiences that gener-

16 EXHIBIT PROGRAM – INTELLIGENT SYSTEMS DEMONSTRATIONS ate body-based and emotive-based feel- cease. Similarly, once the plan has been We demonstrate and discuss how the TPS ings in a digitally simulated coordinate successfully executed and a state satisfy- automated theorem proving system space. ing the externally imposed goals has proves theorems of type theory, which is Identifying Similar Words and Contexts been reached, it has been assumed that also known as higher-order logic and in- in Natural Language with SenseClusters the planning and execution behaviors cludes first-order logic. In a practical Ted Pedersen and Anagha Kulkarni, Uni- will suspend until a new goal set and con- sense, type theory is richer and more ex- versity of Minnesota, Duluth sequent plan has been imposed. These pressive than first-order logic. Notations SenseClusters is a freely available intelli- assumptions do not hold for fully au- used by TPS, which are very close to tra- gent system that clusters together similar tonomous systems, which are capable of ditional notations of logic and mathe- contexts in natural language text. There- directing their own behavior and priori- matics, are explained. tizing their own goals. The problem we after it assigns identifying labels to these QuOnto: Querying Ontologies are most concerned with is determining clusters based on their content. In addi- Andrea Acciarri, Diego Calvanese, tion, SenseClusters can be used to identi- how goals arise during the autonomous Giuseppe De Giacomo, Domenico Lembo, fy sets of related words. It is a purely un- behavior of a system. Maurizio Lenzerini, Mattia Palmieri, and Riccardo Rosati supervised approach that is language in- MGLAIR Agents in Virtual and Other dependent, and uses no knowledge oth- Graphical Environments QuOnto is a query answering system over er than what is available in raw un-anno- Stuart C. Shapiro, Josephine Anstey, David ontologies based on the DL-lite descrip- tated corpora. E. Pape, Trupti Devdas Nayak, Michael tion logic. It provides three basic func- Kandefer, and Orkan Telhan, University at tionalities: (1) specification of the inten- Language-Independent Extractive Buffalo, The State University of New York sional level of the ontology (TBox), (2) Summarization The Trial The Trail is an interactive dra- specification of the extensional level of Rada Mihalcea, University of North Texas ma running on an immersive Virtual Re- the ontology (ABox), and (3) query an- TextRank is an unsupervised system for ality system, in which a human partici- swering over the ontology. To the best of language-independent extractive sum- pant interacts with several SNePS/ our knowledge, this is the first system ex- marization that relies on an innovative MGLAIR actor-agents. Other agents to hibiting the ability to effectively answer application of iterative graph-based rank- be demonstrated include FevahrCassie, a complex queries over ontologies. ing algorithms to graphs encoding the self-aware simulated robot that under- Remote Supervisory Control of a cohesive structure of texts. An important stands and uses a fragment of English, characteristic of the system is that it does and a wumpus world agent. not rely on any language-specific knowl- Michael T. Rosenstein, Massachusetts Insti- Optimal Rhode Island Hold’em Poker tute of Technology; Andrew H. Fagg, Universi- edge resources or any manually con- Andrew Gilpin and Tuomas Sandholm ty of Oklahoma; Robert