A COMPUTATIONAL MODEL OF COLLABORATIVE NEGOTIATION BASED ON THE - - PowerPoint PPT Presentation

a computational model of collaborative negotiation based
SMART_READER_LITE
LIVE PREVIEW

A COMPUTATIONAL MODEL OF COLLABORATIVE NEGOTIATION BASED ON THE - - PowerPoint PPT Presentation

A COMPUTATIONAL MODEL OF COLLABORATIVE NEGOTIATION BASED ON THE INTERPERSONAL RELATION OF DOMINANCE Presented by : Lydia OULD OUALI [LIMSI-CNRS / UPSUD] Directors : Nicolas Sabouret [LIMSI-CNRS / UPSUD] Charles Rich [CS / WPI]


slide-1
SLIDE 1

A COMPUTATIONAL MODEL OF COLLABORATIVE NEGOTIATION BASED ON THE INTERPERSONAL RELATION OF DOMINANCE

Presented by :

  • Lydia OULD OUALI [LIMSI-CNRS / UPSUD]

Directors :

  • Nicolas Sabouret [LIMSI-CNRS / UPSUD]
  • Charles Rich [CS / WPI]
slide-2
SLIDE 2

AlwaysOn Sidner et al 14 Smith et al 10

Companion

REA Bickmore et al 02

Partner

Tutor

SimSensei DeVault et al 14 SimCoach Rizzo et al 11

CONTEXT: CONVERSATIONAL AGENTS

Agents sociaux

Tardis Sabouret et al 15

2

Guide

Alice Glas et al 16 Lea Potdevin 18

slide-3
SLIDE 3

COLLABORATIVE NEGOTIATION

[Lewicki et al 11 , Wollkind et al 04]

Facilitate mutual understanding Agreement making Generation of new ideas

INTERPERSONAL RELATIONSHIP

[Tiedens 03]

Impact of interpersonal relation

  • n the negotiation

Focus on DOMINANCE

Nurses collaborating with MIT Nao Robot Rethink robotics

CONTEXT: CONVERSATIONAL AGENTS

TASK ORIENTED INTERACTION

3

slide-4
SLIDE 4

COLLABORATIVE NEGOTIATION

Interpersonal relation of DOMINANCE

PROBLEMATIC

Decisional model based on behaviors of dominance

4

slide-5
SLIDE 5

Concept of dominance and state of art Computational model of collaborative negotiation Decisional model controlled by dominance behaviors Study and validation: perception of dominance in negotiation Analyze of the partner: Model of ToM Study and validation : complementarity vs similarity Conclusion and perspectives

PLAN

1 2 3 4 5 6 7

5

slide-6
SLIDE 6

CONTEXT: CONCEPT OF DOMINANCE

Personality trait Personality trait DOMINANCE Interpersonal relation Interpersonal relation Social status Social status

6

slide-7
SLIDE 7

CONTEXT: CONCEPT OF DOMINANCE

Personality trait Personality trait DOMINANCE Interpersonal relation Interpersonal relation Social status Social status

7

[Emmons & McAdams 91]

  • Individual trait
  • Ability to exert power
  • Want to control others
  • Desire of celebrity
slide-8
SLIDE 8

CONTEXT: CONCEPT OF DOMINANCE

Personality trait Personality trait DOMINANCE Interpersonal relation Interpersonal relation Social status Social status

  • Position in social hierarchy [Liska 90]
  • Legitimate expression of power and dominance
  • Vested dominance: Birth order, kinship or privileges
  • e.g: college professor, judge, or corporate president
  • Earned dominance: Based on relational or social skill

8

slide-9
SLIDE 9

CONTEXT: CONCEPT OF DOMINANCE

Personality trait Personality trait DOMINANCE Interpersonal relation Interpersonal relation Social status Social status

  • Ability to exert power [Burgoon & Dunbar 98]
  • Power: Ability to influence the behavior of another person [Burgoon et al 98]

Control attempts by one individual are accepted by the interactional partner

[Burgoon & Dunbar 98]

9

slide-10
SLIDE 10

CONTEXT: EXPRESSION OF DOMINANCE

Verbal behaviors Verbal behaviors Non-verbal behaviors Non-verbal behaviors

kinesic cues [Posture, body movement, gestures, facial expressions] Speech frequency , interruption Initiate interaction and behaviors Communication strategies

