KCSB Group Kickoff

RER Rapid Ethnographic Retrieval will pull out data with data fusers to create a searchable database for analysis. Using HRAF, background info on nature of beliefs, conflict, compendium of thesauri to help do the coding. Human in the loop (coded by hand) to train and inform the computers. Thesauri become codebooks. Simulation tools will be driven by data coming out of ORA to make the tools reusable.

Note to self: Huan Liu's tag clouds and blog influence algorithms.





What about networks?



Not just actor-centered networks, multi-modal networks!





Time and Geospatial analysis will be expressly addressed.





META NETWORK IS THE FOCUS. Data or modeling techniques that can't use networks will be ignored.





Integrated System for Adaptive Adversarial Reasoning. ISAAR is the term for the system itself.





From many data sources to formal reasoning





Different data have different qualities, pedigree (quality), etc. How to take this into account?





Development of new pre-processors that can handle metadata. Also post-processers that can handle geospatial information.





Rapid ethnographic retrieval process will preprocess that will develop material for ORA that will generate reports and create metanetwork. Metanetwork data will post process to create ENHANCED metanetwork (drawing from cultural area, gazateer, DIME / PMESII material, put that all through ORA.





NEW KINDS OF SIMULATORS inc. game theory, strategic reasoning, spatio-cultural mapping and other material for COA analysis and other visualizations and analyses.





Visualizations can assist in change detection. Project NOT about visualization but is leveraging previous visualization research for cultural assessment.



In adversarial modeling there is no compendium of tasks! Many challenges to detecting deception and activity modeling; use of network approach for simulation.



Collaborative co-training between institutions.



Understanding Sudan (RL)



Largest country in Africa, almost one million sq. miles (Maine to Florida)

Huge climatic variety

Border nine countries (almost all strategic to US interests)

Borders four oil-producing countries

None of Sudan oil is coming to the US (potential for 1M barrels a day)

Relationship with China very strong

Strategic location on Nile



(Ethiopia--land of the burnt faces)



Modern Africa created by conference in which no Africans were present (arbitrary borders)

Boundaries make no sense; conflict from divisions of some people, arbitrary colliding of peoples who don't want to be together.



Sudan (and Darfur in particularly) is multi-ethnic. Begara (camel-back wandering people), Arabs predominate in Darfur. Multiple varieties of Islam in Darfur.



Other big fault line is the south. Has some Muslims, but mostly Christians and animists. Has all the oil but the pipeline goes to the north.



Election in 2009 coming up. Will be a major turning point. Referendum in 2011 in which Southernists will decide whether or not to succede; popular will is resoundingly for succession. Elites are trying to hold it together.



Piracy in Somalia resulting in a major arms race over major amounts of oil with China in the thick of it!



The centrality of Islam is a fundamental foundation for understanding Sudan. It is the ideological glue that has held governments together, or conversely, caused them to fall apart. Former home of OBL. Sufi traditions also important. Boundary between Sunni and Shia is not hermetically sealed. Sudan very closely related to IRAN as well as China, politically, militarily.

Sudan now self sufficient in arms, including tanks, every manner of small arms. Exception is aircraft. Chinese are trying to handle that.



Sudan a "sort of" Arab country (member of Arab league, multi-cultural, multi-sectarian, multi-faceted, more pyramids than Eqypt.) Has substantial Christian population (Lord's Resistance Army in the South). NO diplomatic relations with Israel.



Khaddafi as a volatile factor in Sudan. Changable, critical factor.



Darfur presented as black/white but there are some significant shades of grey in play.



Refugee aspect is exceedingly complex. We find Southerners who have never been to the South, Westerners who haven't been to the West. The complexity of this situation called for a nuanced approach that takes into account cross-border relationships and the role of China.



Need to make a connected network of all these factors. All of the circles can be filled with data.



Northwestern University Crosstabs became SPSS. Need to find configuration of these variables to develop smart approaches to address interventions or to develop approaches to stay out of conflicts in which the US has no real interest (or no real ability to intercede effectively.)



Rapid Ethnographic Retrieval JJ

Humans good at tagging and coding ethnographic texts, but they are SLOW. Extremely slow.



Identifying subject clusters: use a sample of things to identify clusters to develop methods to speed up coding.



Example is HRAF but any context is possible.



