Open Grid Forum workshop on Social Networks, Semantic Grids and Web

Noshir Contractor Professor, Departments of Speech Communication & Psychology Director, Age of Networks Initiative, Center for Advanced Study Director, Science of Networks in Communities - National Center for Supercomputing Applications University of Illinois at Urbana-Champaign [email_address] From Disasters to WoW How the Science of Networks can inform – and be informed by – the Grid and Web 2.0

Turn on power & set MODE with MODE button. You can confirm the MODE you chose as the red indicator blinks. Lamp blinks when (someone with) a Lovegety for the opposite sex set under the same MODE as yours comes near. FIND lamp blinks when (someone with) a Lovegety for the opposite sex set under different mode from yours comes near. May try the other MODES to “GET” tuned with (him/her) if you like.

Key Takeaways Desire for social networking tools to assist multidisciplinary, interdisciplinary, and transdisciplinary (MIT) collaboration Development of multi-theoretical multilevel (MTML) understandings of why we create, maintain, dissolve and reconstitute social network links. Development of tools and algorithms to – Discover, Diagnose, and Design – (3D) more effective social networks Opportunity to harvest empirical multi-dimensional network data from recent efforts on the Grid and Web 2.0. Opportunity to enable more effective social networking within the Grid Challenge to develop MTML 2.0 to explain creation of links in multidimensional networks

Aphorisms about Networks Social Networks : Its not what you know, its who you know. Cognitive Social Networks : Its not who you know, its who they think you know. Knowledge Networks : Its not who you know, its what they think you know.

Source: Newsweek, December 2000 Cognitive Knowledge Networks

Human Agent to Human Agent Communication Retrieving from knowledge repository Publishing to knowledge repository Non Human Agent to Non Human Agent Communication Non Human Agent (webbots, avatars, databases, “ push” technologies) To Human Agent INTERACTION NETWORKS Source: Contractor, 2001

Human Agent’s Perception of What Another Human Agent Knows Non Human Agent’s Perception of Resources in a Non Human Agent Non Human Agent’s Perception of what a Human Agent knows Human Agent’s Perception of Provision of Resources in a Non Human Agent COGNITIVE KNOWLEDGE NETWORKS * * … Why Tivo thinks I am gay and Amazon thinks I am pregnant ….

Human to Human Interactions and Perceptions Human to Non Human Interactions and Perceptions Non Human to Human Interactions and Perceptions Non Human to Non Human Interactions and Perceptions Non Human Agent Y Non Human Agent X Human C Human B Human A Non Human Agent Y Non Human Agent X Human C Human B Human A

WHY DO WE CREATE, MAINTAIN, DISSOLVE, AND RECONSTITUTE OUR COMMUNICATION AND KNOWLEDGE NETWORKS?

Social Drivers: Why do we create and sustain networks? Theories of self-interest Theories of social and resource exchange Theories of mutual interest and collective action Theories of contagion Theories of balance Theories of homophily Theories of proximity Theories of co-evolution Sources: Contractor, N. S., Wasserman, S. & Faust, K. (2006). Testing multi-theoretical multilevel hypotheses about organizational networks: An analytic framework and empirical example. Academy of Management Review . Monge, P. R. & Contractor, N. S. (2003). Theories of Communication Networks . New York: Oxford University Press.

“ Structural signatures” of Social Drivers Theories of Self interest Theories of Exchange Theories of Collective Action Theories of Balance Theories of Homophily Theories of Cognition F E D B C A - + Novice Expert

Co-evolution of knowledge networks and 21 st century organizational forms NSF KDI Initiative 1999-04. PI: Noshir Contractor, University of Illinois. Co-P.I.s: Monge, Fulk, Bar (USC), Levitt, Kunz (Stanford), Carley (CMU), Wasserman (Indiana), Hollingshead (Illinois). Three dozen industry partners (global, profit, non-profit): Boeing, 3M, NASA, Fiat, U.S. Army, American Bar Association, European Union Project Team, Pew Internet Project, etc.

