seeking consistent stories

50
Lydia L. Chen 01/03/22 Lydia L. Chen <[email protected]> Seeking Consistent Stories Seeking Consistent Stories By Reinterpreting or Discounting By Reinterpreting or Discounting Evidence: Evidence: An Agent-Based Model An Agent-Based Model Decision Consortium May Conference 13-Mar-2009

Upload: keran

Post on 18-Jan-2016

18 views

Category:

Documents


0 download

DESCRIPTION

Seeking Consistent Stories. By Reinterpreting or Discounting Evidence: An Agent-Based Model. Decision Consortium May Conference 13-Mar-2009. Agenda. The Phenomenon Agent-Based Model (ABM) Primer The Model Sample Run Experiments. Real-Life Scenario: Bench Trial. Prosecution ( p ). - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Seeking Consistent Stories

Lydia L. Chen

04/21/23 Lydia L. Chen <[email protected]>

Seeking Consistent StoriesSeeking Consistent StoriesBy Reinterpreting or Discounting Evidence: By Reinterpreting or Discounting Evidence:

An Agent-Based ModelAn Agent-Based Model

Decision Consortium May Conference

13-Mar-2009

Page 2: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 2

AgendaAgenda

1. The Phenomenon

2. Agent-Based Model (ABM) Primer

3. The Model

4. Sample Run

5. Experiments

Page 3: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 3

Real-Life Scenario: Bench TrialReal-Life Scenario: Bench TrialProsecution () Defense Lawyer

“Guilty!

”“Innocen

t!”

Page 4: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 4

Sequential Evidence – What’s NormativeSequential Evidence – What’s Normative

Evidence 1

Evidence 2

Evidence 3

Evidence N

Official Deliberatio

n

Story/ Verdict

Page 5: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 5

Empirical Literature on JDMEmpirical Literature on JDM Form “coherent” stories that support option/verdict

(Pennington & Hastie, 1988)

Confidence in option/verdict increases with “coherence” (Glockner et al., under review)

Decision threshold: “sufficiently strong” (supported by many consistent evidence) or “sufficiently stronger” than other stories (review by Hastie, 1993

Narrative coherence: consistency causality completeness

“Consistency” aspect of “good” stories consistency between evidences in a story consistency of evidence with favored story

Page 6: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 6

Example Case (Pennington & Hastie, 1988)Example Case (Pennington & Hastie, 1988)

Scenario: Defendant Frank Johnson stabbed and killed Alan Caldwell

Evidence := facts or arguments given in support of a story/verdict Facts— “Johnson took knife with him” ; “Johnson pulled out his knife”

Arguments—”Johnson pulled out knife because he wanted revenge” vs. “Johson pulled out knife because he was afraid”

Story := set of evidence supporting a given verdict

Same evidence can be framed to support multiple verdicts/stories!

P D

Desired Verdict “Guilty” “Innocent”

Story “Premeditated murder” “Self-defense”

Interpretation of Evid

“Johnson was angry” “Johnson was afraid”

Page 7: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 7

Sequential Evidence – More DescriptiveSequential Evidence – More Descriptive

Evidence 1

Compare, Deliberate, Interpret

Evidence 2

Evidence 3

Evidence N

Official Deliberatio

n

Story/ Verdict

Premature Story/

Verdict @ Evid n < N

(Brownstein, 2003; Russo et al., 2000)

Page 8: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 8

How People Deal with Incoming EvidenceHow People Deal with Incoming Evidence

People don’t just take evid @ face value, but are selective!

Possible Reactions to New Evid in Light of Old:

“reinforce” each other

“reinterpret” less plausible one (Russo et al., etc.); e.g., misremember the info

“discount” less plausible one (Winston)

actively “seek” more evidence (not modeled here)

Existing evidence A: “Johnson was not carrying a knife.”

New evidence B1: “Johnson is nonviolent.”

Inconsistent new evid B2: “Johnson pulled a knife.”

Reinterpret B2: “Johnson grabbed a knife from Caldwell.” (i.e., explain it was Caldwell’s knife, not Johnson’s)

Discount B2: “Witness must be mistaken.”

Judge asks layers follow-up questions

Page 9: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 9

MotivationsMotivations

Study emergence of consistent stories via reinforcement, reinterpretation, and discounting mechanisms

Process, functions, consequences

Adaptive? aid consistency? speed-accuracy tradeoff? avoid indecisiveness

increases convergence rates? order effects hurt accuracy?

Page 10: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 10

AgendaAgenda

1. The Phenomenon2. Agent-Based Model (ABM) Primer

[1:43]

3. The Model

4. Sample Run

5. Experiments

Page 11: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 11

What is Agent-Based Modeling?What is Agent-Based Modeling? agents + interactions^

start simple; build up^

Key terminology agents system dynamics

agent births and deaths interactions/competitions

parameters

A System of Agents

A

A

A

A A

A

Page 12: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 12

(Contributions ABMs Can Make)(Contributions ABMs Can Make)

Symbiotic relationship:

Behavioral Experiments

ABM

parsimony:

demo emergence of seemingly complex

phenomen from small set of simple rulespredictions:

new observations/predictions

understanding:

study processes in detail

Test

Inputdescription:

informs base assumptions

Page 13: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 13

What’s new and different?

