Está en la página 1de 23

AFOSR

Socio-Cultural Modeling and


Collective Behavior
14 March 2011 Dr. Joseph Lyons
Program Manager
Dr. Terrence Lyons
Program Manager
AFOSR/RSL
Air Force Office of Scientific Research
Distribution A: Approved for public release; distribution is unlimited. 88ABW-2011-1176
Portfolio within the Air Force

―Indeed, it is virtually impossible to disentangle culture


and language, which is why I am proud of Air Force
cultural modeling efforts in direct support of the OSD
Human, Social, Cultural and Behavioral modeling
program. Using the National Operational Environment
Model, the Air Force Research Laboratory and Air Force
Office of Scientific Research are helping to explore
cultural contours, representations of governance,
security institutions, critical infrastructure and social
well-being to model and forecast the human terrain, and
offer cultural insights to Joint analysis and planning.‖

--Secretary of the Air Force, Gen. Schwartz, January 26, 2011


2
Portfolio within the Air Force

―..the prominence of language skills and regional and


cultural appreciation will continue to grow,
facilitating vital face-to-face interaction for which
there is little substitute. The Air Force has made an
institutional commitment to advancing our
capabilities to address this reality, by designing the
building of partnerships and partner capacity as a
Service Core Function.‖
--Secretary of the Air Force, Gen. Schwartz
January 26, 2011

3
2011 AFOSR SPRING REVIEW
2313FX PORTFOLIO OVERVIEW
NAME: Dr. Joseph Lyons

BRIEF DESCRIPTION OF PORTFOLIO:


Sponsoring novel research utilizing multiple disciplines to discover the
foundations of group networks, collective behavior and socio-cultural
influence on beliefs, attitudes, and behavior of several populations, mostly
focused on non-state actors.

LIST SUB-AREAS IN PORTFOLIO:

•Understanding the science of behavioral/cultural influence

•Psychology of collective violence

•Computational social science and the study of social networks

4
Why Socio-Cultural Modeling?

• Understanding of the ―human terrain‖ is critical to Air Force


Operations (Schwartz, 2011)
– Building partnerships and partner capacity is a service core
function
• Influence Operations (AF IFO Roadmap, May 2008)
– Target audience analysis is a top IFO requirement
– IO doctrine emphasizes decision makers, but precision for
groups is needed (Emery, 2008)
• Quadrennial Defense Review (QDR)
– Succeeding in counterinsurgency (COIN)
– Stability and counterterrorism (CT) operations
– Effectively operating in cyberspace through the study of online
collective behavioral networks

5
Strategy

• Fund basic science in three core areas


– Understand culturally-driven methods to shape
behavior/attitudes
– Psychology of collective violence
– Understand and harness the power of social
networks
• Conduct workshops to bridge the gap between
warfighters and scientists
– ACC effects to influence workshop
• Leverage and build collaborations
– Other DoD entities (i.e., Minerva)
– Academia
– AFRL Scientists (Lab tasks)
6
Scientific Challenges

• Human behavior is inherently complex


– Collective behavior is difficult to predict
• Kitty Genovese – ―Bystander Effect‖
• Group decision making, conformity
• Emergent, contagious, - social media is an amplifier
– Culture as a construct is elusive
– Cultural behavior/attitudes are driven by a variety of factors
• Rational (i.e., economic models) may be inappropriate – may
result in adverse effects
• Data is very difficult to collect
– International project challenges
– Privacy issues
• Recipe for terrorism/violence is non-linear & complex
– Scope ranges from individual cognition, to social networks,
shared emotions, to societal conditions
7
Transformational Opportunities

One hundred victories in one hundred battles is not the most


skillful. Subduing the other’s military without battle is the most
skillful. (Sun Tzu, The Art of War; p.9)

• What if DoD had the foundational science to anticipate violence in


different groups/individuals?
– What if we could quantify this?
• What if the DoD had the capability to alter those courses of action
through influence tactics?
• What if DoD could forecast the secondary, and tertiary effects of
their operations on different groups?
• Social media/networks – great potential yet nascent in
understanding the capabilities and limitations

