The Challenge of Cultural The Challenge of Cultural Modeling for - - PowerPoint PPT Presentation

▶
the challenge of cultural the challenge of cultural
SMART_READER_LITE
LIVE PREVIEW

The Challenge of Cultural The Challenge of Cultural Modeling for - - PowerPoint PPT Presentation

Feb 25, 2023 253 likes •812 views

WI/IAT Joint Keynote, Silicon Valley, USA, 2007 The Challenge of Cultural The Challenge of Cultural Modeling for Inferring Modeling for Inferring Intentions and Behavior Intentions and Behavior Eugene Santos Jr. Eugene Santos Jr. Thayer

slide-1
SLIDE 1

November 12, 2007 Dartmouth College Distributed Information and Intelligence Analysis Group (DI2AG) 1

The Challenge of Cultural The Challenge of Cultural Modeling for Inferring Modeling for Inferring Intentions and Behavior Intentions and Behavior

Eugene Santos Jr. Eugene Santos Jr. Thayer School of Engineering Thayer School of Engineering Dartmouth College Dartmouth College Eugene.Santos.Jr@dartmouth.edu Eugene.Santos.Jr@dartmouth.edu

WI/IAT Joint Keynote, Silicon Valley, USA, 2007

slide-2
SLIDE 2

November 12, 2007 Dartmouth College 2 Distributed Information and Intelligence Analysis Group (DI2AG)

Do we need to Do we need to “ “soft soft-

  • stuff

stuff” ”? ?

  • What is culture? What is the

What is culture? What is the “ “soft soft-

  • stuff

stuff” ”? What ? What is the is the “ “human human-

  • factor

factor” ”? ?

  • How does the psychological, social, political,

How does the psychological, social, political, philosophical beliefs or systems of humans and philosophical beliefs or systems of humans and human organizations affect decision human organizations affect decision-

  • making?

making?

  • How do we quantify the

How do we quantify the “ “soft soft-

  • stuff

stuff” ” for for computation? computation?

slide-3
SLIDE 3

November 12, 2007 Dartmouth College 3 Distributed Information and Intelligence Analysis Group (DI2AG)

On the Menu On the Menu – – Are you being served? Are you being served?

  • Found on Beijing/Chinese restaurant menus

Found on Beijing/Chinese restaurant menus

  • “

“virgin chicken virgin chicken” ”

  • “

“burnt lion burnt lion’ ’s head s head” ”

  • “

“Strange Strange Flavour Flavour of inside

  • f inside Freasure

Freasure’” ’”

  • In preparation for 2008 Beijing Olympics

In preparation for 2008 Beijing Olympics – –

  • [But the poor English translations]

[But the poor English translations] “ “either scare or embarrass either scare or embarrass foreign customers and may cause misunderstanding of foreign customers and may cause misunderstanding of China's diet habits China's diet habits” ”. . Beijing Tourism Bureau Beijing Tourism Bureau

  • BBC News (31 August 2007),

BBC News (31 August 2007), “ “China dishes up menu translations, China dishes up menu translations,” ” http://news.bbc.co.uk/2/hi/asia http://news.bbc.co.uk/2/hi/asia-

  • pacific/6972280.stm

pacific/6972280.stm

slide-4
SLIDE 4

November 12, 2007 Dartmouth College 4 Distributed Information and Intelligence Analysis Group (DI2AG)

A Troubled Food Drop A Troubled Food Drop

  • KUKES, Albania, April 11, 1999 (Reuters)

KUKES, Albania, April 11, 1999 (Reuters) – –

  • [Summary] Kosovo refugees are throwing away U.S.

[Summary] Kosovo refugees are throwing away U.S.-

  • donated humanitarian

donated humanitarian rations by the thousands and have even burned some to keep warm, rations by the thousands and have even burned some to keep warm, complaining that the food is inedible and has made people sick. complaining that the food is inedible and has made people sick. Piles of Piles of unopened packages each unopened packages each labelled labelled “ “A Food Gift from the People of the United A Food Gift from the People of the United States of America States of America” ” litter the grounds of makeshift camps housing many of the litter the grounds of makeshift camps housing many of the 150,000 ethnic Albanians who have poured across the border in re 150,000 ethnic Albanians who have poured across the border in recent weeks. cent weeks. The story says that the meals The story says that the meals “ “which include items such as three which include items such as three-

  • bean

bean casserole, legume stew and vegetarian goulash which are foreign casserole, legume stew and vegetarian goulash which are foreign to their to their normal diet normal diet” ” made their children vomit. made their children vomit.

  • U.S. officials say the rations, a civilian version of military r

U.S. officials say the rations, a civilian version of military ready eady-

  • to

to-

  • eat meals,

eat meals, are are suitable for all religions suitable for all religions. .

  • “

“Kosovo refugees spurn U.S. rations as inedible, Kosovo refugees spurn U.S. rations as inedible,” ” Matt Matt Spetalnick Spetalnick, KUKES, Albania, April , KUKES, Albania, April 11 (Reuters) 11 (Reuters)

slide-5
SLIDE 5

November 12, 2007 Dartmouth College 5 Distributed Information and Intelligence Analysis Group (DI2AG)

Jawbreaker Jawbreaker

  • Scenario from Afghanistan Conflict (US War on

Scenario from Afghanistan Conflict (US War on Terror) Terror)

  • Highlights Special Forces interaction with

Highlights Special Forces interaction with indigenous forces indigenous forces

  • Role of complex relationships

Role of complex relationships

  • Jawbreaker: The Attack on Bin Laden and Al Qaeda: A Personal Acc

Jawbreaker: The Attack on Bin Laden and Al Qaeda: A Personal Account by the CIA's Key

  • unt by the CIA's Key

Field Commander Field Commander, , Berntsen Berntsen and and Pezzullo Pezzullo, Crown, 2005. , Crown, 2005.

slide-6
SLIDE 6

November 12, 2007 Dartmouth College 6 Distributed Information and Intelligence Analysis Group (DI2AG)

Cultural Complexity Cultural Complexity

  • All the examples are

All the examples are “ “understandable understandable” ”

  • i.e., can actually get their meanings and anticipate behaviors

i.e., can actually get their meanings and anticipate behaviors

  • They are funny or sad because we only see it from one cultural

They are funny or sad because we only see it from one cultural point of view point of view

  • Examples reflect only American/English cultural viewpoint

Examples reflect only American/English cultural viewpoint

  • Replace

Replace “ “American/English American/English” ” with any group and have similar problems with any group and have similar problems

  • Moral

Moral – – You are already biased when you apply your own You are already biased when you apply your own cultural beliefs and perceptions cultural beliefs and perceptions

  • Challenge

Challenge – – Examples were increasing in complexity Examples were increasing in complexity

slide-7
SLIDE 7

November 12, 2007 Dartmouth College 7 Distributed Information and Intelligence Analysis Group (DI2AG)

Culture is Everywhere Culture is Everywhere

  • Ultimately impacts and drives many areas

Ultimately impacts and drives many areas

  • Planning and Execution

Planning and Execution

  • Logistics and Supply Chain Management

Logistics and Supply Chain Management

  • Negotiation and Mediation

Negotiation and Mediation

  • Economics and Business Practices

Economics and Business Practices

  • …

…

  • Intelligent Agents

Intelligent Agents

  • Many have always been aware of culture

Many have always been aware of culture’ ’s significance s significance

  • Advertising, Politics, Television, Film, and Service Industries,

Advertising, Politics, Television, Film, and Service Industries, to name a few. to name a few.

slide-8
SLIDE 8

November 12, 2007 Dartmouth College 8 Distributed Information and Intelligence Analysis Group (DI2AG)

So, how can we hope to take culture and make it computational?

How do we show culture’s impact on decision-making?

slide-9
SLIDE 9

November 12, 2007 Dartmouth College 9 Distributed Information and Intelligence Analysis Group (DI2AG)

Modeling Others Modeling Others

  • Required in a multitude of domains when

Required in a multitude of domains when actions/reactions/counteractions of others matter actions/reactions/counteractions of others matter

  • Financial/Business Competition

Financial/Business Competition

  • Politics/Elections

Politics/Elections

  • Sports

Sports

  • Warfare/Conflict

Warfare/Conflict

  • Planning and Execution

Planning and Execution

  • Wargaming

Wargaming

  • Adversaries, neutrals, interested parties, communities of

Adversaries, neutrals, interested parties, communities of interests, nation interests, nation-

  • states, conglomerates,

states, conglomerates, … …

slide-10
SLIDE 10

November 12, 2007 Dartmouth College 10 Distributed Information and Intelligence Analysis Group (DI2AG)

What is Intent? What is Intent?