Transactive Memory Perception of Other’s Knowledge Communication to Allocate Information Social Exchange - Retrieval by coworkers on other topics Public Goods / Transactive Memory Allocation to the Intranet Retrieval from the Intranet Perceived Quality and Quantity of Contribution to the Intranet Inertia Components Collaboration Co-authorship Communication Motivations for search In Knowledge Nets Proximity -Work in the same location

1. Social Communication 0.144 2. Perception of Knowledge & Communication to Allocate 0.995 3. Perception of Knowledge & Provision 0.972 4. Perception of Knowledge, Social Exchange, & Social Communication 0.851 5. Perception of Knowledge, Proximity, & Social Communication 0.882 Motivation for Search in Knowledge Networks Using Exponential Random Graph Modeling Techniques

A contextual “meta-theory” of social drivers for creating and sustaining networks

Core Research Social Drivers for Creating & Sustaining Communities Business Applications PackEdge Community of Practice (P&G) Vodafone-Ericsson “Club” for virtual supply chain management (Vodafone) Societal Justice Applications Cultural & Networks Assets In Immigrant Communities (Rockefeller Program on Culture & Creativity) Economic Resilience NGO Community (Rockefeller Program on Working Communities) Entertainment Applications World of Warcraft (NSF) Everquest (NSF, Sony Online Entertainment) Science Applications CLEANER: Collaborative Large Engineering & Analysis Network for Environmental Research (NSF) CP2R: Collaboration for Preparedness, Response & Recovery (NSF) TSEEN: Tobacco Surveillance Evaluation & Epidemiology Network (NSF, NIH, CDC) Projects Investigating Social Drivers for Search/Diffusion in Networks

Contextualizing Goals of Communities Challenges of empirically testing, extending, and exploring theories about networks … until now

Enter: Cyberinfrastructure & Web 2.0

Components of Cyberinfrastructure High Performance Computing services Grid computing services Data, Data Analysis, & Visualization Data Services (access to static, streaming, and sensor data) Workflow & Visual-Analytic Tools (access to software tools, computing power, etc.) Document Libraries (documents, images, data sets, etc.) Collaboratories, Observatories, & Virtual Organizations Collaboration Services (chat, video conferencing, blog, wiki, collaborative editing) Referral Services (people, data, documents, analytic tools) Customization interfaces for communities and sub-communities Education & Workforce Development Training, Outreach, Mentoring services

Science and Engineering Cyberinfrastructures

Multidimensional Networks in CI (Cyberinfrastructure) Multiple Types of Nodes and Multiple Types of Relationships

Its all about “Relational Metadata” Technologies that “ capture ” communities’ relational meta-data (Pingback and trackback in interblog networks, blogrolls, data provenance) Technologies to “ tag ” communities’ relational metadata (from Dublin Core taxonomies to folksonomies (‘wisdom of crowds’) like Tagging pictures and video (Flickr, YouTube) Tagging blogs and news (Technorati, digg) Social bookmarking (del.icio.us, LookupThis, BlinkList) Social citations (CiteULike.org) Social libraries (discogs.com, LibraryThing.com) Social shopping (SwagRoll, Kaboodle, thethingsiwant.com) Social networks (FOAF, XFN, MySpace, Facebook) Technologies to “ manifest ” communities’ relational metadata (Tagclouds, Recommender systems, Rating/Reputation systems, ISI’s HistCite, Network Visualization systems)

Bios, titles & descriptions Personal Web sites Google search results Web of Science Citation Digital Harvesting of Relational Metadata CI-KNOW Analyses and Visualizations http://iknowinc.com/iknow/sb_digital_forum/www/iknow.cgi CATPAC UBERLINK NETDRAW

Core Research Social Drivers for Creating & Sustaining Communities Business Applications PackEdge Community of Practice (P&G) Vodafone-Ericsson “Club” for virtual supply chain management (Vodafone) Societal Justice Applications Cultural & Networks Assets In Immigrant Communities (Rockefeller Program on Culture & Creativity) Economic Resilience NGO Community (Rockefeller Program on Working Communities) Entertainment Applications World of Warcraft (NSF) Everquest (NSF, Sony Online Entertainment) Science Applications CLEANER: Collaborative Large Engineering & Analysis Network for Environmental Research (NSF) CP2R: Collaboration for Preparedness, Response & Recovery (NSF) TSEEN: Tobacco Surveillance Evaluation & Epidemiology Network (NSF, NIH, CDC) Projects Investigating Social Drivers for Communities

CLEANER Grand Challenge How do we detect and predict waterborne hazards in real time? How do we predict the effects of human activities on the quantity, distribution, and quality of water? How do we improve water cycle engineering management strategies to provide water quantity and quality to sustain humans and ecosystems?

