Social Information Processing
March 26-28, 2008, Stanford University, California, USAThank you all for making the symposium such a success. We had many interesting talks and more discussions. I look forward to reconvening with this group at future meetings.
Kristina Lerman introducing the symposium.
Invited talk by Bernardo Huberman.
Gustavo presenting his Flickr work.
Gustavo with Kristina (photo courtesy of Gustavo Glusman).
Luc Steels giving his talk (photo courtesy of Gustavo Glusman).
Introduction to the symposium
DescriptionThe label social media has been attached to a quickly growing number of Web sites whose content is primarily user-driven. Examples of these sites include: blogs --- personal online journals that allow users to share their thoughts and receive feedback on them; Wikipedia --- collectively written and edited online encyclopedia; Flickr, Del.icio.us and Digg --- Web sites that allow users to share, discuss and rank photos, Web pages and news stories respectively. Other sites (e.g., Amazon's Mechanical Turk) allow users to collaborate to find innovative solutions to hard problems.
In the process of using social media sites, users are adding metadata in the form of:
- Tags: annotations of content using free-form keywords
- Ratings: passive (through using) or active (voting) evaluation of content
- Social networks: users designate other users as friends to easily track their activities
Social media facilitate new ways of interacting with information --- what we call social information processing. Social information processing allows users to collaborate implicitly to solve problems by leveraging the opinions and expertise of others. For example, in order to find good images on Flickr, the user may check what images a photographer he admires has marked as her favorite. In addition to collaborative problem solving, social information processing may lead to wholly new kinds of knowledge, that emerge from the distributed activities of many users, for example, "folksonomy" arising from the independent tagging activities of many users.
Social media sites allow for new types of signals, or "digital gestures," and use old ones in new ways. Digital gestures, like forwarding, recommending, sharing, tagging, are ephemeral instances of a web of changing relationships between users and information. We are at a point where enough computing power is available to process an increasingly larger number of these signals, which machine learning or other quantitative multidimensional analysis tools can analyze in real-time to present a highly-filtered view of the world to the end user. Such tools can address the ever-increasing noise on the Web at large. Social Information Processing could be considered as a human-centered way of addressing the pervasive and growing noise problem. Machine learning and related methods will be used to synthesize the vast amounts of socially created data in order to produce an adaptive context, or filter, for an individual user's interactions with information.
At a global level, we can see an information ecosystem arising through the interactions among users, as well as between users and information. The structure of the ecosystem gives meaning to content. For example, a community of users interested in a specific topic may emerge over time, and the linkages to users in other communities may give insight into relationships between these topics. Information ecosystems also change in time, reflecting the dynamic nature of human communities themselves. Tracking these changes may yield additional insights into the inter-relationships between users and information.
This is a rich new area of inquiry, potentially drawing on a number of disciplines from within AI as well as outside of it. The symposium is open to researchers from the academia and industry who are interested in the emergent field of social information processing.
TopicsPapers on the following topics are encouraged for submission. Research on other topics relevant to social information processing is also welcomed.
Tagging has already attracted the interest of the AI community. While the initial
purpose of tagging was to help users organize and manage their own
documents, it has since been proposed that collective tagging of
common documents can be used to organize information via an informal
classification system dubbed a ``folksonomy.'' There is hope that
folksonomies will eventually help fulfill the promise of the Semantic
Human computing and collective intelligence: What type of problems are amenable to human swarm computing approaches? How can we design the "wisdom of crowds" effect to benefit our problem solving needs?
Incentives to participation: How to elicit quality metadata and content from users? How can users resistant to tagging be encouraged to tag content?
Social networks: While users create social networks for a variety of reasons --- e.g., to track lives of friends or work or opinions of the users they respect --- network information is important for many applications. Globally, an information ecosystem may arise through the interactions among users, and between users and content. A community of users interested in a specific topic may emerge over time, with linkages to other communities giving insight into relationships between topics.
Evolution of social media and information ecosystems: How does content, and its quality, change in time? There is increasing interest in peer-production systems, for example in how and why some open-source projects like Linux and Wikipedia are successful. Under what circumstances are user-generated content sites likely to succeed and what implications does this have for information-sharing and learning within communities?
Algorithms: Before we can harness the power of the social information processing, we need new approaches to structured data analysis, specifically algorithms for synthesizing various types of metdata: e.g., social networks and tagging. Research in this area will provide a principled foundation for the development of new algorithms for social search, information discovery and personalization and other approaches that exploit the power of the social information processing.
Invited SpeakersWe are excited to have the following speakers agree to give talks at the symposium
- Bernardo Huberman (HP Labs): "Social Dynamics in the Age of the Web"
- Brian Skyrms(UC Irvine): "Signaling Games: Some Dynamics of Evolution and Learning"
SchedulePreliminary schedule of accepted papers. Unless otherwise specified, talks are 20min long with 30min discussion at the end of the session.
- Kristina Lerman (USC Information Sciences Institute)
- David Gutelius (SRI International)
- Bernardo Huberman (HP Labs)
- Srujana Merugu (Yahoo Inc.)
Program Committee (preliminary)
- Jim Blythe (USC Information Sciences Institute)
- Arindam Banerjee (U Minnesota)
- Sugato Basu (Google)
- Jack Park (SRI International)
- Jeffrey Davitz (SRI International)
- Scott Golder (HP Labs)
- Paolo Massa (Institute for Scientific and Technological Research (IRST))
- Cosma Shalizi (Carnegie Mellong University)
- Ed Chi (PARC)
- Tad Hogg (HP Labs)
- Chris Diehl (JHU APL)
- Sihem Amer-Yahia (Yahoo! Research)
RegistrationRegistration is now open. Invited participants should register by Feb 2, 2008. Registration remains open til Leap Day, 2008.
Instructions for Uploading Camera-Ready Papers
Accepted papers will appear in the symposium proceedings, published as AAAI technical reports. Authors are invited to upload camera-ready papers directly to AAAI by January 25, 2008.
The "accepted authors" button on the left hand side will take you to the Publications area of the web (http://www.aaai.org/Publications/Author/author.php). After reviewing this page, click on "electronic submission form" under Point 5. or "Electronic Submission Instructions" under "Instructions" to access the submission site (http://www.aaai.org/Publications/Author/electronic-submissions.php). Each of these pages contain important author instructions and guidelines, so please spend a few minutes reviewing these in order to complete your submission properly.
When you are ready to submit your paper, select the "AAAI Spring Symposium Series" option, and you will then be asked to
choose a symposium from the list on the submission form. please select
Spring Symposium 6 (SS-06): Social Information Processing
Participants who are unable to use the web-based form should send their abstracts and electronic files to Kristina Lerman
You must complete the AAAI Distribution License form (available from AAAI)
The signed original should be mailed to AAAI, AAAI-08 Symposium Permissions, 445 Burgess Drive, Menlo Park, CA 94025 by January 25. If the original will not reach AAAI by January 25, please also fax a copy to 650-321-4457.
Fill out the A/V form (if necessary) and submit no later than January 25 to firstname.lastname@example.org.