  • Intent inference

Intent inference (or (or user intent inference user intent inference) involves ) involves deducing an entity deducing an entity’ ’s goals based on observations of that s goals based on observations of that entity entity’ ’s actions (Geddes, 1986) s actions (Geddes, 1986)

  • Deduction involves the construction of one or more behavioral

Deduction involves the construction of one or more behavioral models that have been optimized to the entity models that have been optimized to the entity’ ’s behavior patterns s behavior patterns

  • Data/knowledge representing observations of an entity, the

Data/knowledge representing observations of an entity, the entity entity’ ’s actions, or the entity s actions, or the entity’ ’s environment (collectively called s environment (collectively called

  • bservables
  • bservables) are collected and delivered to the

) are collected and delivered to the model(s model(s) )

  • Models attempt to match observables against patterns of

Models attempt to match observables against patterns of behavior and derive inferred intent from those patterns behavior and derive inferred intent from those patterns

  • Useful for generation of advice, definition of future informatio

Useful for generation of advice, definition of future information n requirements, proactive aiding, or a host of other benefits (Bel requirements, proactive aiding, or a host of other benefits (Bell et al., l et al., 2002; Santos, 2003; Santos & Zhao 2006) 2002; Santos, 2003; Santos & Zhao 2006)

slide-11
SLIDE 11

November 12, 2007 Dartmouth College 11 Distributed Information and Intelligence Analysis Group (DI2AG)

Intent Intent – – What can you do with it? What can you do with it?

  • Predict the future

Predict the future: actions, reactions, : actions, reactions, behaviours behaviours, , etc. etc.

  • Explain the present

Explain the present: causes, motivations, goals, : causes, motivations, goals, etc. etc.

  • Understand the past

Understand the past: beliefs, axioms, history, etc. : beliefs, axioms, history, etc.

  • Inferred intent knowledge can help focus and

Inferred intent knowledge can help focus and prune search space, bound optimization, guide prune search space, bound optimization, guide scheduling, better allocate resources, scheduling, better allocate resources, … …

  • (Bell et al. 2002; Santos 2003; Santos & Zhao 2006)

(Bell et al. 2002; Santos 2003; Santos & Zhao 2006)

slide-12
SLIDE 12

November 12, 2007 Dartmouth College 12 Distributed Information and Intelligence Analysis Group (DI2AG)

Approaches to Intent Inferencing Approaches to Intent Inferencing

  • Plan

Plan-

  • goal

goal-

  • graph (PGG)

graph (PGG) – – a network of plans a network of plans and goals, where each high level goal is and goals, where each high level goal is decomposed into a set of plans for achieving it, decomposed into a set of plans for achieving it, and the plans are decomposed into and the plans are decomposed into subgoals subgoals which in turn are decomposed into lower which in turn are decomposed into lower-

  • level

level plans (Geddes, 1994) plans (Geddes, 1994)

  • Intent is finding the path from observables to a plan

Intent is finding the path from observables to a plan

  • r goal
  • r goal
slide-13
SLIDE 13

November 12, 2007 Dartmouth College 13 Distributed Information and Intelligence Analysis Group (DI2AG)

Approaches to Intent Inferencing Approaches to Intent Inferencing

  • Operator function model (OFM)

Operator function model (OFM) – – an expert an expert system using a system using a heterarchic heterarchic-

  • hierarchic network of

hierarchic network of finite finite-

  • state automata, in which nodes represent

state automata, in which nodes represent entity entity’ ’s activities and arcs represent conditions s activities and arcs represent conditions that initiate/terminate certain activities that initiate/terminate certain activities (Bushman et al., 1993; Chu et al., 1995; Rubin et (Bushman et al., 1993; Chu et al., 1995; Rubin et al., 1988) al., 1988)

  • Connect observed action to appropriate activity trees

Connect observed action to appropriate activity trees

slide-14
SLIDE 14

November 12, 2007 Dartmouth College 14 Distributed Information and Intelligence Analysis Group (DI2AG)

Approaches to Intent Inferencing Approaches to Intent Inferencing

  • Generalized plan recognition (GPR)

Generalized plan recognition (GPR) – – recognize recognize the entity the entity’ ’s plan for carrying out the task, based s plan for carrying out the task, based

  • n observations, an exhaustive set of discrete
  • n observations, an exhaustive set of discrete

actions (a plan library), and constraints ( actions (a plan library), and constraints (Lesh Lesh et et al., 1998; al., 1998; Carberry Carberry, 1988; Goodman and , 1988; Goodman and Litman Litman, , 1990) 1990)

slide-15
SLIDE 15

November 12, 2007 Dartmouth College 15 Distributed Information and Intelligence Analysis Group (DI2AG)

Intent Inferencing and Culture Intent Inferencing and Culture

  • Approaches mentioned so far mainly for static

Approaches mentioned so far mainly for static plan recognition plan recognition

  • Can be ambiguous on who/what being modeled

Can be ambiguous on who/what being modeled

  • Only secondary focus on uncertainty

Only secondary focus on uncertainty

  • Culture not directly addressed

Culture not directly addressed

slide-16
SLIDE 16

November 12, 2007 Dartmouth College 16 Distributed Information and Intelligence Analysis Group (DI2AG)

What do you need to know about What do you need to know about “ “others

  • thers”

”? ?

  • Things like:

Things like:

  • Histories of responses and actions in different situations?

Histories of responses and actions in different situations?

  • Social/Economic/Military/Political/Religious doctrine?

Social/Economic/Military/Political/Religious doctrine?

  • Infrastructure and reliability of leadership or command and

Infrastructure and reliability of leadership or command and control? control?

  • Perceptions about us (our force) or other groups?

Perceptions about us (our force) or other groups?

  • Political and cultural factors?

Political and cultural factors?

  • Might provide clues on their propensity for future

Might provide clues on their propensity for future actions? actions?

  • What do we really need and why?

What do we really need and why?

slide-17
SLIDE 17

November 12, 2007 Dartmouth College 17 Distributed Information and Intelligence Analysis Group (DI2AG)

What is What is “ “Other Other’ ’s s” ” Intent? Intent?

  • What

What’ ’s the context of an action they took? s the context of an action they took?

  • What is the rationale behind the action?

What is the rationale behind the action?

  • What are the causes and effects of their

What are the causes and effects of their intended goal? intended goal?

  • What is the motivation behind a specific

What is the motivation behind a specific behaviour behaviour? ?

  • What will happen next?

What will happen next?

  • Why did this behaviour occur?

Why did this behaviour occur?

  • What do they believe?

What do they believe?

slide-18
SLIDE 18

November 12, 2007 Dartmouth College 18 Distributed Information and Intelligence Analysis Group (DI2AG)

Accounting for Human Factors in Accounting for Human Factors in Capturing Intent Capturing Intent

  • Assymetric

Assymetric – – they are not like us; we do not think like them they are not like us; we do not think like them

  • “

“What is rational What is rational” ” is not the same between different is not the same between different individuals or groups especially with different backgrounds. individuals or groups especially with different backgrounds.

  • Differences in decision

Differences in decision-

  • making and behavior come from

making and behavior come from differences in background differences in background

  • Social

Social

  • Cultural

Cultural

  • Economic

Economic

  • Political

Political

  • Psychological

Psychological

slide-19
SLIDE 19

November 12, 2007 Dartmouth College 19 Distributed Information and Intelligence Analysis Group (DI2AG)

“ “Other Other’ ’s s” ” Intent Intent

  • Intent is not just the plan or course of action of the other

Intent is not just the plan or course of action of the other entity entity

  • Not just

Not just “ “The enemy commander The enemy commander intends intends to launch his to launch his SAMs SAMs” ”, , “ “The organization The organization intends intends to undertake a suicide to undertake a suicide bombing bombing” ”, or , or “ “The corporation The corporation intends intends to to ‘ ‘go green go green’ ’ ” ”, , but but also why?? also why??