Platt Jr., John D. structed training data, and thus it is high- Sweeney, and Roderic A. Grupen, Universi- ly portable to new languages or domains. Rhode Island Hold’em Poker is a game ty of Massachusetts Amherst that has been proposed as a testbed for Participants have the opportunity to per- A Learning and Reasoning System AI research. Our research advances in form grasping and manipulation tasks for Intelligence Analysis equilibrium computation have enabled with a humanoid robot located hundreds Mihai Boicu, Gheorghe Tecuci, Cindy Ay- us to solve for the equilibrium for this ers US Army War College, Dorin Marcu, of miles away. In particular, we demon- game. Features of the equilibrium in- Cristina Boicu, Marcel Barbulescu, Bogdan strate the benefits of an assistive inter- Stanescu, William Wagner, Vu Le, Denitsa clude techniques such as bluffing, slow- face that injects artificial intelligence in Apostolova, Adrian Ciubotariu, George Ma- playing, and check-raising. Participants key places, without seizing higher-level son University will compete with our optimal opponent control from a human operator. This demo presents Disciple-LTA, a per- and will experience these strategies first- sonal cognitive assistant that can acquire hand. SenseRelate::TargetWord — A expertise in intelligence analysis directly Generalized Framework for The Proteome Analyst Suite of from intelligence analysts (with limited Word Sense Disambiguation Automated Function Prediction Tools assistance from a knowledge engineer), Siddharth Patwardhan, University of Utah; B. Poulin, D. Szafron, P. Lu, R. Greiner, D. Satanjeev Banerjee, Carnegie Mellon Univer- can train new analysts, and can help ana- Wishart, R. Eisner, A. Fyshe, B. Pearcy, and lysts solve problems through mixed-ini- sity; Ted Pedersen, University of Minnesota, L. Pireddu, Department of Computing Science, Duluth tiative reasoning. Disciple-LTA illustrates University of Alberta SenseRelate::TargetWord is a freely avail- a systematic approach to intelligence Proteome Analyst (PA) is a publicly avail- able Perl package that assigns a meaning analysis based on the task-reduction and able, high-throughput, web- based system to a given target word that appears in solution-composition. for automatically predicting the function natural language text. It selects the sense and properties of proteins. Biologists use MADbot: A Motivated and of the target word that is most related to PA to make predictions based on se- Goal Directed Robot its neighbors in the text according to quence information. Using sequence Alex Coddington, Maria Fox, Jonathan measures of similarity and relatedness Gough, Derek Long, and Ivan Serina analysis and machine learning, PA gives based on the lexical database WordNet. In most work in plan generation and ex- high accuracy and broad coverage for ecution the assumption has been made both molecular function and subcellular SAGA-ML: An Active Learning System that the goals being addressed by the localization predictions. for Semiautomated Gameplay Analysis planning system (and executive) are im- Proving Theorems of Type Theory Finnegan Southey and Robert C. Holte posed externally and that once a plan Automatically with TPS We present SAGA-ML (semi-automated has been constructed to achieve these Peter B. Andrews, Department of Mathemati- gameplay analysis by machine learning), goals the activity of the planner can cal Sciences, Carnegie Mellon University an active learning system for blackbox

