SOCIAM Social Machines and Linked Data
Dominic DiFranzo Missing Piece
e-infrastructure
Big Data The Future! The Fourth Big Compute Quadrant
Conven onal Social online Computa on Networking R&D More machines
More people e-infrastructure
Big Data The Future! Social Big Compute Machines
Conven onal Social online Computa on Networking R&D More machines
More people The Order of Social Machines
Real life is and must be full of all kinds of social constraint – the very processes from which society arises. Computers can help if we use them to create abstract social machines on the Web: processes in which the people do the crea ve work and the machine does the administra on… The stage is set for an evolu onary growth of new social engines.
Berners-Lee, Weaving the Web, 1999 A Big Picture
Machines using people e.g. turk People using machines e.g. collec ve ac on More machines
More people Social Machines the New Fron er An Example
The Kenyan elec on on the 27th December 2007 resulted in a wave of riots, killings and turmoil…
African blogger Erik Hersman read a post by another blogger Ory Okolloh…
Birth of
“Nobody Knows Everything, but Everyone Knows Something.”
Local observers submit reports using the Web or SMS
Social Machine Variations
D.C. "Snowmageddon" Fukushima Tsunami Port au Prince Earthquake
Arab Spring Obama '08 Nigel Shadbolt et al Social Machines are NOT Turing Machines
• they do contain conven onal algorithmic components • but much else is different • a social machine will start with an incomplete specifica on that grows and evolves to cover more of the problem via interac on • a social machine achieves par cipa on through local incen ves which become reinforced as the… • incen ve for an individual to supply data to the algorithm increases as more individuals par cipate • a social machine has a no on of completeness that is a social rather than mathema cal issue • a social machine will not usually have a no on of the correct output or termina on… rather it runs con nuously The dimensions of Social Machines – Social Machines vary depending on
• Number of people • Empowering of • Number of machines individuals, groups or crowds • Scale of data • • Varie es of data Time cri cality • • Type of machine problem Extent of wide area solving communica on • • Type of human problem Need for urgent solving mobiliza on • Specifica on of goal state Lead Investigators
Principal Investigator 5 years 2012-17 Nigel Shadbolt
Co-Investigators EP/J017728/1 Wendy Hall Tim Berners-Lee mc schraefel Luc Moreau
David De Roure
David Robertson Peter Buneman
What will SOCIAM do Theme 6 Web Observatory
• Understand Social Machines through an observatory that observes, monitors and classifies social machines - both those of the project and more widely on the Web - as they evolve; • it will also act as an early warning facility for new disrup ve social machines elsewhere on the Web; • to understand how Social Machines reach pping points, longitudinal observa onal data will reveal how they grow once launched; • whether they coalesce into larger machines or fragment into micro machines that s ll have u lity; • what signals need to be observed, what is a fair and faithful sample of Web behaviour; • this is likely to call a en on to appropriate governance, ethical and legal issues. Web as lens
Web as ar fact
Web Observatories h p://www.w3.org/community/webobservatory/ An observatory is sociotechnical
Technical Social • Mul ple observatories • Researchers • Datasets and dataflows • Managers • Analy cs • Shared artefacts • Tools and so ware (intermediate results, • Automa on annota on) • No fica on (real me) • Shared methods • Provenance • Collabora ve analysis • Secure • Policy • Linkable • Responsible innova on • Programma c access • Training • Visualisa on (eyepiece?) • Standards!
Narra ve Database
VS
• Cause and Effect • Unordered List of • Connects seemingly Items unordered events • Organized for fast together search and retrieval
Narra ve Database
Linked Data Enhancements
Saturday, December 4, 2010
Anatomy of an observatory Plug and Play Query Subscribe analy cs
ongoing collec on
“Non-Consump ve Research” Linked Social Data
• Take everything we learned with LOGD – Provenance/metadata/Linking/Lightweight • Work Like the Web – AAA Principle – Many Truths – Open World Assump on – Bo om Up Bring Back Context to Data
• The events seen in our Big datasets never happen on their own, but are linked to related events. • By linking Large Heterogeneous Data together, we can help capture the larger context of the data we care about. • Allows us to test the correla ons and possible causa ons of the phenomenon we see in our data Bring Context Back
• For exmaple: – In our large video game dataset, we see no ac vity from people playing in Kansas in Early December 2007. Why? – Could it be that a large ice storm in December 6 had a huge power outage from the 6-12th in that area? – Linking to power outages dataset from Wikipedia(h p://en.wikipedia.org/wiki/List_of_major_power_outages) could show us that. Recording of Provenance
• One of the largest difficul es of the social science is reproducibility of experiments. • Recording of provenance of the workflows of how experiments were done (what models, what dataset, what analy cs, etc) could help solve this
Recording of Provenance
• With the more subjec ve and interpre ve nature of Linked Data, it becomes vital that we know WHO said WHAT. • We need to record the authors of Data to know how we should use it, and how it should be accepted. Plug and play observatories Technical and business interface Current Work
• Conver ng SOCIAM data using PRIZMS – Both large Qual and Quant data (Twi er/ Zooniverse) • Leverage our twi er ontology and tools to SOCIAM twi er data. Test how reusable our tools and demos are on other large scale twi er data • Building and collec ng analy cal tools for both Qualita ve and Quan ta ve researchers.