Skip to main content

Conference Logo

Agenda

IBM Research - Almaden - Auditorium A
8:00am - 9:00am Breakfast
9:00am - 9:30am Tessa Lau and Stefan Nusser Welcome
9:30am - 10:15am Oren Etzioni Lessons from A Century of Intelligent User Interface Research
10:15am - 10:45am Break
10:45am - 11:30am Deepak Agarwal Content Optimization on Yahoo! Front Page
11:30am - 12:00pm Poster and Demo Presenters One Minute Poster and Demo Madness
12:00am - 2:00pm Lunch and Posters
2:00pm - 2:45pm Cristina Conati User modeling: Potential and Challenges in Adaptive User Interfaces
2:45pm - 3:30pm Karen Myers CALO: An Intelligent Personal Assistant
3:30pm - 4:00pm Break
4:00pm - 4:45 Damian Isla The Two Faces of Video Game AI
4:45pm - 5:30pm Michael Picheny Statistical Methods for Conversational User Interfaces - An AI for an AI?
5:30pm - 7:00pm Evening Reception

Abstracts and Speaker Biographies


Oren Etzioni photo

Oren Etzioni, University of Washington
Lessons from A Century of Intelligent User Interface Research


Abstract

For the last century or so (measured in person years), my colleagues and I at the University of Washington have been investigating AI methods that seek to enable more powerful, flexible, and capable UIs. We have investigated general-purpose software agents, provably-reliable natural language interfaces, and much more. Along the way we have discovered what works and, more often, what doesn't. My talk will draw on this body of experience to articulate lessons and suggest directions for future work.

About the Speaker

Oren Etzioni is a Professor of Computer Science at the University of Washington's Computer Science Department, and the founder and director of the university's Turing Center. He received his Ph.D. from Carnegie Mellon in 1990, and his B.A. from Harvard in 1986. Etzioni has authored of over 100 technical papers on topics ranging from intelligent agents and natural-language interfaces to information extraction and Web search. In 2007, he received the Robert S. Engelmore Memorial Award for longstanding technical and entrepreneurial contributions to AI. In 2005, he was received the IJCAI Distinguished Paper Award, and in 2003 he was chosen as a AAAI Fellow.

Etzioni is the founder of three companies. In 2003, he founded Farecast (www.farecast.com), a company that utilizes data mining to inform consumers about the right time to buy their air tickets. His work has been featured in the New York Times, Wall Street Journal, NPR, SCIENCE, The Economist, TIME Magazine, Business Week, Newsweek, Discover Magazine, Forbes Magazine, Wired, and elsewhere.

Deepak Agarwal photo

Deepak Agarwal, Yahoo! Research
Content Optimization on Yahoo! Front Page


Abstract

We consider the problem of algorithmically selecting articles to serve to a user on a module (Today Module) when he/she visits the Yahoo! Front Page, choosing from an editorially programmed pool that is frequently refreshed. The system is now in production and selects articles to serve to several hundred million user visits per day, significantly increasing the number of user clicks over the original manual approach, in which editors periodically selected articles to display. We highlight the business and algorithmic challenges encountered in building such a system. The major business challenge was to understand how to blend years of editorial experience in placing articles with automated serving schemes based on statistical algorithms. Some of algorithmic challenges we face include a dynamic content pool, short article lifetimes, non-stationary click-through rates, and extremely high traffic volumes. The fundamental problem we must solve is to quickly identify which items are popular (perhaps within different user segments), and to exploit them while they remain current. A key component of our system is a scalable infrastructure that facilitates near real time model updates enabling us to react fast in embracing emerging new but high quality content and shredding old deteriorating ones. Surprisingly, personalization of content to users/user segments did not succeed with our current editorial programming scheme. We provide insights on why personalization fails and suggest strategies to address them in future.

About the Speaker

Deepak Agarwal is currently a senior research scientist at Yahoo! Research. Prior to joining Yahoo!, he was a member of the statistics department at AT&T Research where he worked extensively on methods to mine information from massive graphs, statistical models for social network analysis. He is currently working on scalable statistical methods for collaborative filtering, online advertising and content optimization. He serves regularly on program committees in Data Mining, Machine Learning and Statistics; he has won several best paper awards at prestigious conferences. He received his PhD in statistics from the University of Connecticut.