INTELLIGENT SYSTEMS DMONSTRATIONS 17 Fourteenth Annual A limited number of extra Proceedings CDs are avail- AAAI Mobile Robot Competition and able for purchase. Purchase one for a Exhibition coauthor or colleague who couldn’t attend! The Fourteenth Annual Robot Competi- (Available in Registration) tion and Exhibition will be held in the foyers of the Westin and at the opening reception. The finals will take place dur- software testing, and SoccerViz, a vizual- The TaskTracer System ing the Poster/Demo session. The com- izer for Electronic Arts’ FIFA Soccer. Simone Stumpf, Xinlong Bao, Anton Dra- petition brings together more than twen- SAGA-ML samples the blackbox, learn- gunov, Thomas G. Dietterich, Jon Herlock- ty teams from universities, colleges, and ing a model of its behaviour that is used er, Kevin Johnsrude, Lida Li, and Jian- Qiang Shen, Oregon State University research laboratories to compete and to both select new samples, and summa- demonstrate cutting edge, state of the art rize the behaviour for developers. Knowledge workers face continual costs as they switch between tasks to retrieve research in robotics and artificial intelli- Solo: A Cognitive Orthosis and create information. The TaskTracer gence. Richard Simpson; University of Pittsburgh, VA project at Oregon State University has Pittsburgh Healthcare System; Edmund Lo- Robot Challenge Presti, AT Sciences; Debra Schreckenghost, developed a system that operates in the Metrica, Inc.; Ned Kirsch, University of Michi- Microsoft Windows environment which The goal of the Robot Challenge is to gan; and Steve Hayashi, VA Pittsburgh records in detail how tasks are completed work toward the development an interac- Healthcare System by tracking most interactions with desk- tive social robot. Toward that end, the Solo is a cognitive assistive device, which top applications as well as tracking challenge requires a robot participate in provides support in remembering when phone calls. In this demonstration we the AAAI conference. Aspects of confer- to perform tasks, executing the steps in a show how we have applied machine ence participation goals include locating task, and recovering from unexpected learning in this environment by develop- the conference registration desk, regis- events. The system includes an interface ing components that intelligently lever- tering for the conference, performing for clients to receive reminders, an inter- age recorded details to reduce costs. volunteer duties, and presenting a talk face for caregivers to enter information (and answering questions) at a pre- about the client’s scheduled tasks, and a TIELT: A Testbed for scribed time and location. Additionally, cognition manager that provides re- Gaming Environments the robot should socially interact with minders and task guidance at appropri- Matthew Molineaux and David W. Aha other conference participants. ate times. Many AI researchers want to test the util- Scavenger Hunt Song Search and Retrieval by Tapping ity of their systems in complex task envi- Geoff Peters, Caroline Anthony, Michael ronments defined by (such as real-time Robots search the conference hotel area Schwartz, Simon Fraser University strategy) gaming simulators or simula- for a checklist of given objects such as Content-based song searching is useful tors of computer generated forces. Also, people or information located at specific when metadata such as the song title is many developers of commercial and mil- locations or at a specific time. This task not known. This demo uses an interac- itary gaming simulators seek behaviors will require robots to navigate and map a tive Java applet that allows a user to tap that can be supported by these systems. dynamic area with moving objects and- the rhythm of a song’s melody to retrieve However, these integrations require people in order to acquire objects and the song’s name. Rhythmic contours are great effort. We will demonstrate the late satisfy a checklist. matched using a minimum edit distance Alpha version of TIELT, a testbed de- algorithm. Our demo is available to try signed to fill these needs. Open Interaction online at cgi.sfu.ca/~gpeters/tapper/ Using the GEMS System for Cancer The goal of this event is to entertain peo- tapper .cgi Diagnosis and Biomarker Discovery ple using robots and provide AI and Swoogle: Searching for Knowledge from Microarray Gene Expression Data robotics researchers a refreshing venue on the Semantic Web Alexander Statnikov, Ioannis Tsamardinos for demonstrating AI techniques for in- Tim Finin, Li Ding, Rong Pan, Anupam Jos- and Constantin F. Aliferis, Discovery System teractive, entertainment, and social hi, Pranam Kolari, Akshay, Java and Yun Laboratory, Department of Biomedical Informat- robots. Some of the topics include navi- Peng, University of Maryland, Baltimore County ics, Vanderbilt University gation, cognitive modeling, perception, Swoogle is a crawler-based semantic web We demonstrate a system GEMS for au- emotional state modeling, natural lan- indexing and retrieval service with infor- tomated development and evaluation of guage processing, and human-robot in- mation on approximately 350K semantic high-quality cancer diagnostic models teraction. web documents containing over 65M and biomarker discovery from microar- triples. We’ll demonstrate Swoogle’s use ray gene expression data. The develop- The Robot Exhibition in finding relevant ontologies, terms and ment of GEMS was informed by a rigor- The mission of the Robot Exhibition is to data; collecting and querying a set of ous algorithmic evaluation. The system demonstrate state of the art research in a documents; ranking documents and was evaluated in cross-dataset applica- less structured environment than the terms; analyzing how classes and proper- tions and using new microarray datasets. competition events. The exhibition gives ties are being used; and navigating the GEMS is freely available for noncommer- researchers an opportunity to showcase semantic web via richer semantic links. cial use from www.gems-system.org current robotics and embodied-AI re-