  • Intent

Intent is the highest is the highest-

  • level

level goal(s goal(s) they are pursuing ) they are pursuing + + the the support for that goal support for that goal + + the plan to achieve it the plan to achieve it

  • Need intent to

Need intent to understand understand and and predict predict other

  • ther’

’s behavior s behavior

  • Must model them based on their

Must model them based on their perceptions perceptions of the world

  • f the world
  • (Santos 2003; Santos & Zhao 2006)

(Santos 2003; Santos & Zhao 2006)

slide-20
SLIDE 20

November 12, 2007 Dartmouth College 20 Distributed Information and Intelligence Analysis Group (DI2AG)

Challenges Challenges

  • Each individual or group is a unique entity

Each individual or group is a unique entity

  • Human factors are difficult to capture accurately

Human factors are difficult to capture accurately and/or completely and/or completely

  • Uncertainty associated with the impacts of human

Uncertainty associated with the impacts of human factors on decision factors on decision-

  • making process is inherent

making process is inherent

  • Intent and behavior evolves over time

Intent and behavior evolves over time

  • Underlying problem of abductive inference

Underlying problem of abductive inference

  • Intuition

Intuition – – explain the source of the observations explain the source of the observations

  • “

“If you build it, does it really work? If you build it, does it really work?” ”

slide-21
SLIDE 21

November 12, 2007 Dartmouth College 21 Distributed Information and Intelligence Analysis Group (DI2AG)

Modeling and Perception Modeling and Perception

  • Our Approach: Model of entity based from entity

Our Approach: Model of entity based from entity’ ’s s perception or point of view perception or point of view

  • How does the entity view the world?

How does the entity view the world?

  • What can the entity observe about us and others?

What can the entity observe about us and others?

  • Explanation of entity behavior grounded in terms of the

Explanation of entity behavior grounded in terms of the entity entity’ ’s world s world-

  • view

view

  • Avoids accidentally imposing our beliefs on the entity

Avoids accidentally imposing our beliefs on the entity

  • Observables and evidence passed to the entity model is

Observables and evidence passed to the entity model is based on the above questions based on the above questions

  • Obviously, the entity does not see everything

Obviously, the entity does not see everything

  • Allows for modeling of deception

Allows for modeling of deception

slide-22
SLIDE 22

November 12, 2007 Dartmouth College 22 Distributed Information and Intelligence Analysis Group (DI2AG)

Objectives Objectives

  • Design and develop a computational model for

Design and develop a computational model for inferring other inferring other’ ’s intent and predicting behavior s intent and predicting behavior

  • Build and employ social, cultural, and political

Build and employ social, cultural, and political data data-

  • driven models to

driven models to explore and explore and explain explain (in (in addition to modeling) their attitudes and addition to modeling) their attitudes and behaviors behaviors

slide-23
SLIDE 23

November 12, 2007 Dartmouth College 23 Distributed Information and Intelligence Analysis Group (DI2AG)

What to do? What to do?

  • “

“Cultural Cultural” ” knowledge fragments knowledge fragments – – human human factors (elements) that define or influence factors (elements) that define or influence decision decision-

  • making central to a particular individual

making central to a particular individual

  • r organization
  • r organization
  • How to build a library of fragments (ingredients)

How to build a library of fragments (ingredients) that can be selected from which to build the that can be selected from which to build the model (soup model (soup – – tasty) tasty)

slide-24
SLIDE 24

November 12, 2007 Dartmouth College 24 Distributed Information and Intelligence Analysis Group (DI2AG)

An Intent Modeling Approach An Intent Modeling Approach

  • Incorporate human factors

Incorporate human factors

  • Intent driven

Intent driven

  • Model the decision making process based on

Model the decision making process based on how how “ “other

  • ther’

’s s” ” views the world views the world

  • Build network fragments for each piece of

Build network fragments for each piece of information / knowledge, and merge them information / knowledge, and merge them together for reasoning together for reasoning

  • Based on Bayesian Knowledge Bases (BKBs)

Based on Bayesian Knowledge Bases (BKBs)

  • Fragments built and validated jointly with social

Fragments built and validated jointly with social scientist/subject matter experts scientist/subject matter experts

slide-25
SLIDE 25

November 12, 2007 Dartmouth College 25 Distributed Information and Intelligence Analysis Group (DI2AG)

Intent Driven Approach Intent Driven Approach

  • Model adversary through 3 formative components:

Model adversary through 3 formative components:

  • Goals/Foci

Goals/Foci: : A prioritized (by probability) list of short and long A prioritized (by probability) list of short and long term goals representing entity intents, objectives or foci. The term goals representing entity intents, objectives or foci. The goal component captures goal component captures what what the entity is doing. the entity is doing.

  • Rationale Network

Rationale Network: : A probabilistic network representing the A probabilistic network representing the influences of the entity influences of the entity’ ’s beliefs, both about themselves and s beliefs, both about themselves and

  • thers, on their goals and on high level actions associated with
  • thers, on their goals and on high level actions associated with

those goals. The rationale component infers those goals. The rationale component infers why why the entity is the entity is behaving in a certain fashion. behaving in a certain fashion.

  • Actions Network

Actions Network: : A probabilistic network representing the A probabilistic network representing the detailed relationships between entity goals and possible actions detailed relationships between entity goals and possible actions to realize those goals. The action component captures to realize those goals. The action component captures how how an an entity might act. entity might act.

  • (Santos 2003; Santos &

(Santos 2003; Santos & Negri Negri 2004; Santos & Zhao 2006) 2004; Santos & Zhao 2006)

slide-26
SLIDE 26

November 12, 2007 Dartmouth College 26 Distributed Information and Intelligence Analysis Group (DI2AG)

Bayesian Knowledge Bayesian Knowledge-

  • Bases

Bases

  • Simple method of knowledge representation

Simple method of knowledge representation

  • “

“if if-

  • then

then” ” rules with conditional probabilities rules with conditional probabilities

  • Mathematically sound model

Mathematically sound model

  • Subsumes existing knowledge representations

Subsumes existing knowledge representations

  • Bayesian Networks [Pearl 1988; Pearl 2000]

Bayesian Networks [Pearl 1988; Pearl 2000]

  • Handles incomplete and cyclic information

Handles incomplete and cyclic information

  • Eases problems in V & V

Eases problems in V & V

(Santos & Santos 96; Santos et al. 97; Santos et al. 97b; (Santos & Santos 96; Santos et al. 97; Santos et al. 97b; Santos & Santos 99; Santos & Santos 99; Shimony Shimony et al. 00; Johnson & et al. 00; Johnson & Santos 00; Rosen et al. 01; Santos et al. 2004) Santos 00; Rosen et al. 01; Santos et al. 2004)

slide-27
SLIDE 27

November 12, 2007 Dartmouth College 27 Distributed Information and Intelligence Analysis Group (DI2AG) (B) Belief (A) Action (X) Axiom (G) Goal

What the entity believes about others What the entity believes about themselves What results the entity wants to achieve How they will carry out their tasks

Basics for BKB fragments and Entity Basics for BKB fragments and Entity Intent Inferencing Model Intent Inferencing Model

(X),(B),(G),(A) Support-Node (S-node). Each S-node has a probability value Instantiation-Node (I-node). Each I-node needs to be supported by a S-node

slide-28
SLIDE 28

November 12, 2007 Dartmouth College 28 Distributed Information and Intelligence Analysis Group (DI2AG) Current Foci Explanation and Enemy Intent Inferred Actions Inferred Rationale Feedback N e w F

  • c

i Feedback New Foci New Rationale Current Rationale Feedback

Processing for Entity Intent Processing for Entity Intent

(Santos 2003; Santos & Zhao 2006) (Santos 2003; Santos & Zhao 2006)

Observables Entity Analysts Entity Foci

Short-Term Long-Term

Entity Actions Entity Rationale

slide-29
SLIDE 29

November 12, 2007 Dartmouth College 29 Distributed Information and Intelligence Analysis Group (DI2AG)

Examples of Capturing Motivations Examples of Capturing Motivations

  • Jawbreaker

Jawbreaker

  • Competing for leadership in the Middle East

Competing for leadership in the Middle East

  • Modeling Multiple Communities of Interest and

Modeling Multiple Communities of Interest and Their Interactions Their Interactions

  • Modeling Intelligence Analysts and their

Modeling Intelligence Analysts and their Cognitive Styles Cognitive Styles

slide-30
SLIDE 30

November 12, 2007 Dartmouth College 30 Distributed Information and Intelligence Analysis Group (DI2AG)

Initial Model (Jawbreaker) Initial Model (Jawbreaker)

slide-31
SLIDE 31

November 12, 2007 Dartmouth College 31 Distributed Information and Intelligence Analysis Group (DI2AG)

Example from Jawbreaker Example from Jawbreaker

slide-32
SLIDE 32

November 12, 2007 Dartmouth College 32 Distributed Information and Intelligence Analysis Group (DI2AG)

Our study: Terror attacks Our study: Terror attacks

  • To maximize data availability use recent

To maximize data availability use recent Palestinian Palestinian-

  • Israeli conflict.