  • 1. Direct strategies
  • 2. Indirect strategies

High vocal tone and frequency Facial expression

10

slide-11
SLIDE 11

Verbal behaviors Verbal behaviors Non verbal behaviors Non verbal behaviors

CONTEXT: EXPRESSION OF DOMINANCE

Style of ommunication

  • Direct communication
  • Indirect communication

Vocal control

Turn taking, interruption, amplitude, disfluencies

Vocal tone and frequency

[Gebhard et al 14]

Gaze orientation

visual dominance ratio, gaze

[Gebhard et al 08 ; Bee et al 10] [Bee et al 10]

Kinesics cues

Posture, body movement, gestures

[Beeet al 10 ; Strassmann et al 16]

11

[Frieze et McHugh 92]

Strategic behavior s

Facial expressions

[Ravenet et al 15]

slide-12
SLIDE 12

NEGOTIATION STRATEGY: BEHAVIORS OF DOMINANCE

DOMINANT SUBMISSIVE BEHAVIOURS OF DOMINANCE IN NEGOTIATION

Level of demand and concession [Dedreu et al 95]

Dominance is associated to a high level of demand and a low level of concessions

Self vs other [Fiske 93, DeDreu et al 95]

Dominant individuals are self-centered and only interested in satisfying their own preferences.

2 1

Lead of the negotiation [Dedreu & VanKleef 04]

Dominant individuals tends to make the first move Control of the flow of the negotiation

3

12

slide-13
SLIDE 13

MODEL OF COLLABORATIVE NEGOTIATION: DECISION BASED ON BEHAVIOR OF DOMINANCE

Implementation of principles of dominance behaviors in a conversational agent

13

slide-14
SLIDE 14

Communication UtteranceSelf Utteranceother Adaptation

OVERVIEW OF THE MODEL OF NEGOTIATION

Mental model

Pow Preferences

1

Decisional model Mental model Pow Preferences Decisional model

Model of the other

Mental model Pow Preferences Decisional model

2 3

Pow

14

slide-15
SLIDE 15

15

Communication

UtteranceSelf Utteranceother Mental model

Pow Preferences

NEGOTIATION DOMAIN: MODEL BASED ON PREFERENCES

slide-16
SLIDE 16

MENTAL MODEL: PREFERENCES

0.16 0.5 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

Goal choose an option [eg : Restaurant]

Option = {C1 , C2 , …, Cn} Criterion = {v1 , v2 , …, vn}

Domain model Domain model

Eg : Restaurant = {cuisine, Price, ambiance} Eg : Cuisine = {Indian, French, Italian, …}

Preferences Preferences

  • Binary relation of

preferences

  • Partial order

Score of satisfaction Sat[v] Inverse of the number of ancestors

16

0.33

slide-17
SLIDE 17

I like v I don’t like v

Computational model

MENTAL MODEL: COMMUNICATION

Share preferences Negotiation

UtteranceSelf Utteranceother Share a preference State Preference[X] Ask for a preference Ask Preference[X] Mental model Shared preferences Other’s preferences

17

slide-18
SLIDE 18

Computational model

MENTAL MODEL: COMMUNICATION

Share preferences Negotiation

UtteranceSelf Utteranceother Share a preference State Preference[X] Ask for a preference Ask Preference[X] Mental model Shared Proposals Make a proposal Propose[X] Reject a proposal Reject[X] Accept a proposal Accept[X]

P :Open, T : Accepted, R : Rejeted

18

slide-19
SLIDE 19

NEGOTIATION STRATEGY: BEHAVIORS OF DOMINANCE

BEHAVIOURS OF DOMINANCE IN NEGOTIATION

Level of demand and concession [Dedreu et al 95]

Dominance is associated to a high level of demand and a low level of concessions

Self vs other [Fiske 93, DeDreu et al 95]

Dominant individuals are self-centered and only interested in satisfying their own preferences.