18 MOBILE ROBOT COMPETITION AND EXHIBITION search that does not fit into the competi- sensing, behaviors, teamwork, localiza- We obtain extremely realistic humanlike tion tasks. tion, world modeling, and locomotion facial expressions using our innovations techniques used by this team of AIBO in advanced polymers, mechanical sys- Workshop robots. We will also be illustrating how tems, and bioinspired design. The robot events culminate with a work- we use the CMDash’05 codebase as part Team: Kansas State University shop where participants describe the re- of the CMRoboBits: Creating an Intelli- Contact: David Gustafson gent Robot course. CMRoboBits is search behind their entries. Event: Exhibition taught in the Computer Science Depart- General Cochairs The Kansas State University team is a ment at CMU and emphasizes learning multi-disciplinary team consisting of two Bill Smart, Washington University in St. Louis robotic concepts from the standpoint of graduate students and three undergrad- Sheila Tejada, University of New Orleans the complete robot system (the sensing/ uates. The team will be using an Activ- cognition/actuation loop). Challenge Chair Media P3AT for the scavenger hunt. The Ashley Stroupe, Jet Propulsion Laboratory Team: Drexel Autonomous Systems Lab team has developed a flexible infrastruc- Drexel University ture for efficient off-board processing of Exhibition Chairs Contact: Paul Oh vision, localization, and control informa- Education: Lloyd Greenwald, Drexel University Events: Exhibition tion. Research: Magdalena Bugajska, Naval Re- The Drexel Autonomous Systems Lab Team: LABORIUS search Labs (DASL) will showcase a six-foot robotic University of Sherbrooke blimp. The team (Shreyansh Shah, Bill Open Interaction Event Chair Contact: Francois Michaud Morgan, and Jason Collins) recently par- Ashley Stroupe, Jet Propulsion Lab Events: Challenge, Scavenger Hunt, Open ticipated in the 2005 Indoor Flying Interaction and Exhibition Robot Competition. This year’s chal- Scavenger Hunt Chair: Our objective is to demonstrate four new lenge involved autonomous line follow- Doug Blank, Bryn Mawr College robot: Spartacus, an autonomous robot ing and teleoperated search-and-rescue with navigation, vision, and audition (for components. A wireless camera, rescue Workshop Chair localization, tracking, and simultaneous tag release mechanism and ground con- Jeffrey Forbes, Duke University separation of sound sources in open en- trol image processing are featured. vironments) capabilities; AZIMUT-2, a Robot Teams Team: Griffins modular and omnidirectional platform; Canisius College Tito, a teleoperated pedagogical robot Team: Academic Autonomy Contact: Debra Burhans for autistic children; and Roball, a spher- Swarthmore College Events: Exhibition ical robot. Spartacus capabilities are go- Contact: Bruce Maxwell We are in the process of expanding the ing to be demonstrated in the Challenge Events: Open Interaction and Exhibition role of robotics in the computer science event, and possibly in the Robot Host We will be demonstrating a social robot curriculum at Canisius College. Starting and Scavenger Hunt events, while the with an emotional model that controls with independent study projects, this other three robots are presented as part the characteristics of physical actions. past semester we included a robotics unit of the Robot Exhibition. The robot will wander through the in a newly developed course on intelli- Team: ND Rudy crowd, looking for people and identify- gent systems. Our demonstration in- University of Notre Dame ing the color of their shirts. Different cludes some of the robots developed by Contact: Matthias Scheutz shirt colors will affect the emotional state students as well as information about our Events: Open Interaction and Exhibition of the robot, causing it to modify its phys- current efforts to introduce robots into a Rudy is an interactive affective entertain- ical behavior based on its emotional breadth-first introductory course and ment robot that uses natural language to state. In addition to the emotional mod- our architecture course. el, the robot will also be demonstrating interact with people and can carry out Team: HMC Hammer the use of the Vioa-Jones Harr-wavelet simple requests. Harvey Mudd College based face detector implemented in the Team: NRL-MU Contact: Zachary Dodds OpenCV open source computer vision Events: Scavenger Hunt and Exhibition Naval Research Laboratory and Universi- project. ty of Missouri-Columbia Team: /DPR Team: Human Emulation Robots Contacts: William Adams (NRL), Alan C. Hanson Robotics, Fedex Institute of Tech- Schultz (NRL), Magdalena Bugajska Carnegie Mellon University and Intel Pitts- nology, ARRI, and UTD (NRL), Dennis Perzanowski (NRL), Scott burgh Research Contact: David Hanson Thomas (NRL), Marjorie Skubic (MU), Contact: Seth Goldstein Derek Anderson (MU), Samuel Blisard Events: Open Interaction and Exhibition Event: Exhibition (MU), J. Gregory Trafton (NRL), E. Vin- This team will show an that is a cent Cross II (NRL) Team: CMDash ‘05 portrait of late science-fiction writer Events: Open Interaction and Exhibition Carnegie Mellon University Philip K Dick, the mind behind Blade Robot Names: Sabretooth, Pyro, Storm, Cy- Contact: Paul Rybski Runner. This android incorporates the clops Event: Exhibition latest machine vision and speech recog- We will exhibit a robot that can play hide CMDash’05 is the current US Open nition technologies, as rich data steams and seek using a cognitive model to de- champion in the RoboCup legged processed by deep, natural language AI. termine where to hide and where to seek league. As part of our exhibit, we will The robot displays its states using me- in a similar fashion to humans. Separate- demonstrate the robust and efficient chanically animated facial expressions. ly, we will exhibit an intuitive sketch in-