This is a high-risk, high-payoff domain


8
Other Organizations That Fund
Related Work
• NSF: Basic research tends to be aligned along disciplines,
AFOSR more multi-disciplinary
• DHS: Culturally-independent hostile intent prediction
• NIH: Focused on social network analysis
• ONR: Interests are close to AFOSR but well coordinated
• ARO: Training, mission rehearsal, face-to-face
negotiation/interaction, etc. Focus on near-term: ―something for
the soldier‖
• DTRA: Focused on WMD scenario
• DARPA: Similar interests, albeit more applied at times, they are a
useful transition partner for AFOSR
• OSD: HSCB Modeling, ops analysis, training (6.2-6.3) – related to
the effective strategic communication in AFOSR portfolio
• Air University Culture & Language Center: focus is mainly on
training related to language and culture
9
Interactions Within Air Force
Modeling, Cyber, Culture, & Lab Tasks

• ACC—Effects to Influence Workshop at William and Mary and Langley


– AFOSR-Sponsored event to help vector our research
• AFRL/RI: Modeling, mathematical assessment, and cyber domain
– DCT: Influence especially in the cyber domain, AFRL/RI Chief Scientist
helped to draft DCT BAA, RI review of white papers
– NOEM Lab Task (AFRL/RIEA): Holistic nation-state PMESII Modeling
Using a systems dynamics approach (Salerno)
• AFRL/RH: Lab tasks
– Cultural variation in precautionary mechanisms (Mort)
– Dynamic Trust Model (Stokes)
– Avenues of Influence (Sutton)
– Cross Cultural Aspects of Categories and Exemplars (Young)
– Whole body deception (Fullenkamp)
• AOARD & EOARD
10
Recent Transitions

• AF Minerva Princeton Project: Direct transition to COIN efforts for ISAF


• AFRL/RI: Modeling, mathematical assessment, and cyber domain
– Singled out by CSAF and AF Chief Scientist
– NOEM Lab Task Holistic nation-state PMESII modeling using a
systems dynamics approach
• CMU: Transition of social network analysis tools to DoD & Intel
– Used by JIOWC in Afghanistan and Iraq: *ORA for dynamic network
analysis was used to train military analysts and support missions;
training of USSTRATCOM and USSOCOM analysts

• Atran/Sageman: Research resulted in credible science-based policy


guidance including a study of evolution of militant networks, impacting
national policies-- U.S. Senate, U.S. State Dept., House of Lords

• Zucker: Transition of work inferring dominant intrinsic dimensions via


geometric harmonics and non-linear dimension reduction techniques to
Google

• UMD MURI: Direct transition of data extraction tools to intelligence


agencies 11
Select Projects

• Minerva projects
– Terrorism, Governance, & Development
– Emotion and Intergroup Relations
• Precautionary Mechanisms (Mort Lab Task)
• Dynamic Trust Model (Stokes Lab Task)
• Avenues of Influence (Sutton Lab Task)
• Cultural Dimensions (Saucier)
• Sacred Values (Atran)
• Computational Methods (Zucker)
• Reality Mining (Pentland)
• Young Investigator Program (Juarez)
12
Minerva: Terrorism, Governance,
and Development
• Eli Berman, (UCS), Jake Shapiro (Princeton), Col. Joe Felter, (Army
and Stanford)
• Objective:
– To understand the impact of policies, security assistance, and
other infrastructure (i.e., education, agriculture, etc.)
development on attitudes and behavior – (i.e., violence)
• Stability Operations are a major part of DoD strategy
• Previous research has shown that interventions sometimes have
unpredicted and deleterious outcomes
– Sometimes increased aid is related to heightened violence
• DoD investment huge > 30 Billion – impact unclear
• Currently working in Afghanistan, Iraq, Columbia, Northern Ireland,
and Philippines
• This research can support new policies based on empirical data
– Already having an influence on COIN strategy 13
Minerva: Emotion and Intergroup
Relations
• Dr. David Matsumoto (San Francisco State U), and Dr.
Mark Frank (U of Buffalo, SUNY)
• Objective:
– To understand the role of emotions (anger,
contempt, and disgust) in predicting violence
• Approach:
– Linguistic analysis of text
– Analysis of video content (key leaders)
– Controlled laboratory experiments
• Induce emotive states and measures outcomes
• Apply results to group scenarios to predict group
behavior
14
Precautionary Mechanisms (Mort
Lab Task)