Israeli conflict.

  • Unambiguous measures: E.g., No. of attacks,

Unambiguous measures: E.g., No. of attacks,

  • No. casualties for 5 factions (PIJ,
  • No. casualties for 5 factions (PIJ, Hamas

Hamas, PLFP, , PLFP, Fateh Fateh, Al , Al-

  • Aqsa

Aqsa Martyr Martyr’ ’s Brigade). s Brigade).

  • Monthly sums January, 1999

Monthly sums January, 1999-

  • Dec. 2005
  • Dec. 2005
  • Four independent sources for each datum

Four independent sources for each datum → → test test intersource intersource reliability reliability. .

slide-33
SLIDE 33

November 12, 2007 Dartmouth College 33 Distributed Information and Intelligence Analysis Group (DI2AG)

Sample preliminary Results: Sample preliminary Results: Palestinian & Israeli Politics Palestinian & Israeli Politics

  • Casualties by IDF decrease Palestinian support

Casualties by IDF decrease Palestinian support for peace process and increase support for for peace process and increase support for attacks against Israeli civilians. attacks against Israeli civilians.

  • Increased Palestinian popular support for

Increased Palestinian popular support for attacks increases the likelihood of attacks by attacks increases the likelihood of attacks by smaller factions (PFLP, PIJ) but not for larger smaller factions (PFLP, PIJ) but not for larger factions ( factions (Hamas Hamas, , Fateh Fateh). ).

  • Perceived corruption in PA relates to support

Perceived corruption in PA relates to support for for Hamas Hamas and attacks by and attacks by Hamas Hamas. .

slide-34
SLIDE 34

November 12, 2007 Dartmouth College 34 Distributed Information and Intelligence Analysis Group (DI2AG)

Empirical problems in predicting Empirical problems in predicting terrorist actions terrorist actions

  • Multiple actions may be coded as

Multiple actions may be coded as “ “terrorism terrorism” ” (e.g., throwing rocks, car bombs, kidnapping). (e.g., throwing rocks, car bombs, kidnapping).

  • Actions are low frequency in most places.

Actions are low frequency in most places.

  • Claims/knowledge of who is responsible for

Claims/knowledge of who is responsible for given action can be suspect. given action can be suspect.

  • Reliability and validity of measures are never

Reliability and validity of measures are never ascertained. ascertained.

  • What about modeling groups and their

What about modeling groups and their interactions? interactions?

slide-35
SLIDE 35

November 12, 2007 Dartmouth College 35 Distributed Information and Intelligence Analysis Group (DI2AG)

A Network for Religious based Insurgent Group A Network for Religious based Insurgent Group

(Santos et al. 2007) (Santos et al. 2007)

Religious insurgent group can be influenced more by religious reasons

slide-36
SLIDE 36

November 12, 2007 Dartmouth College 36 Distributed Information and Intelligence Analysis Group (DI2AG) Sequence 3 1. Coalition Raid 2. Coalition Captures Insurgent Leaders 3. Religious Leader Condemns Heathenry 4. Coalition Distribute Supplies 5. Coalition Meet Religious Leaders, and Religious Leaders Call for Peace (Clear Evidence ‘Religious Leader Condemns’) Group Religous Group Secular

The Dynamic Adversarial Gaming Algorithm Project (DAGA), The Dynamic Adversarial Gaming Algorithm Project (DAGA), Securboration Securboration, Inc., Dartmouth, and , Inc., Dartmouth, and UConn UConn (Santos et al. 2007) (Santos et al. 2007)

slide-37
SLIDE 37

November 12, 2007 Dartmouth College 37 Distributed Information and Intelligence Analysis Group (DI2AG)

DAGA Results DAGA Results

  • Demonstration depicts how Coalition Actions effect

Demonstration depicts how Coalition Actions effect the adversary the adversary’ ’s goals and beliefs and predict the s goals and beliefs and predict the response of response of COIs COIs for for

  • Religious Individual

Religious Individual

  • Religious Groups

Religious Groups

  • Mainstream Individuals

Mainstream Individuals

  • Secular Groups

Secular Groups

  • Demonstrates that cultural aspects can be included in

Demonstrates that cultural aspects can be included in adversarial gaming algorithm vs. providing a adversarial gaming algorithm vs. providing a ‘ ‘validated validated’ ’ model of exactly how the culture influences behavior. model of exactly how the culture influences behavior.

  • See (Santos et al. 2007)

See (Santos et al. 2007)

slide-38
SLIDE 38

November 12, 2007 Dartmouth College 38 Distributed Information and Intelligence Analysis Group (DI2AG)

Lesson Learned Lesson Learned

  • Terrorist groups, no matter religious based or secular

Terrorist groups, no matter religious based or secular based, all have political goals (Bloom 2005). This might based, all have political goals (Bloom 2005). This might make different groups behave similarly in a certain make different groups behave similarly in a certain period time, when their short period time, when their short-

  • term goals are similar.

term goals are similar.

  • Public

Public’ ’s attitudes and behaviors are influenced by s attitudes and behaviors are influenced by leaders, and groups behaviors are also influenced by leaders, and groups behaviors are also influenced by public. public.

  • In a multi

In a multi-

  • agent simulation system, proper algorithms,

agent simulation system, proper algorithms, which consider human factors, on mapping influences which consider human factors, on mapping influences among different entities are needed. among different entities are needed.

slide-39
SLIDE 39

November 12, 2007 Dartmouth College 39 Distributed Information and Intelligence Analysis Group (DI2AG)

Select Study Sources Select Study Sources

  • Mia Bloom (2005) Dying to Kill: the Allure of Suicide

Mia Bloom (2005) Dying to Kill: the Allure of Suicide Terror. Terror.

  • Gary Berntsen and Ralph Pezzullo (2005) Jawbreaker:

Gary Berntsen and Ralph Pezzullo (2005) Jawbreaker: the Attack on Bin Laden and Al Qaeda: a Personal the Attack on Bin Laden and Al Qaeda: a Personal Account by the CIA Account by the CIA’ ’s Key Field Commander. s Key Field Commander.

  • Robert Pape (2005) Dying to Win: the Strategic Logic

Robert Pape (2005) Dying to Win: the Strategic Logic

  • f Suicide Terrorism.
  • f Suicide Terrorism.
  • Ram Sidi (Veteran member of Israel's counter

Ram Sidi (Veteran member of Israel's counter-

  • terrorism

terrorism establishment). (02/2006) Presentation on the establishment). (02/2006) Presentation on the motivations of these Palestinian terrorist groups, at motivations of these Palestinian terrorist groups, at Dartmouth College Dartmouth College

slide-40
SLIDE 40

November 12, 2007 Dartmouth College 40 Distributed Information and Intelligence Analysis Group (DI2AG)

  • Problems in processing information by

Problems in processing information by users/analysts: users/analysts:

  • Massive data from different resources

Massive data from different resources

  • Repetitive tasks

Repetitive tasks

  • Lack of tools for linking known events and knowledge

Lack of tools for linking known events and knowledge

  • Modeling analysts is central to creating a synergistic

Modeling analysts is central to creating a synergistic relationship between an analyst and tools used to relationship between an analyst and tools used to search/ analyze information search/ analyze information

  • What is an analyst actually

What is an analyst actually intending intending to look for? Why to look for? Why are they looking for it? How will they look for it? are they looking for it? How will they look for it?