Enabling Environmental Engineering Communities with Cyberinfrastructure CLEANER Collaborative Large-scale Engineering Analysis Network for Environmental Research Human-dominated, complex environmental systems, e.g. , River basins Coastal margins What researchers requested: Access to “live” and archived sensor data Customized network referrals to people, documents, datasets Analyze, visualize and compare data Organize, automate and share cyber-research processes Users can simultaneously view and discuss data and analyses

CLEANER Community: A multidimensional network NCSA David Streamflow Analyst Mary CHATS WITH USES Hydrologic Information System Status Report C-U County Hydrologic Dataset AFFILIATED WITH watersheds EXPERT IN River Monitoring Systems Contains as Keyword HUDSON RIVER Hydro DATA DOWNLOADS AUTHOR OF Searches for Keyword Annie

CI-KNOW: Harvesting the online community’s relational meta-data INPUTS Cybercommunity Resources Cyberinfrastructure Use External Resources PROCESSES Network Maps Network Referrals OUTPUTS Network Diagnostics Generating a Multi- Dimensional network Network Analysis Users’ Profiles Documents Collaboration Tools Datasets Analysis Tools Bibliographic DBs Personal Websites Organizational Websites Project Websites Patent Databases Linking all data together Algorithms to generate Network Referrals 2. Algorithms to create Network Maps 3. Algorithms to compute Network Diagnostics User activity logs related to cyberinfrastructure Downloading Presentations Using Tools to Analyze Datasets Using Chats, Forum

CI-KNOW: Harvesting the online community’s relational meta-data INPUTS Cybercommunity Resources Cyberinfrastructure Use External Resources PROCESSES Network Maps Network Referrals OUTPUTS Network Diagnostics Generating a Multi- Dimensional network Network Analysis Who to contact for what topic What tools to use for what data What dataset to analyze for what concepts What papers to read for what keywords What nodes are important for what relations The amount of scanning, absorption, diffusion, robustness, vulnerability in a network

Cybercommunity Resources David C. writes a message on a CLEANER forum about hydrological monitoring systems. Documents Collaboration Tools Data sharing Analysis Tools CI-KNOW Portlet

Cybercommunity Resources Documents Collaboration Tools Data sharing Analysis Tools CI-KNOW Portlet

Recommending People David C. retrieves hydrological datasets on C-U County rivers CI-KNOW Portlet