• Focus is on how people react to potential threats, against which one


must take precautions
• Exploring cultural differences in threat detection
- Are there cultural differences in antecedents and reactions
• Leveraging collaborations/data collection venues in South Africa,
Israel, Ireland
15
Stokes Lab Task
Dynamic Trust Model
• Trust = the willingness of individuals to make themselves
vulnerable to others (Mayer et al., 1995)
• This process may vary by culture, but most studies have relied on
western samples/models
– Research suggests that collectivistic and individualistic
societies have different antecedents to trust (Branzei et al., 2007;
Huff & Kelley, 2003)

• But little experimental evidence exists


• This project will examine the relative weight of the three
trustworthiness dimensions (ability, benevolence, and integrity)
on trust in collectivistic versus individualistic cultures
– Utilize a novel experimental platform
• Leveraging a collaboration with AOARD
16
Sutton Lab Task
Avenues of Influence
• This project will examine the relative impact of social influence
tactics and affect-laden messages on different cultures
– Initial work will focus on the social proof bias (Cialdini, 2001)
– Penalties for violating social norms of category memberships
are tougher for collectivists who identify more with social
groups (Markus & Kitayama, 1991)
– Different cultures use and are effected by influence tactics
differently (Fu & Yukl, 2000)
– Emotional arousal via messages can influence behavior (Bator
& Cialdini, 2000; Schneider et al., 2009)

• Current project will leverage the NATO School in Germany


– 9K+ international students/yr
– Conduct a series of experiments
– Dr. Sutton is a Research Fellow at the NATO School 17
Cultural Dimensions

• PI: Gerard Saucier, University of Oregon


• Objective:
– Determine dimensionality of psychological measures of
personality, attitudes, values, and usefulness of these
dimensions for characterizing cultural differences
– Collect data from 45 countries
• Estimated N = 10K
– Should help operationalize culture
– Conduct psychometric analyses
• Factor structure
• Item Response Theory
• Measurement equivalence 18
Sacred Values

• PI: Dr. Scott Atran, John Jay College


• Background: Path to collective violence poorly understood
– Thus, attempts to thwart violence may be ineffective
• Objective: Understand the behavioral tendencies and cognitions
associated with collective violence
• Sacred values = moral imperatives drive behavior, attempts to
persuade behaviors that conflict with these are ineffective
– Sacred values may ―trump‖ economic considerations
– Attempts to alter scared values via economic means may have
negative consequences
• Group dynamics – path to violence largely influenced by
immediate peers/network
• Using test cases of inherent inter-group conflict
– Use combination of survey and interviews
19
Social Data Analysis Through Non-
linear Embedding
• PI: Steven Zucker (Yale)
• Background: Computational methods for mining large datasets is
needed for the social sciences
– Data may be incomplete or inaccurate
• Objective: to develop a novel
approach to organize knowledge
– Reveal knowledge structure
– Identify patterns amid complexity
– Based on diffusion maps
• Leverage existing social databases
– Might reveal intrinsic dimensions
– Support Intel analysts
– Identify trends – i.e., potential problems in advance
20
Reality Mining

• PI: Alex Pentland, MIT Media Lab


• Background: Social networks are a powerful source of social
movement/influence
– But how do we harness this power?
• Objective: leverage data mining and machine learning to model
network data over time
– Identify behavioral patterns – ―tribes‖
– Identify influence/leverage points
– Much human activity is predictable
• Seeks to understand the mechanisms that foster social
movement – contagion among individuals
– Tactics that motivate task activity and recruitment of
additional resources effective
– Time Critcial Social Mobilization: The DARPA network
challenge winning strategy – accepted by Science Magazine
21
Young Investigator Program
• PI: Ruben Juarez (University of Hawaii)
• Background: Online communities are pervasive but the formation
of these communities is poorly understood
– What drives a coalition to form and how can we model that?
• Objective: To create a model using game theory to predict the
formation of coalitions in distributed communities
– Power and altruism effects
– Validate this model using experimental methods and
culturally-diverse participants
• Use a variety of scenarios (i.e., tournaments) to test the influence
of various agent-based parameters
– Power = multidimensional
– Shared outcomes
– Costs for success
22
Program Trends

Experimental Social Science


Field Studies in Social Science
Study of Social Networks
Computational Methods
Agent-based Modeling

23

También podría gustarte