Institute for2009 Collaborative Innovation
Institute for 2009 I I Patterns of Interaction Collaborative & - - PowerPoint PPT Presentation
Institute for 2009 I I Patterns of Interaction Collaborative & - - PowerPoint PPT Presentation
Institute for 2009 I I Patterns of Interaction Collaborative & Perception of Intent Innovation Michael OBrien Kok Keng Tan Tal Sack Matthieu Branlat Lauren Grinstein Michael Smith Institute for 2009 I I Intent from COMINT Patterns
Understanding intent is hard
because it is complex. Looking at the patterns involved in interpretation of signals, and interactions between agents, helps structure things in a meaningful way.
Intent from COMINT
Challenges to 3rd Party Observers in Inferring Goals and Constraints
Institute for2009 Collaborative Innovation
I I
Patterns of Interaction & Perception of Intent
HEY YOU!
Intent from COMINT
Challenges to 3rd Party Observers in Inferring Goals and Constraints
Institute for2009 Collaborative Innovation
I I
Patterns of Interaction & Perception of Intent
Framework for Understanding Issues of Reading Intent Case Studies of Intent Misperception
- Ruby Ridge
- Yom Kippur/October 1973 War
Lab Studies of Interaction and 3rd Party Analysis
- Cooperative/Competitive Game
- Drone Mission Negotiation Scenario
Directions for Support (Design Seeds)
What is Intent?
Powerful explanatory mechanism for forecasting behavior So useful we use it everywhere, even when it might not be really accurate
What is Intent?
Evolving Sets
- f Goals
Opportunities Constraints Changing State of World Evolving Understanding
- f World
Actions
- f Agents
What is Intent?
Multiple Roles
Direct Perception Projection Perspective Taking
Continuum of Ways to Read Intent
Perspective Taking
Continuum of Ways to Read Intent
Experts
- Social Scientists
- Historians
- Global Leaders
- Consider immediate and
larger contexts
- Use mental simulation
- Apply domain knowledge
when assessing information
Framework for Understanding Issues of Reading Intent
Political Psychology models of signaling and (mis)perception Cognitive Systems Engineering models of distributed anomaly response and process tracing
⚔ $
$
...
Signaling & (Mis)Perception
Patterns of Misperception of Intent
- ver-
homogenization
- ver-centralization;
presumption of coherence, planning, unitary action fundamental attribution error stereotyping
- ver-estimation
- f own
importance predisposition for causal explanations misperception of others' perceptions of us
- ver-confidence
(in reading & signaling intent) lack of empathy projection, mirroring view others as static, unchanging assumption of proportionality in effort and importance
Time Nature of Constraints
Causal Intentional
Data Events Analyses Stance Goals Plans Activities Expectations Data
get integrated spawn (diagnostic/ predictive) support resolves set set context shape structure
Distributed Process Monitoring & Replanning
Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention
More Detailed Model of Signaling & Perception
Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention
More Detailed Model of Signaling & Perception
Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention Data Events Analysis Stance Goals Plans Activities Expectations- Attention Expectations- Attention
More Detailed Model of Signaling & Perception
Methodological issues
- How can we study the process of inferring
intent?
- How can we model dynamics of behavior that
reveal underlying dimensions of intent (stance, mindset, goals)?
Studies
- Case studies
- Ruby Ridge
- Yom Kippur
- Experiments
- Collaborative/Competitive Game
- Negotiation Scenario
Studies
- Case studies
- Ruby Ridge
- Yom Kippur ➠ poster session
- Experiments
- Collaborative/Competitive Game
- Negotiation Task ➠ poster session
Case Study: Ruby Ridge
Motivations
- Purposes
- Analyze dynamics of interaction
- Explore representations
- Scope
- Available documentation
- Multiple perspectives
- Implement process-tracing methodology
Case Study: Ruby Ridge
Result: Perspectives
Initial Shooting Marshall killed Weaver will resist arrest Weaver is a serious threat Protect society from Weaver Get additional resources Get FBI involved General attention to white supremacists, militia movements, conspiracy theorists... File built on Weaver (Aryan Nations) Initial troubles with Weaver Threatening letters received by judge(s)
LAW ENFORCEMENT / FBI
Case Study: Ruby Ridge
Result: Perspectives
Son killed Police will shoot them Increased animosity and mistrust of government Protect themselves from law enforcement Lock themselves in cabin Check movement
- n property
Initial Shooting Moved away from corrupt society Autonomous life Tricked by local police Growing animosity and mistrust against governmental agencies Knew he was in trouble with the law Thought he would not have a fair trial Law enforcement will come with hostile purposes
WEAVER FAMILY
Case Study: Ruby Ridge
Result: Patterns
time alignment conflict conflict
Experiment
Collaborative/Competitive Game
- Purposes
- Analyze dynamics of interaction from multiple
perspectives
- Rules designed to produce cooperative and
competitive behaviors
- Experiment
- Participants verbalize
- Activity is captured, then analyzed by outside observer
Experiment
Collaborative/Competitive Game
Round 1
Experiment
Collaborative/Competitive Game
Round 2
Inferring intent is difficult
- We are very attuned to it...
- we are very good at it in many situations
- we will infer intent
- ... but it is easy to misperceive intent
- people’s behavior is not always deliberate
- history of interactions creates powerful expectations
- “language” of interaction can be ambiguous
- Especially difficult for third party observers
Patterns of Misperceptions of Intent
- ver-
homogenization
- ver-centralization;
presumption of coherence, planning, unitary action fundamental attribution error stereotyping
- ver-estimation
- f own
importance predisposition for causal explanations misperception of others' perceptions of us
- ver-confidence
(in reading & signaling intent) lack of empathy projection, mirroring view others as static, unchanging assumption of proportionality in effort and importance
Perspective Taking
- Mental simulation
- Structure (patterns, dynamics)
- Reflection
What supports intent reading
Directions for Support
Rigor as Framework for Reflection Archetypical Patterns as Hypothesis Space Anchors Post-Hoc Automated Critique
Archetypical Patterns as Hypothesis Space Anchors
Escalation Balancing process Rapprochement
Rivalry
Misperceptions
Recognition of misread signal
Post-Hoc Automated Critique
Patterns of Misperceptions of Intent
- ver-
homogenization
- ver-centralization;
presumption of coherence, planning, unitary action fundamental attribution error stereotyping
- ver-estimation
- f own
importance predisposition for causal explanations misperception of others' perceptions of us
- ver-confidence
(in reading & signaling intent) lack of empathy projection, mirroring view others as static, unchanging assumption of proportionality in effort and importance
Understanding intent is hard
because it is complex. Looking at the patterns involved in interpretation of signals, and interactions between agents, helps structure things in a meaningful way.
Intent from COMINT
Challenges to 3rd Party Observers in Inferring Goals and Constraints
Institute for2009 Collaborative Innovation
I I
Patterns of Interaction & Perception of Intent