ROBOT COMPETITION AND EXHIBITION 19 terface to a team of mobile robots, which Team: Tekkotsu Project bines AI planning techniques with the resolves sketched information with the Carnegie Mellon University human-robot interface. As part of the robots’ local sensor data. Contacts: David S. Touretzky and Ethan J. scavenger hunt and exhibition partici- Tira-Thompson Team: Penn State Abington Robotics pants can interact with multiple robots to Event: Exhibition Contact: Bob Avanzato accomplish a task. Event: Exhibition Tekkotsu (“framework” in Japanese) is an open source application development Team: Pink Team Searching framework for the Sony AIBO. The General Information Carnegie Mellon University Tekkotsu architecture provides a level of Contact: Marek Michalowski abstraction above OPEN-R, and includes Admission Event: Exhibition a suite of remote monitoring and teleop- The robot GRACE will be playing a game eration tools written in Java for portabili- Each conference attendee will receive a that involves human-robot social interac- ty. Tekkotsu is currently used by 20 uni- name badge upon registration. This tion, navigation, and interface design. versities for education, research, and ro- badge is required for admittance to the The task will be for Grace to locate and bosoccer. technical, tutorial, IAAI and workshop programs. Workshop attendees will also rendezvous with one of our team mem- Team: UML Robotics Lab be checked off a master registration list bers, who will be wearing a pink hat. The University of Massachusetts Lowell at individual rooms. Tutorial attendees game can be seen as a social version of Contact: Holly Yanco must present syllabi tickets to receive syl- “tag” or “Marco Polo,” where the robot Events: Scavenger Hunt and Exhibition senses the target not through the modal- labi volumes, and attendance tickets for We will demonstrate human-robot inter- ities of sight or sound, but rather admittance to the tutorial rooms. Smok- action with remote robot systems. The through social interactions with ing, drinking and eating are not allowed system includes a novel interface, varying strangers in the environment. The robot in any of the technical, tutorial, work- autonomy levels and a multicamera vi- will identify humans, approach them, ask shop or IAAI sessions. sion system. for directions, and follow the directions. Team: University of Pittsburgh Banking Team: Pyro Contacts: Richard Simpson, Rory Cooper, Bryn Mawr, Swarthmore and UMass Lowell An ATM machine is located adjacent to Songfeng Guo and Vinod Sharma, Univer- the gift shop in the main lobby of the Contact: Holly Yanco sity of Pittsburgh; Ed LoPresti, AT Sciences; Events: Scavenger Hunt and Exhibition Steve Hayashi, VA Pittsburgh Healthcare Sys- Westin. Pyro, which stands for Python Robotics, tem Federal Reserve Bank of Cleveland is a Python-based robotics programming Event: Exhibition (Pittsburgh Branch) environment that enables students and The Smart Wheelchair Component Sys- 717 Grant Street (3 blocks from Westin) researchers to explore topics in robotics. tem (SWCS) and Smart Power Assis- Dollar Bank Pyro provides abstractions that allow tance Module (SPAM) can be added to programs written for a commercial wheelchairs from several dif- Corner of Stanwix Street and Forbes Av- small robot to be used, without any mod- ferent manufacturers to convert them in- enue (6 blocks from Westin) ifications, to control a much larger to smart wheelchairs. The SWCS is “in- Parkville Bank robot, even one with different sensors. serted” into a power wheelchair’s control 559 Grant Street (4 blocks from Westin) system between the user’s input device Team: Stony Brook Robot Design Team and the wheelchair’s motor controller. National City Bank Stony Brook University The SPAM modifies the signal transmit- 445 Smithfield Street (5 blocks from Contacts: Matthew Marge, Nader Alrawahi, Westin) Ayman Sawas, Juan Carlos Liberato, Mur- ted to motorized “power assist” wheels taza M. Karim, Manish Muttreja, Zeynep Al- developed for manual wheelchairs. Both tinbas, Taurean Dyer, Brain Fink, Carrado systems share control of the wheelchair Business Centers Grant, Warren Halbig, Milan Karunaratne, with the operator to provide smooth, col- The following business centers are avail- Michael Molloy, Thomas Nasti, G. Oarap- lision-free navigation. The wheelchair ather, Sagar Pilania, Betson Thomas, Albert able in the area: operator is responsible for choosing Wu, Huang Yang, Rodney Yeung, and Mo- Westin hammad Yusuf when and in which direction the Affiliations: Department of Computer Sci- wheelchair moves, while the SWCS or Self Service Business Center open 24 ence, Department of Mechanical Engineer- SPAM modifies the speed of travel based hours. Accepts credit card payment. ing, and Department of Electrical and on the proximity of obstacles in the Kinko’s Computer Engineering of Stony Brook University wheelchair’s current direction of travel. 210 Grant Street (approximately 6 blocks Events: Scavenger Hunt and Exhibition Team: UNO Robotics Team from Westin). Open 24 hours. Robot Name: NavBot University of New Orleans The Stony Brook Robot Design team has Contact: Sheila Tejada Career Information focused on two main areas of research in Event: Scavenger Hunt and Exhibition A bulletin board for job opportunities in the creation of NavBot, our new robot The UNO Robotics team has designed the artificial intelligence industry will be created for the Scavenger Hunt Event: an interface for people to interact and made available in the registration area, navigation and computer vision. The collaborate with a group of heteroge- on the third level of the Westin. Atten- purpose is to create an intelligent ma- nous robots such as Aibos, wheeled dees are welcome to post job descrip- chine that is able to navigate the confer- robots and blimps. Students will demon- tions of openings at their company or in- ence floor for specific objects. strate their AI class project, which com- stitution.