  • (Santos et al. 1999; Santos et al. 2003; Nguyen et al. 2004; San

(Santos et al. 1999; Santos et al. 2003; Nguyen et al. 2004; Santos et al. 2005) tos et al. 2005)

User/Analyst Intent

slide-41
SLIDE 41

November 12, 2007 Dartmouth College 41 Distributed Information and Intelligence Analysis Group (DI2AG)

Problem Problem

  • Want to help analysts better retrieve info

Want to help analysts better retrieve info

  • Challenges of employing User Modeling (UM)

Challenges of employing User Modeling (UM) techniques for IR techniques for IR

  • Partial information

Partial information

  • Uncertainty

Uncertainty

  • Vagueness

Vagueness

  • Dynamics

Dynamics

slide-42
SLIDE 42

November 12, 2007 Dartmouth College 42 Distributed Information and Intelligence Analysis Group (DI2AG)

Key Factors Key Factors

  • Important factors for building user models for

Important factors for building user models for IR: IR:

  • Relevance feedback.

Relevance feedback.

  • Adaptation.

Adaptation.

  • Intention.

Intention.

  • Author

Author’ ’s intention s intention

  • User

User’ ’s intention s intention

  • (Nguyen et al. 2004)

(Nguyen et al. 2004)

slide-43
SLIDE 43

November 12, 2007 Dartmouth College 43 Distributed Information and Intelligence Analysis Group (DI2AG)

Sentence Chopper Texts documents Link Parser Relation Generator Noun phrase extractor DAG Query Document Graph Query Graph Query Modifier Graph Matching Pro Query Graph User Interest User Context

Preference Network

BN Engine

Feedback

User Modeling Functional Architecture In User Modeling Functional Architecture In Information Retrieval System Information Retrieval System

slide-44
SLIDE 44

November 12, 2007 Dartmouth College 44 Distributed Information and Intelligence Analysis Group (DI2AG)

Create and update context network Create and update context network

  • Context network is constructed

Context network is constructed “ “on

  • n-
  • the

the-

  • fly

fly” ” by finding a common set of by finding a common set of sub sub-

  • graphs among the relevant

graphs among the relevant documents documents

  • Captured as BKB

Captured as BKB

  • Each document is represented as a

Each document is represented as a document graph (DG). document graph (DG).

  • A sub

A sub-

  • graph

graph X X of DG

  • f DG Y

Y is: is:

  • X

X is a DG is a DG

  • ∀

∀ node node a a ∈ ∈X X → → a a ∈ ∈Y Y

slide-45
SLIDE 45

November 12, 2007 Dartmouth College 45 Distributed Information and Intelligence Analysis Group (DI2AG)

Interest Set

Interest set: Interest set: terrorism, terrorism, level=0.96 level=0.96 terrorist, terrorist, level=0.9 level=0.9 money money_ _laundering, level=0.93 laundering, level=0.93 al al_ _quaeda quaeda, level=0.85 , level=0.85 deposit, deposit, level=0.82 level=0.82 withdraw, withdraw, level=0.81 level=0.81 explosion, explosion, level=0.8 level=0.8 terrorist terrorist_ _training, level=0.79 training, level=0.79 forged forged_ _document, level=0.77 document, level=0.77 bank bank _ _account, level =0.76 account, level =0.76

slide-46
SLIDE 46

November 12, 2007 Dartmouth College 46 Distributed Information and Intelligence Analysis Group (DI2AG)

Context Network

related _to al_queda isa

terrorism_ group

hezabola isa taliban related _to terrorist related _to isa

  • rganization

related _to

terrorism_ activity

related _to

terrorism_ attack

isa

terrorism

bombing retroactive_ material related _to explosive _training related _to explosive training related _to money_ laundering related _to related _to banking_ transaction deposit withdraw isa isa

related _to related _to

bank_ account

related _to related _to related _to

related _to isa account related _to al_quaeda isa

terrorism_ group

hezabola isa taliban related _to terrorist related _to isa

  • rganization

related _to

terrorism_ activity

related _to

terrorism_ attack

isa

terrorism

bombing radioactive_ material related _to explosive _training related _to explosive training related _to money_ laundering related _to related _to banking_ transaction deposit withdraw isa isa

related _to related _to

bank_ account

related _to related _to related _to

related _to isa account

slide-47
SLIDE 47

November 12, 2007 Dartmouth College 47 Distributed Information and Intelligence Analysis Group (DI2AG)

Modified Query Graph

Original query graph Modified query graph

banking_ transaction transaction banking Related_to Isa banking_ transaction transaction banking Related_to Isa money_ laundering Related_to deposit withdraw Isa bank_ account Related_to Isa Related_to Related _to

Related_to Related_to

slide-48
SLIDE 48

November 12, 2007 Dartmouth College 48 Distributed Information and Intelligence Analysis Group (DI2AG)

Human Subject Evaluation with Human Subject Evaluation with CNS collection CNS collection

  • Evaluation took place in NIST laboratory.

Evaluation took place in NIST laboratory.

  • Test bed: Center for Nonproliferation Studies (CNS) database

Test bed: Center for Nonproliferation Studies (CNS) database ( (sept

  • sept. 2003).

. 2003).

  • Country profiles on WMD

Country profiles on WMD

  • Arms control

Arms control

  • WMD Terrorism

WMD Terrorism

  • UM vs. Verity Query Language

UM vs. Verity Query Language. .

  • Subjects use two different systems in parallel.

Subjects use two different systems in parallel.

  • There are 10 scripted queries on

There are 10 scripted queries on “ “Iran R&D supporting Iran R&D supporting BW BW” ”. .

  • Only first 10 documents have been reviewed for relevancy.

Only first 10 documents have been reviewed for relevancy.

  • Three analysts took part in the experiments

Three analysts took part in the experiments

  • Different experience, styles, and training level.

Different experience, styles, and training level.

slide-49
SLIDE 49

November 12, 2007 Dartmouth College 49 Distributed Information and Intelligence Analysis Group (DI2AG)

User Model Helps Find Documents Based on User Model Helps Find Documents Based on Individual Interests Individual Interests

(Santos et al. 2005) (Santos et al. 2005)

8 15 1 1 5 9

User 1 User 2 User 3

3 3 3 3 13 2

User 1 User 2 User 3

Overlap of Unique Relevant Documents

User Model Verity Query Language

User Model Verity Total unique relevant documents 39 27 Documents marked as relevant by all 3 analysts 8 3 Documents marked as relevant by more than 2 analysts 15 19 Documents marked as relevant by only 1 analyst 24 8 When faced with a search problem do you tend to: 1 Both 2 Start with big picture 3 Start from details

slide-50
SLIDE 50

November 12, 2007 Dartmouth College 50 Distributed Information and Intelligence Analysis Group (DI2AG)

What did we learn? What did we learn?

  • Continue to develop tools and methodologies for capturing cultur

Continue to develop tools and methodologies for capturing cultural al aspects of intent aspects of intent

  • Developing formal measures for fragment selection, fusion, and m

Developing formal measures for fragment selection, fusion, and model

  • del

composition composition

  • Resolve missing data and probabilities by developing models

Resolve missing data and probabilities by developing models (Bayesian knowledge fragments) that can be evaluated, at least (Bayesian knowledge fragments) that can be evaluated, at least subjectively, by the subject matter experts (social psychologist subjectively, by the subject matter experts (social psychologists, s, politic scientists, etc.) politic scientists, etc.)

  • Iterative process

Iterative process

  • Special Dartmouth Women in Science Program (WISP) supporting 3

Special Dartmouth Women in Science Program (WISP) supporting 3 highly enthusiastic freshmen to study: Afghanistan culture and s highly enthusiastic freshmen to study: Afghanistan culture and social

  • cial

structure, Iranian leadership and organizations, and Indonesian structure, Iranian leadership and organizations, and Indonesian culture and social structure culture and social structure

  • Successful due to neutral POV with regards to modeling

Successful due to neutral POV with regards to modeling

slide-51
SLIDE 51

November 12, 2007 Dartmouth College 51 Distributed Information and Intelligence Analysis Group (DI2AG)

Challenges of Validation Challenges of Validation

  • HowHowHowHowHowHowHowHow

HowHowHowHowHowHowHowHow? ?