The strongest recommendation in the entire Knowledge network is given a value of 1. Other recommendations are scaled by this global maximum. Structural recommendations reflect the profile similarity of items and take the user’s connections into account. Recommendation lists from multiple modules are merged and duplicates are removed, retaining highest scores A user’s recommendation history is saved and recommendations are generally not repeated within a session When users get recommendations for entities to which they do not have direct access, they are offered visual network paths to reach those entities. CI-KNOW: Cyberinfrastructure Knowledge Networks on the Web A Recommender System for Locating Resources in a Knowledge Network Elements of CI-KNOW Multi-disciplinary research and policy communities have indicated substantial interest in a network referral system that would identify available resources of interest to their members, e.g.: Researchers with similar or complementary expertise Popular, new, or relevant data sets Appropriate analysis tools Models or model systems Visual analytic tools, workflows, etc To meet this interest, the Science of Networks in Communities (SONIC) group’s social networking technology, CI-KNOW, is integrated into portals and other collaborative environments. CI-KNOW is a suite of software tools that leverages automated data collection and network extraction techniques to infer a scientist’s interests and expertise based on their interactions and activities within a Knowledge Network. It implements advanced algorithms to provide network referrals that take into account socio-technical incentives for knowledge sharing. At the User level, CI-KNOW provides tools for Knowledge network recommendations and visualizations Network neighborhood visualization and navigation At the Administrator level CI-KNOW provides portlets for Recommendations Global network visualization/navigation Portfolio management and network diagnostics subsystems What is a Knowledge Network? CI-KNOW Team Noshir Contractor, Hank Green, York Yao Steven Harper, Nat Bulkley, Andy Don Abstract. A Knowledge Network is a multi-dimensional network created from the interactions and interconnections among the people, documents, data, analytic tools, and interactive collaboration spaces (like forums and wikis) associated with a collaborative environment. CI-KNOW is a suite of software tools that leverages automated data collection and social network theories, analysis techniques and algorithms to infer an individual’s interests and expertise based on their interactions and activities within a Knowledge Network. Recent developments in social network theories and methods provide the backbone for a modular system that creates recommendations from relational metadata. The CI-KNOW Recommender mines the Knowledge Network associated with communities’ use of cyberinfrastructure tools and uses relational metadata to record connections among entities in the Knowledge Network. A network navigation portlet allows users to locate colleagues, documents, data or analytic tools in the network and to explore their networks through a visual, step-wise process. An internal auditing portlet offers administrators diagnostics to assess the growth and health of the entire Knowledge Network. The CI-KNOW recommender system is integrated with the CLEANER portal, created by the Environmental Cyberinfrastructure Demonstration Project, which supports the activities of environmental science communities (CLEANER and CUAHSI) under the umbrella of the WATERS network ( http://cleaner.ncsa.uiuc.edu ). It is also integrated with the ToBIG portal, created by the SONIC group to support the activities of tobacco researchers and policy makers ( http://tobig.ncsa.uiuc.edu ). CI-KNOW development is supported primarily by grants from the National Science Foundation and the National Cancer Institute and the National Institutes of Health. When individuals log in and uses a collaborative environment, their behavior is logged within a unique session that is linked to their user ID. Each session stores use logs as Resource Description Framework (RDF) triples that can be called ‘metadata’ Metadata summarizes objects (whether users or items), locates objects in geographical space, indexes content, stores and defines network ties among objects, and records usage and access information. A triple is information stored in the form: <subject><predicate><object> For example : <user 1><is author of><forum post 1> Predicates form the structural linkages between items in the knowledge network . CI-KNOW harvests structural metadata to record connections among entities in the knowledge network A Knowledge Network is a multi-dimensional network created from the interactions and interconnections among people, documents, data, analytic tools, and interactive collaboration spaces (like forums and wikis) associated with a collaborative environment. A Knowledge Network grows organically as users publish, access data, use analysis tools, post to forums, etc. Linkages among items are stored as meta-data, which is accessed by CI-KNOW. The CI-KNOW recommender system mines the Knowledge Network associated with communities’ use of cyberinfrastructure tools like the portal, combining that data with information gathered from other sources like bibliographic databases and patent information. Funding for CI-KNOW development comes from NSF grants BES-0414259, BES-0533513, IIS-0535214, SBE-0555115, and SCI-0525308 and Office of Naval Research grant N00014-04-1-0437. Automated Metadata Collection CI-KNOW Recommender System Tools: Current and in Development The CI-KNOW portlet window “returns” a folded list of recommendations to users while they are