Cristina Conati photo

Cristina Conati, University of British Columbia
User modeling: Potential and Challenges in Adaptive User Interfaces


Abstract

Adaptive interfaces aim to improve human-computer interaction by providing users with tailored support for complex tasks. This support involves acquiring a model of user traits relevant to adequately tailor the interaction, i.e., a user model. Depending on the nature of the task, the relevant user traits may include simple performance measures (such as frequencies of interface actions), domain-dependent cognitive traits (such as knowledge and goals), meta-cognitive processes that cut across tasks and domains (such as reasoning by analogy), and affective states. Arguably, the higher the level of the traits to be captured, the more difficult it is to assess them unobtrusively from simple interaction events. In this talk, I will first describe some of our solutions to this challenge, and then I will discuss two important open questions in user modeling research: how much does an adaptive system really need to know about its user to provide a good service? And how much does the user need to know about what the system knows?

About the Speaker

Dr. Conati is an Associate Professor of Computer Science at the University of British Columbia. She received a "Laurea" degree (M.Sc. equivalent) in Computer Science at the University of Milan, Italy (1988), as well as a M.Sc. (1996) and Ph.D. (1999) in Artificial Intelligence at the University of Pittsburgh. Dr. Conati's research goal is to integrate research in Artificial Intelligence (AI), Cognitive Science and Human Computer Interaction (HCI) to make complex interactive systems increasingly more effective and adaptive to the users' needs. Her areas of interest include Adaptive Interfaces, User Modeling, Affective Computing and Intelligent Tutoring Systems. Dr. Conati has served on program committees and as a reviewer for major AI and HCI conferences/journals. She was program co-chair of User Modeling 2007, the 11th International Conference on User Modeling, and she is conference co-chair of IUI 2009, the International Conference on Intelligent User Interfaces. She published over 40 strictly referred articles, and her research has received awards from the International Conference on User Modeling (1997), the International Conference of AI in Education (1999), the International Conference on Intelligent User Interfaces (2007) and the Journal of User Modeling and User Adapted Interaction (2002).


Karen Myers photo

Karen Myers, SRI International
CALO: An Intelligent Personal Assistant


Abstract

In today's internet-enabled world, a typical knowledge worker is flooded with relevant information while productivity demands have increased sharply. Intelligent personal assistants hold promise for addressing the problems of information and task overload. To be effective, however, an intelligent assistant must complement user work practices while supporting significant adaptation to and directability by the user. In addition, the assistant must evolve and extend its knowledge over time to ensure its relevance and utility. This talk describes a large-scale effort at SRI to build an intelligent personal assistant with these qualities, called CALO (Cognitive Assistant that Learns and Organizes). The CALO project draws on contributions from 25 different research organizations across the country in areas that include learning, speech and language understanding, reasoning, and agent technologies.

About the Speaker

Dr. Karen Myers is a Principal Scientist at SRI International, as well as Director of the Intelligent Mixed-initiative Planning and Control Technologies program within SRI's Artificial Intelligence Center. Dr. Myers joined SRI in 1991 after completing a Ph.D. in computer science at Stanford University. Her technical interests lie with the development of technologies that enable humans and machines to solve problems collaboratively. One major thrust for her work has been the development of advisable systems, in which a user can direct automated technology by providing a range of guidance to influence its operations. Her work on mixed-initiative planning systems pioneered concepts for sketch-based planning, interactive exploration of complex solution spaces, and plan summarization and explanation techniques, spanning both military and space applications. Currently, she is leading a multi-institutional team focused on developing intelligent personal assistants that can acquire and evolve their problem-solving knowledge over time. Dr. Myers recently completed a three-year term as a member of the Executive Council for the Association for the Advancement of Artificial Intelligence (AAAI). She is on the Executive Council for the International Conference on Automated Planning and Scheduling and the editorial board for Artificial Intelligence. She previously served on the editorial board for the Journal of Artificial Intelligence Research as well as numerous program commitees for conferences and workshops in the areas of intelligent agents, planning and scheduling, knowledge representation, and applications of AI.


Damian Isla photo

Damian Isla, Bungie Studios
The Two Faces of Video Game AI


Abstract

Some of the most interesting AI in the world right now is being built for computer games, where players routinely pit their wits and skills against those of (apparently) living (apparently) breathing (apparently) intelligent computer-controlled characters. To the player, the AI represents one of the most intuitive user interfaces imaginable: an entity that is not only expressive but also highly intentional: you have no doubts as to the goals, desires and internal states of a Brute Chieftan when you're facing down the business end of his fuel-rod gun.

At the same time, AI, in most important ways, IS gameplay, and thus its functioning is every bit as much about art and design as it is about engineering -- player's are liable to consider the AI "smart" if it is well-animated, for example. So there are really TWO important human-computer interfaces that a good video game AI system needs to support: one which communicates intention to the player, and another through which designers and other content-creators generate, modify and analyse decision-making and behavior. And these two "faces" have very different, often contradictory needs.