  • History not necessarily indicative of the future

History not necessarily indicative of the future

  • People/groups/nations change

People/groups/nations change

  • Can be drastic, can be long

Can be drastic, can be long-

  • termed

termed

slide-52
SLIDE 52

November 12, 2007 Dartmouth College 52 Distributed Information and Intelligence Analysis Group (DI2AG)

Thanks to Fellow Researchers Thanks to Fellow Researchers

  • Academic Collaborators

Academic Collaborators

  • Felicia

Felicia Pratto Pratto ( (UConn UConn) ) – – cultural and social psychology of individuals and effects of cultural and social psychology of individuals and effects of groups groups

  • George

George Cybenko Cybenko (Dartmouth) (Dartmouth) – – process modeling and the human terrain process modeling and the human terrain

  • Jeff Bradshaw and Paul

Jeff Bradshaw and Paul Feltovich Feltovich (IHMC) (IHMC) – – organizational behavior modeling and joint

  • rganizational behavior modeling and joint

policy management policy management

  • Eunice E. Santos (Virginia Tech)

Eunice E. Santos (Virginia Tech) – – social networks analysis and computational social social networks analysis and computational social science science

  • Industrial and Government Collaborators

Industrial and Government Collaborators

  • Richard Warren (AFRL/Human Effectiveness)

Richard Warren (AFRL/Human Effectiveness)

  • Duane Gilmour (AFRL/Information Directorate)

Duane Gilmour (AFRL/Information Directorate)

  • Lee Krause (Securboration, Inc.)

Lee Krause (Securboration, Inc.)

  • Jore

Jore Park and Park and Wylci Wylci Fables ( Fables (IndaSea IndaSea) )

  • Greg

Greg Jannarone Jannarone (Air War College/Behavioral Influence Analysis) (Air War College/Behavioral Influence Analysis)

  • Thanks to many others as well!

Thanks to many others as well!

  • For more information on our work, see

For more information on our work, see

  • http://di2ag.thayer.dartmouth.edu/~eugene

http://di2ag.thayer.dartmouth.edu/~eugene

  • http://di2ag.thayer.dartmouth.edu

http://di2ag.thayer.dartmouth.edu

slide-53
SLIDE 53

November 12, 2007 Dartmouth College 53 Distributed Information and Intelligence Analysis Group (DI2AG)

Select References for Others (Adversarial) Modeling Select References for Others (Adversarial) Modeling

  • Santos, Eugene, Jr., Zhao,

Santos, Eugene, Jr., Zhao, Qunhua Qunhua, , Pratto Pratto, Felicia, Pearson, Adam R., , Felicia, Pearson, Adam R., McQueary McQueary, Bruce, Breeden, Andy, and Krause, Lee, , Bruce, Breeden, Andy, and Krause, Lee, “ “Modeling Modeling Multiple Communities of Interest for Interactive Simulation and Multiple Communities of Interest for Interactive Simulation and Gaming: The Dynamic Adversarial Gaming Algorithm Project Gaming: The Dynamic Adversarial Gaming Algorithm Project, ,” ” Proceedings of the SPIE: Defense & Security Symposium Proceedings of the SPIE: Defense & Security Symposium, , Vol Vol: 6564, Orlando, FL, 2007. : 6564, Orlando, FL, 2007.

  • Santos, Eugene, Jr. and Zhao,

Santos, Eugene, Jr. and Zhao, Qunhua Qunhua, , “ “Adversarial Models for Opponent Intent Inferencing Adversarial Models for Opponent Intent Inferencing, ,” ” in in Adversarial Reasoning: Computational Adversarial Reasoning: Computational Approaches to Reading Approaches to Reading the Opponents Mind the Opponents Mind (Eds. A. (Eds. A. Kott Kott and W. and W. McEneaney McEneaney), pp 1 ), pp 1-

  • 22, Chapman & Hall/CRC Computer and Information

22, Chapman & Hall/CRC Computer and Information Science Series, Chapman & Hall/ Science Series, Chapman & Hall/CRC:Boca CRC:Boca Raton, 2006. Raton, 2006.

  • Santos, Eugene, Jr., Zhao,

Santos, Eugene, Jr., Zhao, Qunhua Qunhua, Johnson, Gregory, Nguyen, Hien, and Thompson, Paul, , Johnson, Gregory, Nguyen, Hien, and Thompson, Paul, “ “A Cognitive Framework For Information A Cognitive Framework For Information Gathering with Deception Detection For Intelligence Analysis Gathering with Deception Detection For Intelligence Analysis, ,” ” Proceedings of the 2005 International Conference on Intelligence Proceedings of the 2005 International Conference on Intelligence Analysis Analysis, , McClean McClean, VA, 2005. , VA, 2005.

  • Lehman, Lynn A., Krause, Lee S., Gilmour, Duane A., Santos, Euge

Lehman, Lynn A., Krause, Lee S., Gilmour, Duane A., Santos, Eugene, Jr., and Zhao, ne, Jr., and Zhao, Qunhua Qunhua “ “Intent Driven Adversarial Modeling Intent Driven Adversarial Modeling, ,” ” Proceedings of the Tenth International Proceedings of the Tenth International Command and Control Research and Technology Symposium: The Futur Command and Control Research and Technology Symposium: The Future of e of C2 C2, McLean, VA, 2005. , McLean, VA, 2005.

  • Santos, Eugene, Jr. and Johnson, Gregory,

Santos, Eugene, Jr. and Johnson, Gregory, “ “Toward Detecting Deception in Intelligent Systems Toward Detecting Deception in Intelligent Systems, ,” ” Proceedings of the SPIE: Defense & Security Proceedings of the SPIE: Defense & Security S Symposium ymposium, Vol. 5423, 131 , Vol. 5423, 131-

  • 140, Orlando, FL 2004.

140, Orlando, FL 2004.

  • Santos, Eugene, Jr. and

Santos, Eugene, Jr. and Negri Negri, , Allesandro Allesandro, , “ “Constructing Adversarial Models for Threat Intent Prediction and Constructing Adversarial Models for Threat Intent Prediction and Inferencing Inferencing, ,” ” Proceedings of Proceedings of the the SPIE Defense & Security Symposium SPIE Defense & Security Symposium, Vol. 5423, 77 , Vol. 5423, 77-

  • 88, Orlando, FL 2004.

88, Orlando, FL 2004.

  • McQueary

McQueary, Bruce, Krause, Lee, Santos, Eugene, Jr., Wang, , Bruce, Krause, Lee, Santos, Eugene, Jr., Wang, Hua Hua, and Zhao, , and Zhao, Qunhua Qunhua, , “ “Analysis and Visualization of Uncertainty in the Analysis and Visualization of Uncertainty in the Battlespace Battlespace, ,” ” Proceedings of the 16th Proceedings of the 16th IEEE International Conference on Tools with Artificial Intellige IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2004) nce (ICTAI 2004), 782 , 782-

  • 784, Boca Raton, FL,

784, Boca Raton, FL, 2004. 2004.

  • Santos, Eugene, Jr.,

Santos, Eugene, Jr., “ “A Cognitive Architecture for Adversary Intent Inferencing: Knowl A Cognitive Architecture for Adversary Intent Inferencing: Knowledge Structure and Computation edge Structure and Computation, ,” ” Proceedings of Proceedings of the SPIE: the SPIE: 17th Annual International Symposium on Aerospace/Defense Sensing 17th Annual International Symposium on Aerospace/Defense Sensing and Controls: and Controls: AeroSense AeroSense 2003 2003, Vol. 5091, 182 , Vol. 5091, 182-

  • 193, Orlando, FL,

193, Orlando, FL, 2003. 2003.

  • Surman

Surman, Joshua, Hillman, Robert, and Santos, Eugene, Jr., , Joshua, Hillman, Robert, and Santos, Eugene, Jr., “ “Adversarial Inferencing for Generating Dynamic Adversary Behavio Adversarial Inferencing for Generating Dynamic Adversary Behavior r, ,” ” Proceedings of Proceedings of the SPIE: 17th Annual International Symposium on Aerospace/Defen the SPIE: 17th Annual International Symposium on Aerospace/Defense Sensing and se Sensing and Controls: Controls: AeroSense AeroSense 2003 2003, Vol. 5091, 194 , Vol. 5091, 194-

  • 201,

201, Orlando, FL, 2003. Orlando, FL, 2003.

