Christoph Barrett - Policy Informatics at Societal Scale
0 likes2,544 views
This document discusses policy informatics at a societal scale for massively interactive socially-coupled systems. It argues that policy modeling must be responsive to evidence-based policymaking and end simplistic prediction models. With embedded computing and social networks, individual reasoning is socially influenced. It proposes using synthetic data and simulations of interacting individuals and infrastructure to model these complex systems, while preserving privacy. This would involve generating synthetic populations and networks that represent regional interactions and integrate with built systems.
Christoph Barrett - Policy Informatics at Societal Scale
- 1. Network Dynamics & Simulation Science Laboratory Policy Informatics at Societal Scale: Massively Interactive Socially-Coupled Systems Christopher L. Barrett Scientific Director Virginia Bioinformatics Institute Virginia Tech
- 2. Network Dynamics & Simulation Science Laboratory Policy informatics • Must be responsive to actual evidential policy making – Principles, standards, objectives – Processes and procedures – Measurement of performance wrt objectives • End of monolithic models of these complex systems • End of simplistic ideas about prediction • End of the “great man theories” of decision making – Sociality in reasoning processes is not an abstraction now • Embedded, pervasive computing and information networks • Costs drivers have shifted from data to analytics
- 3. Network Dynamics & Simulation Science Laboratory PI in complex systems essentially change them • Co-evolution and branching are at the heart of the real world of big data – Margin trading example of co-evolution – “Arbitrage law” drives branching – Taditional data sources, no matter how big will always be “measure zero” • Viable ICT approaches: – replace positivist prediction paradigms with abductive, counter factual/fictive, evidence-driven, systems – are inherently privacy preserving – Delivers problems to computing and deploys pervasive computing
- 4. Network Dynamics & Simulation Science Laboratory Is this necessary? Is it possible to “package”? • We’ll look at: – How to think about this – Tools, methods, resources – Rationale • What is the necessary R&D program? – What are the theoretical and practical issues? • Relevance to policy making problems and organizations
- 5. Network Dynamics & Simulation Science Laboratory Intro to Synthetic Socio-technical information • Start by using surveys and other individual information
- 6. Network Dynamics & Simulation Science Laboratory Add relevant individual behavior • Attached to every synthetic individual • Connects individual properties to plans and to joint plan information
- 7. Network Dynamics & Simulation Science Laboratory Project onto activity locations (geographic & virtual) Edge labels • activity type: shop, work, school • (start time 1, end time 1) • (start time 2, end time 2) Location Vertex: • (x,y,z) • land use . • Business type People Vertex: • age • household size • gender • income .. • Demographically match schedules • Assign appropriate locations by activity and distance • Determine duration of interaction • Generate social network
- 8. Network Dynamics & Simulation Science Laboratory Produce synthetic data libraries & networks • “Megapolitan” Regional networks • Interaction with built socio- technical infrastructures • Methodology Advances – Software scale to national scope – Graph library to calculate graph measures of large networks Simulate Composed Interactions
- 9. Network Dynamics & Simulation Science Laboratory Sources of information • Social media sources • Existing and new crowd sourcing & embedded pervasive sources • Micro surveys • Aggregators • Conventional sources • Enterprise information • Biological information in detail • Medical information…….etc
- 10. Network Dynamics & Simulation Science Laboratory Properties of synthetic information • Synthetic information is inherently: – Privacy preserving, yet – Extremely granular – Very large – Dynamic – Customizable by product lines – Reusable and modifiable • HPC and pervasive computation-oriented – Changes how HPC must be delivered – Emphasizes data services and synthesis, not modeled prediction
- 11. Network Dynamics & Simulation Science Laboratory Tools and Methods •
- 12. Network Dynamics & Simulation Science Laboratory Synthetic information environments: Big data synthesizers creates and enables
- 13. Network Dynamics & Simulation Science Laboratory User & context–driven Structured and Unstructured Data Sources in the context of a query…
- 14. Network Dynamics & Simulation Science Laboratory Overview Structured and Unstructured Data Sources and transforms them…
- 15. Network Dynamics & Simulation Science Laboratory Very large synthetic information libraries Structured and Unstructured Data Sources…into
- 16. Network Dynamics & Simulation Science Laboratory Example: Train a “reach back” response system • Use decision analytics platform and crowd source interface to create training environment – Stakeholder integration – Complex scenario – Diverse component interactions with user – Maintain non-specialist, application focus – Use leading edge HPC and pervasive computing tools and methods • This is an introductory movie for the students
- 17. Network Dynamics & Simulation Science Laboratory
- 18. Network Dynamics & Simulation Science Laboratory Motivation: Large scale interaction problems
- 19. Network Dynamics & Simulation Science Laboratory Individual behaviors and populations • Socially-coupled systems involve people, their behaviors and their environments • They co-evolve and branch • Behavior is structured by individual biological state, cognitive state, individual motivations, perception and situational reasoning, economic and social reasoning, strategies and plans, technological and environmental properties, functionalities and constraints, etc • What matters?
