The Intel Cycle in Practice

1
The Intel Cycle in Practice
Intel infers needs
Ops receives reacts
Task Collection
?
Dissemination
Raw intel
Treverton 2001
Analysis Production
? Intel Systems operate in adhocracy they must
support analysts opportunistic use of
never-quite-structured data
2
Proteus Interface Concepts

• Proteus will be a general platform for gathering
  any kind of data together and visually exploring
  semantic, geospatial, and temporal relationships.
  The first two important things that it needs to
  be able to do are
• easily and repeatedly import new data sources and
  formats into the database
• provide an integrated environment where analysts
  can easily pull the thread to search for and
  display entities by text query, semantic
  relation, temporal sequence, or geospatial
  position, and repeat the operation on the results
  in the same application they are displayed.
• Getting this right will lay the groundwork for
  more advanced capabilities like frame queries,
  anomaly-detection, combinatorial analysis,
  consistency-checking, etc. Further analyst
  assistance in the forms of tools to structure
  investigation should also be provided. But first
  it must be possible to bring everything into one
  place and then rapidly search and display it
  semantically, geospatially, and temporally.

3
Pulling the thread on semantic, geographical,
and temporal relationships

• From any of various GIS, Graph, Browser, Query,
  Table applications, provide intuitive and
  standardized capability to
• Select a set of entities to start with
• Select types of related entities to look for
• Look for entities that are related by
• Semantic (graph) relationship
• explicitly stored DB relationship
• possible relationship due to document
  co-occurrence
• Collocation within geographic parameters
  (indicated by drawing circle, rectangle, or
  route)
• Temporal co-occurrence
• Display results in
• GIS
• Semantic Graph
• Table
• Repeat

4
GIS App (Google Earth, FalconView, LLNL open
source)
(right click for query options form)
Select types of entities to plot, adding a GIS
layer
Select an area (rectangle or circle or draw a
route)
Plot semantic links among displayed entities,
adding layer
plot semantic links from displayed entities,
adding layer
5
Plot (certain types of) entities active at in the
same time as selected entities
Graph entities in semantic network to show
relations to abstract entitles or unlocatable
entities
Plot locatable entities selected from a semantic
graph
6
Sort, Edit, Select subset, and plot in GIS, or
display relationships in semantic graph
Perform complex query, or use a saved query
Select subset, and display in sortable, editable
table
7
If entity extractor supports training,
correction, allow analyst to do this on the fly
Select an entity or relationship, and display
sources in a browser. If item was automatically
extracted, display markup
Inserts, Deletes, Edits, coordinated across
applications, with big Undo button!
8
Cognitive Organizational Bias(consider when
generating/evaluating hypotheses/evidence)

• Belief Fixity
• Assimilate new data to existing assumptions
• Confirmation Bias
• Ignore/discount disconfirming evidence
• Incremental Interpretation
• More data ? more confidence, not necessarily more
  accuracy
• Satisficing
• First available answer/analogy, rather than
  systematic search evaluation
• Framing (Prospect)
• Presentation influences judgment
• False confidence
• Im an expert, nothing surprises me
• Attribution Error
• Their behavior is coordinated, centralized,
  intentional
• Our effort ? their good behavior butTheir
  character ? their bad behavior
• Mirror Imaging/Ethnocentrism
• They think and act like us
• They know what we know
• We understand each others intentions

• Consistency Bias
• If we change our assessment well lose
  credibility
• Selection Bias
• The best info is classified (ignore opn src)
• Readily available or exclusive sources
• Consensus
• Groupthink
• Interagency bargaining
• Politicization
• Intel to please
• Parochialism
• Policy advocacy
• Complexity
• Sharing, Communication
• Circular Reporting
• Time Constraints
• Short Term Production/time horizon
• Measures of Effectiveness
• Quantity over Quality

Jervis 1976, Herman 1995, Heuer 1999, Lowenthal
2000, Johnston 2005
9
Tools to structure analysis

• Refine reformulate question
• Brainstorm evidence, alternatives
• Identify factors/variables, potential causal
  relations, feedbacks (arrow/system diagram)
• Sort/group factors/evidence/data from varying
  perspectives (list, outline, or bivariate matrix)
• Multi-factor weighted pair-wise ranking
• Pros, cons, alternatives
• Analysis of competing hypotheses matrix devils
  advocacy
• Chronological alternative/decision trees
• Expected value analysis (assess probability
  utility represent variables/outcomes/options in
  matrix or tree)

Jones 1998
10
Causal Diagrams
Somalisstarving

• Identify observable factors
• Label direct () and inverse (-) causal
  relationships (avoid further quantification)
• Identify unstable (U) and self-sustaining (SS)
  feedbacks(Loops with all direct causation are
  unstable two coupled inverse arrows add up to
  one direct arrow)
• Analyze sensitivity to adding removing
  variables, redrawing reweighting influences
• Each arrow and circuit is a hypothesis that can
  be evaluated in ACH

SS
Media Coverage
Warlord Profiteering
-
U


Foreign assistance
-
U

Foreign aid workers
US Military Presence
-
SS


Warlords attacking foreigners
-
U

US Military casualties
Jones 1998
11
Analysis of Competing Hypotheses (ACH)

• Generate hypotheses
• List observable evidence, pro con, for each
  hypothesis (list known, historical and potential
  future evidence)
• Assess strength of each item(is it a certain
  and/or unique prediction of each hypothesis?)
• Refine analyze hypotheses evidence
  matrix(order evidence chronologically attempt
  to disprove hypotheses)
• Sensitivity assessment (what if specific evidence
  is wrong, misleading, misinterpreted?)

Heuer 1999
12
Strong v. Weak Evidence
Certain (Explanation definitely predicts)
Yes
No
Yes
Decisive
Smoking Gun
Unique (no other explanation)
Circumstantial
Hoop Test
No
Van Evera 1999
13

• Herman, Michael. 1996. Intelligence Power in
  Peace and War. Cambridge, UK Cambridge
  University Press.
• Heuer, Richards J. 1999. Psychology of
  Intelligence Analysis. Washington, DC Center For
  the Study of Intelligence.
• Jervis, Robert. 1976. Perception and
  Misperception in International Politics.
  Princeton, NJ Princeton University Press.
• Johnston, Rob. 2005. Analytic Culture in the US
  Intelligence Community an Ethnographic Study.
  Washington, DC Center For the Study of
  Intelligence.
• Jones, Morgan D. 1998. The Thinker's Toolkit 14
  Powerful Techniques For Problem Solving. New
  York, NY Three Rivers Press.
• Lowenthal, Mark M. 2000. Intelligence From
  Secrets to Policy. Washington DC CQPress.
• Treverton, Gregory F. 2001. Reshaping National
  Intelligence in an Age of Information. Cambridge,
  UK Cambridge University Press.
• Van Evera, Stephen W. 1997. Guide to Methods For
  Students of Political Science. Ithaca, NY
  Cornell University Press.