working inside the portal. The recommendation itself is an active link to the contact information for the user or to the particular forum post, document, or data that is recommended. Clicking on the “Why” button in the portlet provides more detail about why a user has been given that particular recommendation and provides a text description and a network visualization of the path between the user, the search term and the recommendation. Metadata are classified according to the context from which the data are drawn and the primary descriptive characteristic that the data possess. Data contexts from collaborative environments include, but aren’t limited to: user profiles chat environments forums and solution centers wiki and collaborative editing spaces individual ‘notebooks’ data repositories analysis tool areas libraries of relevant documents external links to data, tools and documents data mined from bibliographic data bases or other sources Knowledge Networks in CI (Cyberinfrastructure) CI-KNOW Metadata Taxonomies Java-based Run on a dedicated Apache Tomcat server Platform independent Uses a web services/API paradigm Calls <.25 second for current system settings, calculation dependent (based on table lookup and sorting) Scalable to 10 million links and 1 million nodes with no increase in call time CI-KNOW Technical Specifications CI-KNOW tools are being developed as independent web applications that can appear as portlets within a portal or collaborative environment. An intuitive user-interface enable members of the community to use the network referral system. To make the system sustainable, CI-KNOW continuously retrieves data from the community, processes these data, and uses internal network referral algorithms to update its network referrals for use within the community. The network navigation tools allow users to locate themselves or their colleagues, documents or workspaces in the knowledge network and to explore their networks through a visual, step-wise process. Users will have the option to explore these networks in an optimal ‘topological’ space, or to have the networks connected to geographical maps through an approach similar to the one used by Google Maps A Hybrid, Modular Recommender System Traditional referral systems do not take into account: (i) multiple relationships that link Knowledge Network entities; (ii) the multiple socio-technical incentives that influence the effectiveness of a network-based recommendation. Recent developments in Social Network theories and methods provide ways to fill in this gap. CI-KNOW recommendations emerge from a modular system that allows for the combination and hybridization of multiple mechanisms for creating recommendations from structural metadata. Modules are based on Social Network Analysis Theory and can be updated at any time. Current and Planned Modules Structural similarity of users and items to search terms: what is most alike? Structural closeness of users and items to search terms what is closest? Popularity of an item: what is conventional? Structural influence: Who or what are key links in the network? Innovation: what’s new? Burst activities: what’s ‘hot’? Exchange: who you previously helped and now “owes you one”? Proximity: who is geographically near you? Users can specify preferences and settings for how they use the CI-KNOW system and what they share with others who use the system. Module importance is user-determined; they determine which modules should contribute to the total recommendation function and how much. Users can determine what sorts of recommendation results they will see and modify the type and quantity of recommendations they receive. For ‘Push’ recommendations: CI-KNOW relies on user preferences and inferred search terms to provide pro-active recommendations For ‘Pull’ recommendations: users can specify preferences and search terms for CI-KNOW in real-time, with the capability to set default specifications User Modified Rather than providing traditional search results (such as documents that contain a specific keyword), users receive recommendations to entities (datasets, documents, people) that are associated through transactions, interactions, and interconnections with items that contain the keyword. A user enters a word or phrase into the ‘Recommendation’ search box. Those items that contain the word or phrase become the ‘seeds’ for CI-KNOW CI-KNOW uses relational metadata to crawl outward from the seeds, recording which entities in the knowledge network are connected directly or indirectly to these seeds Recommendations are based on algorithms that take into account the structural paths to relevant entities as well as the social motivations and incentives for knowledge sharing How CI-KNOW Searching Works Internal Modifications CI-KNOW Tools in Development Concerns Recommendation quality and validity must be investigated. Test cases provide insight User feedback for live recommendations User tracking Privacy/Access/Permissions policies must be set Access to documents, contact information, and usage tracking determined by owners Visualization of users, documents and data in KN determined by owners Administrative subsystem for searching the KN and reporting of a standard set of network diagnostics for knowledge portfolio management Instantiations of CI-KNOW Waters Network Portal CyberIntegrator ToBIG Portal The ‘Tell Me Why’ page presents information about the shortest path between the user, the search term and the recommendation. The ‘Tell Me Why’ visualization shows the shortest path between the user, the search term and the recommendation. With one click, the user can show all the shortest paths, hide node labels, and access more tools for Knowledge Network exploration