Terms are embedded in other terms . Development of the "node-list" approach that created a two mode relationship (culture, term). Matrix with weighted edges used to create visualization, SNA exploration.



Problem with HRAF is it's not in the form that can be used for this method. Text prep for 20 cultures on one term (conflict) took 2 days.



Network as "tag cloud" method.





Technology Being Developed



Problem of human data input, far too labor intensive



Development of DYNETML 2.0



Semantic network a key addition to the DYNETML 2.0

The automatic of a "metanetwork thesauri set" to enhance pre and post processing is another improvement

This is then fed into ORA for new kinds of reports



Key needs

Increase speed of processing

Greater control over components

Improved thesauri creation/handling

Emotion tracking (sentiment tracking)

Triad extraction

Rule extraction

Arabic translation



DYNEtmL Change Rationale



Reliability and Pedigree

Delays in positive ID of unknown nodes

Nodes and edges (relations) that may appear, disappear, change in strength

Multi-nodal, multi link and multi-level data



EER: Enhanced Edge Representation

ENR: Enhanced Node Representation

EMR: Enhanced Meta-Data Representation



Five types of data:

String

Date

Numeric

Category-Ordered (income, low to high)

Category-Non-ordered (nationality)



Sudan Studies Org



DB's LLNL SNA of leadership might be of use (Glen?)



Change, Deception and Strategy



KC, IM



Complex interactions have impact on the agent's transactive memory

Network dynamc is the beh of the network over time

Network over time: they can evolve plus do other things

- Stability (modeled with link probability)--rather rare

- Evolutionary change

- modeled with multi-agent simulation

- modeled with MCMC

- Random chain

- Understanding network dynamics of a pop is critical for accurate inference

- Require more and better tools than one-off models



Need to detect change very quickly with few number of observations



Periodicity (Impact)

Collect email

Chunk data

Calculate graph

Fourier: you can create a sine wave from such data to identify periodicity (Fisher's test)

Drop out noise. Inverse Fourier form.

Subtract out periodicity

What's left shows periods of significant change



Detecting Change in Action

Focus on Network based actions

Currently only applied to social network;

PLANNED: application to non-social networks



Bluffing, stealth, baiting, and misrepresentation



Key to theory building is lots of virtual experimentation

Red teaming;



NEED for high performance computing!



Strategic Adversarial Modeling



Need to improve cognitive models



Don't want tradition cog models like ACT-R or SOAR

Make social agents more cognitively realistic, more strategic.



CONSTRUCT 4.0



Have knowledge, beliefs about others, beliefs are a function of knowledge and influence. Socio-demographics and social network are inter-correlated.



Butterfly effect: creation of cascades of change based on shifts in knowledge and/or influence, socio-demographics, networks



Creation of a virtual city (and initialize actors, social networks)

(No birth module, not used for long term sims)



Rapid Re-Use--Construct can be called from ORA.



Toy use = front end; serious use=back end



Adversarial Agent Models



Political capital: network and non-network component



Enriched EGT.

Visualization of Belief Systems (VIBES)
ORA Software under the hood
At the Pentagon

Try to replace VIBES with integrated tool chain we're developing now
And on into IBAC
Integrated Behavioral Assessment Common User Interface
Sits on top of Internet, will use VIBES as one of its component tools

Sudan to Afghanistan
Spiral deliv

Collecting info on human socio cultural models, put into THE BRAIN via a questionnaire.JFCOM CATALYST. Army Research Lab. Jeff Hansburger. Data will be available to EVERYONE.

Ask Sue about Bob Fost'er's integrated architecture for all HCSB models.

Mega model of Sudan? (RL talked about it)--applied metrics.

NRL DIME / PMESII requirement board with specs due in February.

More Notes

Can you integrate models by simple data passing?

But not when they are relatively far apart in theory or context.

Granularity, domain big issues of context.

How do we deal with the critical factors of decision making? Current motivations? Sentiment?

For strings, can we build taxonomies that could grow, evolve?

3 ontologies: individual-level; enviromental-level (inc. politics, geography, economic) ; If you're interested in their activities, this is another rich set of taxonomies (such as WMD)

ACT R SOAR are task based models; social models are activity based (higher level of abstraction, temporality)

connections between terrorists and crime