2 1

Lead of the negotiation [Dedreu & VanKleef 04]

Dominant individuals tends to make the first move Control of the flow of the negotiation

3

19

slide-20
SLIDE 20

MENTAL MODEL: SATISFIABILITY [LEVEL OF DEMAND]

Preferences Power Initial value of dominance : POW [O,1]

1 0.7

  • S: Set of satisfiable values

Express the liking

  • f the agent

[StatePreference] Communication

20

0.16 0.5 0.33 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

slide-21
SLIDE 21

MENTAL MODEL: SATISFIABILITY [LEVEL OF DEMAND]

Preferences Power Initial value of dominance : POW [O,1]

1 0.6

  • S: Set of satisfiable values

Express the liking

  • f the agent

[StatePreference] Communication

21

0.16 0.5 0.33 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

slide-22
SLIDE 22

MENTAL MODEL: SATISFIABILITY [LEVEL OF DEMAND]

Preferences Power Initial value of dominance : POW [O,1]

1 0.6

Values the agent doesn’t like Values the agent like

Communication I like French cuisine I don’t like Korean cuisine

22

0.16 0.5 0.33 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

slide-23
SLIDE 23

MENTAL MODEL: ACCEPTABILITY [CONCESSION]

Preferences Power Initial value of dominance : POW [O,1]

1 0.6 Express Proposals [Propose] [Accept] [Reject] Communication

23

0.16 0.5 0.33 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

slide-24
SLIDE 24

MENTAL MODEL: ACCEPTABILITY [CONCESSION]

Preferences Power Initial value of dominance : POW [O,1]

  • 2. Acceptability

t Self[pow,t] τ powa powb

1 0.6

Acceptable values Concessions

24

0.16 0.5 0.33 0.66 0.66 0.83 1

French Italian Indian Korean Mexican Japanese Chinese

slide-25
SLIDE 25

MENTAL MODEL: DECISION ABOUT PROPOSALS [SELF VS OTHER]

25

self 1 - self Implementation: Choose the value of a proposal Take into account self preferences and other preferences High-dominance self 1 - self Low-dominance

slide-26
SLIDE 26

 Decision rules  Define a priority in the choice of the utterance

MENTAL MODEL: UTTERANCE CHOICE [LEAD OF DIALOGUE]

High-dominance Negotiation acts

(Propose, CounterPropose)

Low-dominance Information acts (AskPreference)

26

slide-27
SLIDE 27

A: "Let's go to a Chinese restaurant." B: "I don't like Chinese restaurants, let's choose something else." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like Italian restaurants?" A: "I don't like Italian restaurants." B: "Do you like French restaurants?" A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like French restaurants?" A: "I don't like French restaurants." B: "Do you like Korean restaurants?" A: "Let's go to a cheap restaurant." B: "Okay, let's go to a cheap restaurant." A: "Let's go to a restaurant on the south side." B: "Okay, let's go to a restaurant on the south side." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Okay, let's go to the Shanghai restaurant.“

Pow(A) = 0.9, Pow(B) = 0.4

MENTAL MODEL: UTTERANCE CHOICE

Agent A Dominant Agent B Submissive

27

slide-28
SLIDE 28

Level of demand & concessions Level of demand & concessions

MENTAL MODEL: UTTERANCE CHOICE

A: "Let's go to a Chinese restaurant." B: "I don't like Chinese restaurants, let's choose something else." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like Italian restaurants?" A: "I don't like Italian restaurants." B: "Do you like French restaurants?" A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like French restaurants?" A: "I don't like French restaurants." B: "Do you like Korean restaurants?" A: "Let's go to a cheap restaurant." B: "Okay, let's go to a cheap restaurant." A: "Let's go to a restaurant on the south side." B: "Okay, let's go to a restaurant on the south side." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Okay, let's go to the Shanghai restaurant.“ Pow(A) = 0.9, Pow(B) = 0.4 Agent A Dominant Agent B Submissive

28

slide-29
SLIDE 29

self-centeredness self-centeredness

Pow(A) = 0.9, Pow(B) = 0.4 Agent A Dominant Agent B Submissive

MENTAL MODEL: UTTERANCE CHOICE

A: "Let's go to a Chinese restaurant." B: "I don't like Chinese restaurants, let's choose something else." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like Italian restaurants?" A: "I don't like Italian restaurants." B: "Do you like French restaurants?" A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like French restaurants?" A: "I don't like French restaurants." B: "Do you like Korean restaurants?" A: "Let's go to a cheap restaurant." B: "Okay, let's go to a cheap restaurant." A: "Let's go to a restaurant on the south side." B: "Okay, let's go to a restaurant on the south side." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Okay, let's go to the Shanghai restaurant.“

29

slide-30
SLIDE 30

A: "I don't like French restaurants." B: "Do you like Korean restaurants?" A: "Let's go to a cheap restaurant." B: "Okay, let's go to a cheap restaurant." A: "Let's go to a restaurant on the south side." B: "Okay, let's go to a restaurant on the south side." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Okay, let's go to the Shanghai restaurant.“

Lead of dialogue

Pow(A) = 0.9, Pow(B) = 0.4 Agent A Dominant Agent B Submissive

MENTAL MODEL: UTTERANCE CHOICE

A: "Let's go to a Chinese restaurant." B: "I don't like Chinese restaurants, let's choose something else." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like Italian restaurants?" A: "I don't like Italian restaurants." B: "Do you like French restaurants?" A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like French restaurants?"