20 GENERAL INFORMATION Handicapped Facilities The Westin Convention Center, Pitts- Registration burgh is equipped with handicapped fa- cilities. Conference registration is located on the third level of the Westin Convention Cen- ter, Pittsburgh, beginning Saturday, July 9. Registration hours are: Housing Saturday, July 9 7:30 AM – 5:00 PM Sunday, July 10 7:30 AM – 6:00 PM For information regarding hotel reserva- Monday, July 11 8:00 AM – 5:30 PM Tuesday, July 12 8:00 AM – 5:30 PM Wednesday, July 13 8:30 AM – 12:00 PM tions, please contact hotels directly. For student housing, please contact Only checks drawn on U.S. banks, VISA, MasterCard, American Express, govern- Duquesne University at 412-396-5083. ment purchase orders, and traveler’s checks will be accepted. We cannot accept for- eign currency or checks drawn on foreign banks. Internet Access Registration Fees Wireless Internet access is available to AAAI-05 attendees in the lobby areas of The AAAI-05/IAAI-05 technical program registration fee includes admission to the the Westin. For access information, technical sessions, the Workshop Program (preregistration only), the Opening Re- please go to onsite registration. ception, the Poster/Demonstration Session, and the AAAI-05/IAAI-05 Conference Proceedings CD. Students must present proof of full-time student status to qualify for List of Attendees the student rate. Onsite technical program fees are as follows: A list of preregistered attendees of the Technical Registration Fees conference will be available for review at Regular Member $695 the AAAI Desk in the registration area. Regular Nonmember $850 Attendee lists will not be distributed. Student Member $235 Student Nonmember $315 Parking AAAI Platinum Fees (Includes one year new or renewal membership in AAAI) Parking is available at the Westin Con- Regular US/Canada $790 vention Center, Pittsburgh fro $20.00 per Regular International $830 24 hours with no in and out privileges. Student US/Canada $270 Student International $310 Valet parking is $22.00 per 24 hours with in and out privileges. Tutorial Forum Includes admittance to up to four consecutive tutorials and the accompanying tuto- Printed Materials rial forum notes. Display tables for the distribution of pro- Regular $100 motional and informational materials of Student $25 interest to conference attendees will be Workshop Program located in the registration area. Workshop registration is limited to those active participants determined by the or- Proceedings CD ganizer prior to the conference. All workshop participants must register for the AAAI-05 technical program. Each technical registrant will receive a ticket with the registration materials for Opening Reception (Sunday, July 10) one copy of the conference CD. Tickets Adult Guest $30.00 Child $10.00 can be redeemed at the proceedings dis- tribution center, located on the second Poster/Demo Session Reception (Tuesday, July 12) level of the Westin during registration Adult Guest $30.00 Child $10.00 hours. All tickets must be redeemed on- site by Wednesday, July 13 at 11:00 AM. Registration hours are subject to change. AAAI cannot mail CDs to registrants af- ter the conference. Extra copies of the Proceedings CD may be purchased at the Registration Desk for $15.00 each. Supplies are limited. that includes a homemade soup. Open vided for your convenience, but are sub- daily from 6:00 AM – 4:00 PM. ject to change without notice. Please Restaurants (Westin Outlets) check with the provider before obtaining Orchard Café is open for breakfast and Shipping service. lunch daily from 6:30 AM – 2:00 PM. Cool The Westin can assist with small deliver- Bean Coffee offers gourmet blended cof- Taxi ies, and can provide shipping labels for fee and other beverages, assorted baked Taxi service is available at the front en- FedEx and UPS. Federal Express is also goods and pastries and light lunch items. trance to the Westin and at the Airport. Open daily from 6:00 AM – 4:00 PM. located at Kinko’s. Brown Bag Deli offers coffee, tea, break- Westin Shuttle Service fast sandwiches, assorted pastries and Transportation The Westin provides complimentary baked goods, and daily lunch specials Transportation times and fees are pro- shuttle service around the downtown

