Marital Agent TraitBased Emotion System

1
Marital Agent Trait-BasedEmotion System

• System collects information about a pre-marital
  couple.
• Use questionnaires to determine
• Personality- IPIP NEO
• Relationship concerns (Gottman)
• Goals with emotion vectors
• Model couple in negotiation

2
Enhancing Marital Stability

• Vast majority in US marry (83)
• 43 of first marriages end in divorce
• marital unhappiness negatively affects adults and
  children
• health
• loss of social and financial support
• psychopathology (e.g. depression, anxiety)
• lower work productivity
• poverty
• alienation
• Societal costs are enormous

3
My wife and I had words, but I never got to use
mine. -Fibber McGee

• The basic cause of divorce today is faulty mate
  selection and inadequate preparation for a
  companionship type of marriage. Waller

4
Variables for prediction of divorce

• characteristics of each person
• combinations of couple characteristics
• compatibility in areas such as personal values
  and leisure interests

• dyadic characteristics
• length of courtship
• feelings for partner
• evaluation of relationship
• social network
• cohabitation
• premarital birth
• communication patterns

5
Communication Important

• Finding right partner in terms of all variables
  is very slim conflict is inevitable!
• Communication patterns can be changed whereas
  race, social class, feelings for partner,etc.
  typically cannot!
• GOAL Improve communication by modeling couples
  communication

6
Benefits to Couple

• See themselves ah ha moment
• Demonstrate concerns
• Give multiple forms of feedback
• visual
• written evaluation
• interactive
• model best practices/suggest changes
• Accessible and non-threatening

7
Contributions

• Model participants themselves
• Use self-reporting to assess personality, goals,
  concerns
• Create an Algebra of Emotion- unified approach
• Knowledge engineering

8

• Alice Want
• to go to a
• movie?
• joy,fear
• Bob What
• movie?
• joy,interest

Symbolic representation of request/inform/accept/r
eject/counter Emotions explicitly
conveyed. Emotional transforms filter, combine,
enhance, degrade
9
The System

• Library of plans indexed by goals/personalities
• Goals extracted from questionnaire
• select appropriate plan
• Personality restricts actions through micro
  theory selection. (Open Cyc)
• Emotions generated from goal achievement or
  likelihood/unlikelihood of goal achievement.
• Emotions are used to select means of achieving
  subgoals
• Events may motivate plan repair simplistic
  search through conditional alternatives

10
The Details

• Agent is scripted using Prolog-like knowledge
  base
• Some goals are actively pursued
  (e.g. have stable relationship)
• Other goals are hoped for but not actively
  pursued (e.g. be appreciated)
• Goals are populated with emotion vector for each
  subgoal.
• Emotions combined, filtered, degraded as
  specified by parameters of algebra of emotion.

11
Personality Effects

• Which emotions are felt
• Intensity of emotions
• Which actions are selected
• Which attributes the person is willing/able to
  modify
• Accurate modeling is important confidence advice
  is appropriate

12
Where do numbers come from?

• Personality surveys valid for self-reporting?
• Couple surveys typically done by observation via
  trained professionals. Still accurate for
  self-reporting?
• What factors control intensity of response?
• What actions choices are appropriate for history,
  personality, emotion?
• Knowledge engineering is critical!