30

slide-31
SLIDE 31

Lead of dialogue

Pow(A) = 0.9, Pow(B) = 0.4 Agent A Dominant Agent B Submissive

MENTAL MODEL: UTTERANCE CHOICE

A: "Let's go to a Chinese restaurant." B: "I don't like Chinese restaurants, let's choose something else." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like Italian restaurants?" A: "I don't like Italian restaurants." B: "Do you like French restaurants?" A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Do you like French restaurants?" A: "I don't like French restaurants." B: "Do you like Korean restaurants?" A: "Let's go to a cheap restaurant." B: "Okay, let's go to a cheap restaurant." A: "Let's go to a restaurant on the south side." B: "Okay, let's go to a restaurant on the south side." A: "Let's go to the Shanghai. It's a quiet, cheap Chinese restaurant on the south side." B: "Okay, let's go to the Shanghai restaurant.“

31

slide-32
SLIDE 32

EVALUATION: PERCEPTION OF BEHAVIORS OF DOMINANCE

32

slide-33
SLIDE 33

MENTAL MODEL: AGENT/AGENT STUDY

H1 H2 H3 H4

The higher-power agent will more strongly be perceived as self-centered than the lower-power agent The lower-power agent will be more strongly perceived as making larger concessions than the higher-power agent The higher-power agent will more strongly be perceived as demanding than the lower-power agent The higher-power agent will more strongly be perceived as taking the lead in the negotiation than the lower-power agent

33

Principle 2 Principle 1 Principle 3

slide-34
SLIDE 34

MENTAL MODEL: AGENT/AGENT STUDY

CONDITIONS PROCEDURE

  • Init of power

Pow(A) = 0.9, Pow(B) = 0.4 Pow(A) = 0.7, Pow(B) = 0.4 Pow(A) = 0.7, Pow(B) = 0.2

  • Agent preferences.

Similar preferences Different preferences External judges evaluate both agent behaviors during their negotiation. A between-subject study on the online site CrowdFlower.com. Fill in a questionnaire about each principle Agents described as two friends negotiating about restaurant where to have dinner. Total participants: 120

34

  • 4 dialogues generated
slide-35
SLIDE 35

MENTAL MODEL: AGENT/AGENT STUDY

H1: Self centeredness H2: Concessions

  • Agent A is more self-centered and makes less concessions.
  • Agent B tries to find the best trade-off for both parties, and is able to make larger concessions.

35

slide-36
SLIDE 36

MENTAL MODEL: AGENT/AGENT STUDY

H3: Level of demand H4: Lead of the dialogue

  • Agent A is more demanding than agent B.
  • Agent A is the one who leads the dialogue. (even in the similar condition)

36

slide-37
SLIDE 37

MENTAL MODEL: AGENT/AGENT STUDY

H1 H2 H3 H4

The higher-power agent will more strongly be perceived as

self-centered than the lower-power agent

The lower-power agent will be more strongly perceived as making larger concessions than the higher-power agent The higher-power agent will more strongly be perceived as

demanding than the lower-power agent

The higher-power agent will more strongly be perceived as taking the lead in the negotiation than the lower- power agent

Perception of power’s behaviors in Human/Agent negotiation

UttSelf Uttother

37

slide-38
SLIDE 38

UttSelf Uttother

MENTAL MODEL: AGENT/HUMAN STUDY

CONDITIONS PROCEDURE

  • Init agents

Bob : Dominant behaviors (Pow =0,8) Arthur: Submissive behaviors (Pow =0,4)

  • Presentation of the negotiation

task: Choose a restaurant using criteria : { Cuisine, Price , Athmosphere, Location} Total : 420 restaurants