  • Brown, Scott M., Santos, Eugene, Jr., and Bell, Benjamin,

Brown, Scott M., Santos, Eugene, Jr., and Bell, Benjamin, “ “Knowledge Acquisition for Adversary Course of Action Prediction Knowledge Acquisition for Adversary Course of Action Prediction Models Models, ,” ” Proc of the AAAI 2002 Fall Symposium on Intent Inference for Use Proc of the AAAI 2002 Fall Symposium on Intent Inference for Users, Teams, and Adversaries rs, Teams, and Adversaries, Boston, MA, 2002. , Boston, MA, 2002.

  • Bell, Benjamin, Santos, Eugene, Jr., and Brown, Scott M.,

Bell, Benjamin, Santos, Eugene, Jr., and Brown, Scott M., “ “Making Adversary Decision Modeling Tractable with Intent Inferen Making Adversary Decision Modeling Tractable with Intent Inference and ce and Information Fusion Information Fusion, ,” ” Proceedings of the 11th Conference on Computer Generated Forces Proceedings of the 11th Conference on Computer Generated Forces and Behavioral Representation and Behavioral Representation, 535 , 535-

  • 542, Orlando, FL, 2002.

542, Orlando, FL, 2002.

slide-54
SLIDE 54

November 12, 2007 Dartmouth College 54 Distributed Information and Intelligence Analysis Group (DI2AG)

Select References for Others (User) Modeling Select References for Others (User) Modeling

  • Nguyen, Hien and Santos, Eugene, Jr.,

Nguyen, Hien and Santos, Eugene, Jr., “ “An Evaluation of the Accuracy of Capturing User Intent for Infor An Evaluation of the Accuracy of Capturing User Intent for Information Retrieval mation Retrieval, ,” ” Proc. Of the Intl Conf on AI

  • Proc. Of the Intl Conf on AI, Las Vegas,

, Las Vegas, NV, 2007. NV, 2007.

  • Nguyen, Hien and Santos, Eugene, Jr.,

Nguyen, Hien and Santos, Eugene, Jr., “ “Effects of Prior Knowledge on the Effectiveness of a Hybrid User Effects of Prior Knowledge on the Effectiveness of a Hybrid User Model for Information Retrieval Model for Information Retrieval, ,” ” Proceedings of the SPIE: Proceedings of the SPIE: Defense & Security Symposium Defense & Security Symposium, , Vol Vol: 6536, Orlando, FL, 2007. : 6536, Orlando, FL, 2007.

  • Nguyen, Hien, Santos, Eugene, Jr., Smith, Nathan, and

Nguyen, Hien, Santos, Eugene, Jr., Smith, Nathan, and Schuett Schuett, Aaron, , Aaron, “ “Hybrid User Model for Information Retrieval Hybrid User Model for Information Retrieval, ,” ” Proceedings of the AAAI Workshop Proceedings of the AAAI Workshop on Modeling

  • n Modeling

Others from Observations (MOO Others from Observations (MOO-

  • 06)

06), 61 , 61-

  • 76, Boston, MA, 2006.

76, Boston, MA, 2006.

  • Santos, Eugene, Jr., Zhao,

Santos, Eugene, Jr., Zhao, Qunhua Qunhua, Nguyen, Hien, and Wang, , Nguyen, Hien, and Wang, Hua Hua, , “ “Impacts of User Modeling on Personalization of Information Retri Impacts of User Modeling on Personalization of Information Retrieval: An Evaluation with eval: An Evaluation with Human Intelligence Analysts Human Intelligence Analysts, ,” ” in S. in S. Weibelzahl Weibelzahl, A. , A. Paramythis Paramythis, and J. , and J. Masthoff Masthoff (Eds.), (Eds.), Proceedings of the Fourth Workshop on the Proceedings of the Fourth Workshop on the Evaluation of Adaptive Systems Evaluation of Adaptive Systems (held in (held in conjunction with the 10th International Conference on User Model conjunction with the 10th International Conference on User Modeling (UM ing (UM’ ’05)), 27 05)), 27-

  • 36, Edinburgh, UK, 2005.

36, Edinburgh, UK, 2005.

  • Nguyen, Hien, Santos, Eugene, Jr., Zhao,

Nguyen, Hien, Santos, Eugene, Jr., Zhao, Qunhua Qunhua, and Wang, , and Wang, Hua Hua, , “ “Capturing User Intent for Information Retrieval Capturing User Intent for Information Retrieval, ,” ” Proceedings of the 48th Annual Meeting of the Proceedings of the 48th Annual Meeting of the Human Human Factors and Ergonomics Society (HFES 2004) Factors and Ergonomics Society (HFES 2004), 371 , 371-

  • 375, New Orleans, LA, 2004.

375, New Orleans, LA, 2004.

  • Nguyen, Hien, Santos, Eugene, Jr., Zhao,

Nguyen, Hien, Santos, Eugene, Jr., Zhao, Qunhua Qunhua, and Lee, Chester, , and Lee, Chester, “ “Evaluation of Effects on Retrieval Performance for an Adaptive U Evaluation of Effects on Retrieval Performance for an Adaptive User Model ser Model, ,” ” Adaptive Adaptive Hypermedia Hypermedia 2004: Workshop Proceedings 2004: Workshop Proceedings -

  • Part I

Part I, 193 , 193-

  • 202, Eindhoven, The Netherlands, 2004.

202, Eindhoven, The Netherlands, 2004.

  • McQueary

McQueary, Bruce, Krause, Lee, Santos, Eugene, Jr., Wang, , Bruce, Krause, Lee, Santos, Eugene, Jr., Wang, Hua Hua, and Zhao, , and Zhao, Qunhua Qunhua, , “ “Analysis and Visualization of Uncertainty in the Analysis and Visualization of Uncertainty in the Battlespace Battlespace, ,” ” Proceedings of the Proceedings of the 16th 16th IEEE International Conference on Tools with Artificial Intellige IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2004) nce (ICTAI 2004), 782 , 782-

  • 784, Boca Raton, FL, 2004.

784, Boca Raton, FL, 2004.

  • Santos, Eugene, Jr., Nguyen, Hien, Zhao,

Santos, Eugene, Jr., Nguyen, Hien, Zhao, Qunhua Qunhua, and Wang, , and Wang, Hua Hua, , “ “User User Modelling Modelling for Intent Prediction in Information Analysis for Intent Prediction in Information Analysis, ,” ” Proceedings of the 47th Annual Proceedings of the 47th Annual Meeting Meeting for the Human Factors and Ergonomics Society (HFES for the Human Factors and Ergonomics Society (HFES-

  • 03)

03), 1034 , 1034-

  • 1038, Denver, CO, 2003.

1038, Denver, CO, 2003.

  • Santos, Eugene, Jr., Nguyen, Hien, Zhao,

Santos, Eugene, Jr., Nguyen, Hien, Zhao, Qunhua Qunhua, and , and Pukinskis Pukinskis, Erik, , Erik, “ “Empirical Evaluation of Adaptive User Modeling in a Medical Info Empirical Evaluation of Adaptive User Modeling in a Medical Information Retrieval rmation Retrieval Application Application, ,” ” Lecture Notes in Artificial Intelligence 2702: User Modeling 20 Lecture Notes in Artificial Intelligence 2702: User Modeling 2003 03 (Eds. P. (Eds. P. Brusilovsky Brusilovsky, A. Corbett, and F. de , A. Corbett, and F. de Rosis Rosis), 292 ), 292-

  • 296, Springer, 2003.

296, Springer, 2003.

  • Santos, Eugene, Jr., Nguyen, Hien, and Brown, Scott M.,

Santos, Eugene, Jr., Nguyen, Hien, and Brown, Scott M., “ “Kavanah Kavanah: An Active User Interface Information Retrieval Application : An Active User Interface Information Retrieval Application, ,” ” Proceedings of the 2nd Asia Proceedings of the 2nd Asia-

  • Pacific

Pacific Conference on Intelligent Agent Technology Conference on Intelligent Agent Technology, 412 , 412-

  • 423, 2001.