This talk will discuss the general role of AI in game development, and will introduce the audience to some of the ways in which the Halo AI system balances the needs of the developers and the needs of the player.

About the Speaker

Damian Isla is the Artificial Intelligence Lead at Bungie Studios, where he was responsible for the AI for mega-hit first-person shooters HALO 2 and HALO 3, two of the most successful and fastest-selling video games of all time. Before coming to Bungie, he earned an M.Eng. at the MIT Media Lab with Bruce Blumberg's Synthetic Characters Group, where he did research on learning and behavior for artificial creatures. He has spoken on games and character AI at the Game Developers Conference, the International Joint Conference on Artificial Intelligence (IJCAI), the AI and Interactive Digital Entertainment Conference (AIIDE) and at Siggraph, and was a contributor to Game AI Programming Wisdom 1 and 3. He also has a B.Sc. in Computer Science, again from MIT. He enjoys drama of all kinds.


Michael Picheny photo

Michael Picheny, IBM TJ Watson Research Center
Statistical Methods for Conversational User Interfaces - An AI for an AI?


Abstract

Traditional UI design has typically consisted of a set of user interface experts studying a design problem in depth and manually iterating on the parameters with a set of users. However, with the exponential growth of the Internet and its associated data, it is now possible to utilize statistical models in the design of novel UIs. Such models, infeasible ten or more years ago because of lack of computation and data, now enable a whole new set of novel applications. This talk will describe some of the basic technology advances in statistical modeling specifically aimed at conversational user interfaces, with specific examples drawn from such areas as mobile search, speech to speech translation, telematics, and analysis of human-human interactions in call centers. The talk will conclude with some speculations on how future interface design may become dominated by statistical methods as access to data becomes more and more prevalent.

About the Speaker

Michael Picheny is the Senior Manager of the Speech and Language Algorithms Group at the IBM TJ Watson Research Center. Michael has worked in the Speech Recognition area since 1981, joining IBM after finishing his doctorate at MIT. He has been heavily involved in the development of almost all of IBM's recognition systems, ranging from the world's first real-time large vocabulary discrete system through IBM's current product lines for telephony and embedded systems. He has published numerous papers in both journals and conferences on almost all aspects of speech recognition. He has received several awards from IBM for his work, including three outstanding Technical Achievement Awards and two Research Division Awards. He is the co-holder of over 20 patents and was named a Master Inventor by IBM in 1995 and again in 2000. Michael served as an Associate Editor of the IEEE Transactions on Acoustics, Speech, and Signal Processing from 1986-1989, was the chairman of the Speech Technical Committee of the IEEE Signal Processing Society from 2002-2004, and is a Fellow of the IEEE. He served as an Adjunct Professor in the Electrical Engineering Department of Columbia University in 2004-2005 and co-taught a course in speech recognition. He is currently a member of the board of ISCA (International Speech Communication Association).

About the conference organizers

Tessa Lau photo

Tessa Lau is a Research Staff Member at IBM Research - Almaden. She joined IBM in 2001 after completing a PhD in computer science at the University of Washington. Dr. Lau's research goal is to integrate techniques from artificial intelligence and human-computer interaction to build systems that enhance human productivity and creativity. She is an expert in the area of programming by demonstration, whose goal is to enable regular end users to automate routine tasks simply by demonstrating how to perform their desired task. Dr. Lau currently leads the CoScripter project, which is building a platform for collaborative scripting of web-based tasks. She has served on program committees and as a reviewer for major AI and HCI conferences and journals. Dr. Lau was Program Co-Chair of IUI 2007, the international conference on intelligent user interfaces. She also serves on the board of CRA-W, the CRA committee on the status of women in computing research.


Stefan Nusser photo Stefan Nusser is Senior Manager of the User-Focused Systems research group at the IBM Research - Almaden in San Jose, California. He oversees a team of researchers responsible for understanding and improving how people interact with information technology. Stefan has held various positions in both research and software development over his ten year career with IBM; focusing on a variety of topics from digital media, digital rights management, content management to most recently web 2.0 and collaborative software. Before taking on his current role, Stefan was managing a research team focused on applying web 2.0 client technologies and software trends to enterprise collaboration applications. Prior to that, he managed a Content Protection research group at the Almaden Research Center while acting as the IBM Research Relationship Manager for the Media and Entertainment Industry. Stefan received his PhD in Management Information Systems from Vienna University of Business Administration and Economics in 1997.

[an error occurred while processing this directive]