  • Negotiate with each agent
  • Fill in a survey

Total participants: 40

38

PaParler de l’interfacer

EZ1

slide-39
SLIDE 39

MENTAL MODEL: AGENT/HUMAN STUDY

39

1 2 3 4 5 SelfVsOther Demand Concessions Lead Bob Arthur * * * P =0.055

slide-40
SLIDE 40

MENTAL MODEL: PERCEPTION OF BEHAVIORS OF DOMINANCE

SIMULATION OF INTERPERSONAL RELATION Adapt the behaviors of the agent in

  • rder to simulate an interpersonal

relation BEHAVIORS OF DOMINANCE Behaviors of dominance correctly perceived in agent/agent and human/agent interaction

CAN WE SIMULATE THE USER’S BEHAVIORS USING THE AGENT MENTAL MODEL

40

slide-41
SLIDE 41

TOM: EVALUATION OF THE PARTNER’S BEHAVIORS

41

slide-42
SLIDE 42

SIMULATION THEORY Revision

. . .

Pow 1 Utteranceother

MODEL OF THE OTHER: NAÏVE APPROACH

. . .

SELF DECISIONAL MODEL

Utterance1

. . .

UtteranceN

Possible powi

Compare Pow N

X X X

42

slide-43
SLIDE 43

SIMULATION THEORY Revision

. . .

Pow 1 Utteranceother

MODEL OF THE OTHER: NAÏVE APPROACH

. . .

SELF DECISIONAL MODEL

Utterance1

. . .

UtteranceN

Possible powi

Compare Pow N

X X X

COMBINATORY EXPLOSION

43

slide-44
SLIDE 44

MODEL OF THE OTHER: REASONING WITH UNCERTAINTY

Total order on preferences For each Powi Compute nb Sat values Pow = 0.6 |S| = 4 Extract hypotheses Set of possible sat values 1. S = {Jap,It ,Ch, Fr} 2. S= {Ind, Mex, Kor, Ch} 3. S ={ Fr, It, Ch, Jap} 4. . . . Adapt decisional model Satisfiability Acceptability Utterance type Revision

1 2 3 4 5

SIMULATION THEORY

Highest number of accurate hypotheses

0.16 0.33 0.66 0.5 0.83 1 French Italian Indian Korean Mexican Japanese Chinese

44

slide-45
SLIDE 45

EVALUATION: REASONING WITH UNCERTAINTY

Goal

  • Evaluate the accuracy of predictions
  • Time execution for each prediction

Utt1 Utt2

Dominant Submissive

Conditions:

1 2

Initial value of power

Submissive 0.3 0.4 0.5 Dominant 0.6 0.7 0.8 0.9

Initial preferences

Domain size Nb possible hypotheses Small 1296 Medium 4.14 x 105 Large 2.6 x 109

Total dialogues : 1080

45

slide-46
SLIDE 46

RESULTS: ACCURACY OF PREDICTIONS

46

  • 2. Correct predictions
  • 1. Accuracy
  • 3. Convergence

Powother True Pow Comparison

0.55 Base line Model MEA RMSE Base line 0.15 0.2 ToM 0.07 0.12

97.4%

Correct prediction 2.53 RMSE < 0.14 3.25 Final prediction 3.79

slide-47
SLIDE 47

MODEL OF COLLABORATIVE NEGOTIATION: SIMULATE AN INTERPERSONAL RELATION OF DOMINANCE

47

slide-48
SLIDE 48

STUDY: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

COMPLEMENTARITY

Increase coordination Value creation is better Better common gain Improve the comfort felt and create mutual liking

SIMILARITY

People are attracted to partners who express similar behaviors Increase value creation Improve attractiveness

48

Tiendens et al, 03 Olekalns & Smith, 13

slide-49
SLIDE 49

STUDY: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

CONDITIONS

ARTHUR : Similar BOB: Complementary KEVIN: Neutral

INITIAL BEHAVIORS

  • POW = 0.55
  • Decision among 630 restaurants
  • Manipulation of preferences:
  • Distance(Agent, User) > 0.7
  • Distance(Agent, Agent_i) > 0.35
  • Most liked values are different

49

slide-50
SLIDE 50

STUDY: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H1 H2 H3 H4

The behaviors of dominance expressed by the agent; whether complementary or similar;

are correctly perceived by the participants

Negotiation converges more quickly when negotiators establish a complementary relation of dominance Participants feel more comfortable with a partner who expresses complementary behaviors of dominance Complementarity in the dominance relation increases appreciation between negotiators