423, 2001.

  • Santos, Eugene, Jr., Brown, Scott M.,

Santos, Eugene, Jr., Brown, Scott M., Lejter Lejter, , Moises Moises, , Ngai Ngai, Grace, Banks, Sheila B., and , Grace, Banks, Sheila B., and Stytz Stytz, Martin R., , Martin R., “ “Dynamic User Model Construction with Bayesian Dynamic User Model Construction with Bayesian Networks for Intelligent Information Queries Networks for Intelligent Information Queries, ,” ” Proceedings of the 12th International FLAIRS Proceedings of the 12th International FLAIRS Conference Conference, 3 , 3-

  • 7, Orlando, FL, 1999.

7, Orlando, FL, 1999.

  • Brown, Scott M., Santos, Eugene, Jr., and Banks, Sheila B.,

Brown, Scott M., Santos, Eugene, Jr., and Banks, Sheila B., “ “Utility theory Utility theory-

  • based user models for intelligent interface agents

based user models for intelligent interface agents, ,” ” Lecture Notes in Artificial Intelligence Lecture Notes in Artificial Intelligence 1418: Advances in Artificial Intelligence 1418: Advances in Artificial Intelligence – – AI AI ’ ’98 98, 378 , 378-

  • 392, Springer

392, Springer-

  • Verlag

Verlag, 1998. , 1998.

  • Brown, Scott M., Santos, Eugene, Jr., Banks, Sheila B., and Oxle

Brown, Scott M., Santos, Eugene, Jr., Banks, Sheila B., and Oxley, Mark, y, Mark, “ “Using Explicit Requirements and Metrics for Interface Agent User Using Explicit Requirements and Metrics for Interface Agent User Model Correction Model Correction, ,” ” Proceedings of the Second International Conference on Autonomous Proceedings of the Second International Conference on Autonomous Agents Agents, 1 , 1-

  • 7, Minneapolis/ St. Paul, MN, 1998.

7, Minneapolis/ St. Paul, MN, 1998.

  • Brown, Scott M., Harrington, Robert A., Santos, Eugene, Jr., and

Brown, Scott M., Harrington, Robert A., Santos, Eugene, Jr., and Banks, Sheila B., Banks, Sheila B., “ “User Models, Intelligent Interface Agents and Expert Systems User Models, Intelligent Interface Agents and Expert Systems, ,” ” Proceedings of Proceedings of the the Sixth International Conference on User Modeling Workshop on Embe Sixth International Conference on User Modeling Workshop on Embedding User dding User Models in Intelligent Applications Models in Intelligent Applications, 12 , 12-

  • 16, Sardinia, Italy, 1997.

16, Sardinia, Italy, 1997.

slide-55
SLIDE 55

November 12, 2007 Dartmouth College 55 Distributed Information and Intelligence Analysis Group (DI2AG)

Select References for Bayesian Knowledge Bases Select References for Bayesian Knowledge Bases

  • Santos, Eugene, Jr. and

Santos, Eugene, Jr. and Dinh Dinh, Hang T., , Hang T., “ “Automatic Knowledge Validation for Bayesian Knowledge Bases Automatic Knowledge Validation for Bayesian Knowledge Bases, ,” ” to appear to appear Data and Knowledge Engineering Data and Knowledge Engineering, 2007. , 2007.

  • Rosen,

Rosen, Tzachi Tzachi, , Shimony Shimony, Solomon , Solomon Eyal Eyal, and Santos, Eugene, Jr., , and Santos, Eugene, Jr., “ “Reasoning with BKBs Reasoning with BKBs – – Algorithms and Algorithms and Complexity Complexity, ,” ” Annals of Mathematics and Artificial Intelligence Annals of Mathematics and Artificial Intelligence 40(3 40(3-

  • 4)

4), 403 , 403-

  • 425, 2004.

425, 2004.

  • Santos, Eugene, Jr., Santos, Eugene S., and

Santos, Eugene, Jr., Santos, Eugene S., and Shimony Shimony, Solomon , Solomon Eyal Eyal, , “ “Implicitly Preserving Semantics During Implicitly Preserving Semantics During Incremental Knowledge Base Acquisition Under Uncertainty Incremental Knowledge Base Acquisition Under Uncertainty, ,” ” International Journal of Approximate Reasoning International Journal of Approximate Reasoning 33(1) 33(1), 71 , 71-

  • 94,

94, 2003. 2003.

  • Santos, Eugene, Jr.,

Santos, Eugene, Jr., “ “Verification and Validation of Knowledge Verification and Validation of Knowledge-

  • Bases Under Uncertainty

Bases Under Uncertainty, ,” ” Data and Knowledge Data and Knowledge Engineering Engineering 37 37, 307 , 307-

  • 329, 2001.

329, 2001.

  • Johnson, Gregory and Santos, Eugene, Jr.,

Johnson, Gregory and Santos, Eugene, Jr., “ “Generalizing Knowledge Representation Rules for Uncertain Knowle Generalizing Knowledge Representation Rules for Uncertain Knowledge dge, ,” ” Proceedings of the 13th International FLAIRS Conference Proceedings of the 13th International FLAIRS Conference, 186 , 186-

  • 190, Orlando, FL, 2000.

190, Orlando, FL, 2000.

  • Santos, Eugene, Jr. and Young, Joel D.,

Santos, Eugene, Jr. and Young, Joel D., “ “Probabilistic Temporal Networks: A unified framework for reasoni Probabilistic Temporal Networks: A unified framework for reasoning with ng with Time and Uncertainty Time and Uncertainty, ,” ” International Journal of International Journal of Approximate Reasoning Approximate Reasoning 20 20, 263 , 263-

  • 291, 1999.

291, 1999.

  • Santos, Eugene, Jr. and Santos, Eugene S.,

Santos, Eugene, Jr. and Santos, Eugene S., “ “A Framework for Building Knowledge A Framework for Building Knowledge-

  • Bases Under Uncertainty

Bases Under Uncertainty, ,” ” Journal Journal

  • f Experimental and Theoretical Artificial Intelligence
  • f Experimental and Theoretical Artificial Intelligence 11

11, 265 , 265-

  • 286, 1999.

286, 1999.

  • Santos, Eugene, Jr., and

Santos, Eugene, Jr., and Shimony Shimony, Solomon , Solomon Eyal Eyal, , “ “Deterministic Approximation of Marginal Probabilities in Deterministic Approximation of Marginal Probabilities in Bayes Bayes Nets Nets, ,” ” IEEE Transactions on Systems, Man, and IEEE Transactions on Systems, Man, and Cybernetics Part A: Systems and Humans Cybernetics Part A: Systems and Humans 28(4) 28(4), 377 , 377-

  • 393, 1998.

393, 1998.

  • Santos, Eugene, Jr.,

Santos, Eugene, Jr., “ “On Linear Potential Functions for Approximating Bayesian Computa On Linear Potential Functions for Approximating Bayesian Computations tions, ,” ” Journal of the ACM Journal of the ACM 43(3) 43(3), 399 , 399-

  • 430, 1996.

430, 1996.

  • Santos, Eugene, Jr.,

Santos, Eugene, Jr., “ “On the Generation of Alternative Explanations with Implications On the Generation of Alternative Explanations with Implications for Belief Revision for Belief Revision, ,” ” Proceedings Proceedings

  • f the Seventh Conference on Uncertainty in
  • f the Seventh Conference on Uncertainty in Artificial Intelligence

Artificial Intelligence, 337 , 337-

  • 347, Los Angeles, CA, 1991.

347, Los Angeles, CA, 1991.

  • Santos, Eugene S., and Santos, Eugene, Jr.,

Santos, Eugene S., and Santos, Eugene, Jr., “ “Reasoning with Uncertainty in a Knowledge Reasoning with Uncertainty in a Knowledge-

  • Based System

Based System, ,” ” Proceedings of Proceedings of the Seventeenth International Symposium on the Seventeenth International Symposium on Multiple Multiple-

  • Valued Logic

Valued Logic, 75 , 75-

  • 81, Boston, MA, 1987.

81, Boston, MA, 1987.