3D Strategy for Enhancing CoP Networks D iscovery: Effectively and efficiently foster network links from people to other people, knowledge, and artifacts (data sets/streams, analytic tools, visualization tools, documents, etc.). “If only we knew what we knew.” D iagnosis: Assess the “health” of CoP’s internal and external networks - in terms of scanning, absorptive capacity, diffusion, robustness, and vulnerability to external environment D esign or re-wire networks using social and organizational incentives (based on social network research) and network referral systems to enhance evolving and mature communities.

Mapping Flows in the PackEdge CoP Network

PackEdge CoP Vital Statistics Exploration Scanning : Access to expertise external to CoP Absorption : Ability to absorb expertise external to CoP Vulnerability : Brokered by members external to CoP Exploitation Diffusion : Ability to diffuse expertise throughout CoP Robustness : Not relying on few critical CoP members to keep things together

“ Pre-wired” PackEdge CoP Network

“ Re-wired” PackEdge CoP Network

Wiring the PackEdge CoP Network for Success Increase the likelihood to give and get information to the right target and source respectively Benefits for CoP Increase absorptive capacity from 45.3% to 53.4% Reduce number of steps for diffusion from 4.3 to 2.6 Costs for CoP Increase communication links of network leaders from 28 to 38 (~ 150 new links). Increase criticality of network leaders from 26.7 % to 48.5%

Core Research Social Drivers for Creating & Sustaining Communities Business Applications PackEdge Community of Practice (P&G) Vodafone-Ericsson “Club” for virtual supply chain management (Vodafone) Societal Justice Applications Cultural & Networks Assets In Immigrant Communities (Rockefeller Program on Culture & Creativity) Economic Resilience NGO Community (Rockefeller Program on Working Communities) Entertainment Applications World of Warcraft (NSF) Everquest (NSF, Sony Online Entertainment) Science Applications CLEANER: Collaborative Large Engineering & Analysis Network for Environmental Research (NSF) : Collaboration for Preparedness, Response & Recovery (NSF) TSEEN: Tobacco Surveillance Evaluation & Epidemiology Network (NSF, NIH, CDC) Projects Investigating Social Drivers for Communities

WoW: Massively Multiplayer Online Role Playing Game

Source: http://www.mmogchart.com/ Rise of WoW

Goals of WoW Community Teams perform diverse quests within the game environment, typically varying in length from one hour to several days, with the goal of achieving an objective, gaining resources, and increasing experience. Exploiting, Bonding & Swarming

Mapping Goals to Theories: WoW Gaming Community

Data Collection Data were collected from all 184 individuals who belonged to 16 guilds at 3 points in time. T1: initial contact and survey administration T2: two weeks after initial survey administration T3: four weeks after initial survey administration Demographic information: Gender: Female members (21.7%) Male members (78.3%) Ethnicity: Caucasian (79.3%) Asian/ Pacific Islander (15.2%) African American (2.2%) Hispanic/Latino (1.1%) Native American (1.1%)

Information Retrieval - Time One

Information Retrieval - Time Two

Information Retrieval -Time Three

Density of Communication Ties Decreases over Time

Unraveling the “Structural Signatures” Incentive for creating a WoW link with someone = -1.08 (cost of creating a link) [Self-interest] + 0.29 (benefit of reciprocating) [Exchange] + 3.07 (benefit for being a friend of a friend) [Balance] + 0.04 (benefit of connecting to an expert) [Cognition] All coefficients significant at 0.05 level

Summary Theories about the social motivations for creating, maintaining, dissolving and re-creating social network ties in multidimensional networks Development of cyberinfrastructure/Web 2.0 provide the technological capability to capture relational metadata needed to more effectively understand (and enable) communities. Computational modeling techniques to model network dynamics in large-scale multi-agent systems Exponential random graph modeling techniques to empirically validate the local structural signatures that explain emergent global network properties

SONIC Research Team Members Steven Harper Postdoctoral Research Associate NCSA, UIUC Hank Green Research Scientist NCSA, UIUC Chunke Su Graduate Research Assistant Speech Communication, UIUC Andy Don Research Programmer NCSA, UIUC Annie Wang Graduate Research Assistant Speech Communication, UIUC Mengxiao Zhu Graduate Research Assistant Speech Communication, UIUC Nat Bulkley Postdoctoral Research Associate NCSA, UIUC York Yao Research Programmer NCSA, UIUC Diana Jimeno-Ingrum Graduate Research Assistant Labor & Industrial Relations, UIUC

Open Grid Forum workshop on Social Networks, Semantic Grids and Web

More Related Content

What's hot

Similar to Open Grid Forum workshop on Social Networks, Semantic Grids and Web

Recently uploaded

Open Grid Forum workshop on Social Networks, Semantic Grids and Web