- 20. Network Dynamics & Simulation Science Laboratory Consider what is involved in urban mobility
- 21. Network Dynamics & Simulation Science Laboratory How socially-coupled systems compose
- 22. Network Dynamics & Simulation Science Laboratory Composition: Wireless interference among vehicles
- 23. Network Dynamics & Simulation Science Laboratory The size of the problem: person to country • From individuals: their state, motivations, activities and • From locations: their functionalities, services, constraints, supply chains, etc • Individuals and related groups are defined • Order 107 to roughly order 1010 interacting elements (now)
- 24. Network Dynamics & Simulation Science Laboratory
- 25. Network Dynamics & Simulation Science Laboratory
- 26. Network Dynamics & Simulation Science Laboratory
- 27. Network Dynamics & Simulation Science Laboratory
- 28. Network Dynamics & Simulation Science Laboratory
- 29. Network Dynamics & Simulation Science Laboratory
- 30. Network Dynamics & Simulation Science Laboratory Composed dynamics and behavior: disease, individuals, populations, interregional travel, health care system
- 31. Network Dynamics & Simulation Science Laboratory Distribution of day of first arrival of disease
- 32. Network Dynamics & Simulation Science Laboratory Reporting of Adenovirus variant
- 33. Network Dynamics & Simulation Science Laboratory Day of overwhelmed hospital treatment resources
- 34. Network Dynamics & Simulation Science Laboratory Infrastructure catastrophe example
- 35. Network Dynamics & Simulation Science Laboratory Physical disaster in a social context • Event put “on top of” a normally functioning day’s population dynamics • National Planning Scenario 1 • Unannounced detonation • Time: 11:15 EDT • Date: May 15, 2006
- 36. Network Dynamics & Simulation Science Laboratory Damage to power network and long term power outage area • Probability of damage to individual substations • / / : High/medium/low: probability of damage Aggregated outage area • Long-term outage area devised by geographically relating the location of substations in the city with the blast damage zones. • Loss of a substation has a much more widespread impact on provided power to the customers. Time 0:00
- 37. Network Dynamics & Simulation Science Laboratory Infrastructure: initial laydown • Positions and demographic identities of individual synthetic people in the DC region were calculated at the time of detonation. • Street addresses mapped to geo-functional data • Persons traveling to destinations were placed outside on transportation networks –walk, roadway, metro, bus. • Power outage, damage, collapse, rubble, blast temp, radiation dose rate assigned to each location and transportation network node Built Infrastructure Power Outages Position of People Time 0:00
- 38. Network Dynamics & Simulation Science Laboratory Building Collapse DistributionTime 0:00
- 39. Network Dynamics & Simulation Science Laboratory Damage to transportation networks • Red: completely damaged • Orange: highly damage; reduced travel speed • Green: medium damage • Blue: light damage • White: No damage Walk network Road Time 0:00
- 40. Network Dynamics & Simulation Science Laboratory No communication – green Partial Communication Restoration – Blue First 29 hours Social-behavioral Event in a Physical Context
- 41. Network Dynamics & Simulation Science Laboratory Composite behavior differences w & w/o early restored comms
- 42. Network Dynamics & Simulation Science Laboratory Aggregate behavioral details & exposure to injury • Each individuals' daily or event context- driven activities take them inside and outside periodically, the details affect their injury level at the time of, as well as after, the blast. • Injury traversing rubble • Delay of access to care, etc Outdoors Indoors
- 43. Network Dynamics & Simulation Science Laboratory Transportation load comparison Time +0:00 to +0:10 Blue - Higher load in No Restoration case Purple - Higher load in Partial Restoration case
- 44. Network Dynamics & Simulation Science Laboratory Interdependent, contextual, intentional individual avatar behaviors induce social level effects w/o scripting
- 45. Network Dynamics & Simulation Science Laboratory A drama in machine intelligence: Reuniting a family after the disaster Cliff • Father • +0:00 - At work • Uninjured Clair and Denise • Mother and infant daughter • +0:00 - Home • Both uninjured Theo • Son • +0:00 Daycare • Uninjured
- 46. Network Dynamics & Simulation Science Laboratory Initial Panic Cliff • +0:00 – Panics, abandon’s car, heads to nearest hospital • Exposed to 0.4cGy first 50 minutes Clair and Denise • +0:00 – Shelter at home • Repeatedly calls 911 • Both exposed to 10cGy first 10 minutes Theo • +0:10 – Workers bring children to nearby building for shelter • No exposure
- 47. Network Dynamics & Simulation Science Laboratory Calls finally go through Cliff • +3:00 – Call to Clair successful • Stops panicking and finds shelter • +3:10 – Call to Theo (i.e., daycare worker) successful Clair and Denise • +3:05 - Evacuate City • Doesn’t know where Theo is Theo • Continues shelter in Daycare
- 48. Network Dynamics & Simulation Science Laboratory Family Reconstitution Cliff • 44:30 Leaves shelter Theo • Remains at daycare
- 49. Network Dynamics & Simulation Science Laboratory Evacuation Cliff • +45:00 – Arrives at daycare • Evacuates city with Theo
- 50. Network Dynamics & Simulation Science Laboratory Data Intensive Computing Resources Compute TimeModule Wall Time Compute Time Transportation 13.75 hr 8911 hr 648 cores Behavior 3.92 hr 397 hr 96 cores Communication 9.53 hr 9.53 hr Health 4.3 hr 4.3 hr Infrastructure 1.4 hr 1.4 hr Data Initial Dynamic (1 run) Complete Design (20 cells, 30 replicates) 2M individuals, 2 weeks, full design Database 3.55 GB 27 GB 25TB 250TB Disk 1.16 GB 15 GB 20TB 175TB *Summary over all iterations r1413
- 51. Network Dynamics & Simulation Science Laboratory thanks