REGISTRATION INFORMATION 21 area, based on availability from 7:00 AM – Tutorial Forum Syllabi are the providers of the service. Because 11:00 PM, Monday thru Friday. The Extra copies of AAAI-05 tutorial syllabi AAAI has no control over the personnel, “Vroom Service” shuttle to and from the volume will be available for purchase in equipment or operations or providers of Westin sister property (Sheraton Station AAAI onsite registration area, beginning accommodations or other services in- Square Hotel) allows guests to dine in Monday, July 11. Supplies are limited. cluded as part of the AAAI-05/IAAI-05 their restaurant, Pittsburgh Rare, the The cost is $25.00 per volume (includes program, AAAI assumes no responsibility 2002 winner of Best New Restaurant and all tutorials). Preregistration tutorial syl- for and will not be liable for any person- third place for Best Steak Restaurant by labi tickets may be redeemed at the Pro- al delay, inconveniences or other dam- the readers of Pittsburgh Magazine. Guests ceedings distribution center on the sec- age suffered by conference participants also have the option to transfer the ond level of the Westin during registra- which may arise by reason of (1) any restaurant charges to their hotel room at tion hours. All tickets must be redeemed wrongful or negligent acts or omissions the Westin. on the part of any Supplier or its em- onsite by Wednesday, July 13 at 11:00 AM. Express Shuttle USA AAAI cannot mail books to registrants af- ployees, (2) any defect in or failure of any vehicle, equipment or instrumentali- This airport shuttle bus provides conve- ter the conference. ty owned, operated or otherwise used by nient transportation between the Westin any Supplier, or (3) any wrongful or neg- Convention Center and Pittsburgh Inter- Volunteer Station ligent acts or omissions on the part of national Airport. Pick up and drop off is The volunteer station will be located in any other party not under the control, di- from the Hotel lobby and outside the the onsite registration area. All volun- rect or otherwise, of AAAI. baggage claim area of the airport. Shut- teers are required to sign in prior to their tle time from the Westin is five minutes shift, and sign out when they finish. past each hour from 6:05 AM – 10:05 PM, Sunday – Friday and 7:05 AM – 4:05 PM, Saturday. The cost is $17.00 one-way, and Disclaimer $32.00 round trip. In offering the Westin Convention Cen- City Transit System ter, Pittsburgh, Duquesne University, The “T” light rail system begins opera- Stetson Convention Services, Inc., tion at 6:00 AM and runs throughout the AVHQ, Pittsburgh International Airport, day until 11:00 PM. The downtown areas and all other service providers (here- are free. Out of town service and Station inafter referred to as “Supplier(s)” for Square are between $1.75–$2.25 each the National Conference on Artificial In- way. For more information, visit telligence and the Innovative Applica- www.ridegold.com or contact the front tions Conference), AAAI acts only in the desk of the Westin. capacity of agent for the Suppliers that

22 DISCLAIMER Get the 2005 Proceedings in Print!

Order a copy of the 2005 Proceedings for yourself or a colleague! During the conference, you can place an order for the printed proceedings at a special discounted price (only available from July 9–13) and save. (No additional AAAI member discount may be applied)

For details and to complete your order, please go to registration.

While you’re there, pick up an extra order form to take back to your library!

A limited number of the conference CDs are also available for purchase. See the registration desk for details

23 Please join us in Boston as we celebrate 50 years of artificial intelligence research at the Twenty-First National Conference on Artificial Intelligence (AAAI-06)

July 16-20, 2006 Boston, Massachusetts

Sponsored by the American Association for Artificial Intelligence

Call for Papers: http://www.aaai.org/Conferences/National/2006/