H5

Negotiators achieve a greater common gain when negotiators establish a

complementary relation of dominance

50

slide-51
SLIDE 51

STUDY: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

PROCEDURE Stage 1

Tutorial Use the interface to know the utterances

Stage 2

Information about the preferences

Stage 3

Collaborative negotiation with the agent to choose a restaurant

Stage 4

Questionnaire about the negotiation

DATA

  • User preferences
  • Complete dialogue
  • User_Pow computed by the ToM at each negotiation turn

51

slide-52
SLIDE 52

RESULTS: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H1 : Perception of self Dominance Vs other

1 2 3 4 5

SelfVsOther Demand Concession Lead

Perception of dominance Complementary

Other Self * * *

52

slide-53
SLIDE 53

RESULTS : COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

1 2 3 4 5

SelfVsOther Demand Concession Lead

Perception of dominance Similar

Other Self

53

H1 : Perception of self Dominance Vs other

slide-54
SLIDE 54

RESULTS : COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

1 2 3 4 5

SelfVsOther Demand Concession Lead

Perception of dominance Neutral

Other Self * * *

54

H1 : Perception of self Dominance Vs other

slide-55
SLIDE 55

RESULTS : COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H2 : Common gain obtained during negotiation

55

1 2 3 4 5 Comp. Similar Neutral

Perception of satisfaction

1 2 3 4 5 Comp. Similar Neutral

Perception of fairness

slide-56
SLIDE 56

RESULTS: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H2: Common gain obtained during negotiation

0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8

  • Mean. Common gain

Compl. Similar Neutral * * *

56

slide-57
SLIDE 57

RESULTS : COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H3 : Convergence of negotiation [number of negotiation turns]

5 10 15 20 25 30

Mean Nb Utterances

Comp. Similar Neutral * *

57

slide-58
SLIDE 58

RESULTS : COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H3 : Convergence of negotiation [ease of collaboration]

1 2 3 4 5

Ease of collaboration

Comp. Similar Neutral * *

58

slide-59
SLIDE 59

RESULTS: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H4 : Comfort felt during negotiation

1 2 3 4 5

Score of comfort

Comp. Similar Neutral * *

59

slide-60
SLIDE 60

RESULTS: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

H5 : Appreciation of the negotiator’s agent

1 2 3 4 5

Appreciation scores

Comp. Similar Neutral * *

60

slide-61
SLIDE 61

RESULTS: COMPLEMENTARITY VS SIMILARITY IN NEGOTIATION

61

Behaviors of dominance

  • Demand,

concession, self centeredness

  • Lead of negotiation

1 5 3 2 4

Common gain

  • No difference in

perception

  • Significative higher

common gain with Complementary agent

Rapidity of convergence

  • Efficient exchange
  • f information with

the complementary agent

  • Easer collaboration

Comfort during negotiation

Participants felt significatively more comfortable while negotiating with complementary agent

Appreciation of the agent

Participants appreciate significatively more the negotiation with complementary agent

slide-62
SLIDE 62

CONLUSION: A COMPUTATIONAL MODEL OF COLLABORATIVE NEGOTIATION

62

slide-63
SLIDE 63

Validation Agent/agent Human/agent

Communication UtteranceSelf Utteranceother Complementary Relation

CONCLUSION: MODEL OF NEGOTIATION BASED ON DOMINANCE

Mental model

Pow Preferences

1

Decisional model Mental model Pow Preferences Decisional model

Model of the other

Mental model Pow Preferences Decisional model

2 3

Pow

63

Validation Human/agent

slide-64
SLIDE 64

PERSPECTIVE: WHAT’S NEXT ?

64

Evaluation on repeated interactions Integrate individual traits

  • 1. Create specific topics of negotiation
  • 2. Study

personality traits and social hierarchy

Verbal expressivity

  • 1. Linguistic style
  • 2. Natural language in interaction

Argumentation in negotiation

Influence of interpersonal dominance on argumentation's strategies Evolution of the strategies

  • ver interactions

Non verbal behaviors

  • Integrate non verbal behaviors to improve expressivity
  • Study the relation of non verbal behaviors and perception of

negotiation strategies Social conversational agent

slide-65
SLIDE 65

65

Lydia.ouldouali@gmail.com