Page 14: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 14

(Other Models (Hastie, 1993))(Other Models (Hastie, 1993))

Page 15: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 15

Algebraic (additive), Bayesian (multiplicative): “single meter” of overall plausibility

ABM allows: revisiting and reconsidering previously-processed

evidence

interaction/competition between individual evidences (not just conglomerate)

(Contrast with Bayesian & Algebraic Models)(Contrast with Bayesian & Algebraic Models)

conglomerate of all previous evidence

new evidence

Page 16: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 16

Contrast with Story Model & ECHOContrast with Story Model & ECHO Explanatory Coherence Model := Thagard’s Theory

of Explanatory Coherence (TEC) + Story Model

Only implemented discounting, not reinterpretation

Local evidence-agent-level consistency, as opposed to global story-level consistency

Unlike previous ABMs, model agents within individual as system

Page 17: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 17

AgendaAgenda

1. The Phenomenon

2. Agent-Based Model (ABM) Primer3. The Model [1:46]

4. Sample Run

5. Experiments

Page 18: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 18

GoalGoal

Model consistency-seeking process in story formation

Present evidence-agents to judge-system

Judge-system compares evidence-agents => keep, reinterpret, or discount evidence (agents “interact” & “compete”)

Until sufficiently strong/stronger story emerges

Page 19: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 19

Evidence-agents, composed of “features”

Operationalize consistency /b/ agents: “similarity” in abstract features (Axelrod’s culture ABM,

1997) “inverse Hamming distance” := % feature matches

AgentsAgents

G

x

x

89%

Evid 1

N

x

y

34%

Evid 2

Verdict (“G”,”N”,…)

Abstract features (binary)

Plausibility index (0%-

100%)

N

x

x

14%

Evid 1

Page 20: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 20

(Parameters of “Base Case”)(Parameters of “Base Case”)Parameter Variabl

eBase

# agents to present

- Initial A0 3

- Additional A 20

# possible verdicts V 2

# features/evidence (excluding verdict) F 2

% feature matches for:

Sufficiently consistent C 100%

Close enough for reinterpretation N < C 50%

# interactions /b/ evidence-agent births I (A0+A -1)

Change rate of agent plausibility (increase upon winning, decrease upon losing)

D 0.1

Rule(s) and threshold(s) for winning story-- “sufficiently strong” and/or “sufficiently stronger”

S, Sd Both — 0.5, 0.2

Page 21: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 21

(Java GUI)(Java GUI)

Page 22: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 22

System & Agent BirthsSystem & Agent Births

Judge-system represents judge’s mind

Evidence presentation = “agent birth”:

Initialization of N0 Agents: Set up randomly-generated agents, OR…

uniform distributions—even for plausibility index due to prior beliefs (Kunda, 1990) or knowledge (Klein, 1993)

User-specified

Page 23: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 23

(Topology)(Topology)

results\sysout_081118_0455.log

Printing 8 agents:

01 02 03 04 05 06 07 08

N G I G N I N N

y y y y y y y y

y y y y y y y y

100% 100% 100% 100% 100% 100% 100% 80%

Dead/Rejected Evidence--Printing 9 agents:

01 02 03 04 05 06 07 08 09

I G G I G I N N N

y y x x y x x x y

y x x x y x x y y

00% 00% 00% 00% 00% 00% 00% 00% 00%

To keep track of evidence-agents and their order of presentation:

lists of agents in order of birth/presentation to the system.

latest-born agent always appears at the end of a list.

no geographical “topology”

Page 24: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 24

StoriesStories:= {evidence-agents supporting a verdict}

Consistency = inverse Hamming distance amongst evidence (:= evidencePairs # feature matches / # evidencePairs / F )

Plausibility Index (“Strength”) = average of plausibility indices

G

x

x

89%

Evid 1N

x

y

51%

Evid 2

N

x

x

14%

Evid 1

Story promoting

“Guilty” verdict

Story promoting “Innocent”

verdict

strength = 89%

strength = 99% / 3

= 33%

N

y

y

34%

Evid 3

Page 25: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 25

Births & InteractionsBirths & Interactions1. If time period t = k * I, where k is some constant, birth new

agent.

2. Random selection of agent to compare with newest-born.

3. Interaction. Compare agents and compute consistency. Depending upon consistency [next slide]:

“reward consistency” “increase consistency” or “punish consistency”

Agents with plausibility = 0% => death & removal from system

4. Gather stories in system. Check strengths.

“Winning story found” if 1 ! story with strength >= S and/or |strength-strength | >= Sd for all competing stories; stop run early.

Page 26: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 26

Operationalizing Consistency--ExamplesOperationalizing Consistency--Examples

Possible Reactions to New Evid :

Completely consistent (e.g., 100% features match) => => both are “winners” => “reinforce” both; “reward consistency”

Inconsistent…

but salvageable (50% features match) => “reinterpret” less plausible “loser” (by plausibility); “increase inconsistency”

not salvageable (0% features match) => “discount” “loser”; “punish inconsistency”

=> plaus(Evid1) > plaus(Evid2) => Evid2 “loser”

x

y

51%

Evid 1

x

y

34%

Evid 2

x

y

61%

Evid 1

x

y

44%

Evid 2

x

y

51%

Evid 1

y

y

34%

Evid 2

x

y

51%

Evid 1

x

y

34%

Evid 2

x

x

51%

Evid 1

y

y

34%

Evid 2

x

x

51%

Evid 1

x

y

24%

Evid 2

Page 27: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 27

AgendaAgenda

1. The Phenomenon

2. Agent-Based Model (ABM) Primer

3. The Model4. Sample Run [1:54]

5. Experiments

Page 28: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 28

Sample Output^Sample Output^ Live Evidence--Printing 8 agents:

01 02 03 04 05 06 07 08

N G I G N I N N

y y y y y y y y

y y y y y y y y

100% 100% 100% 100% 100% 100% 100% 80%

Dead/Rejected Evidence--Printing 9 agents:

01 02 03 04 05 06 07 08 09

I G G I G I N N N

y y x x y x x x y

y x x x y x x y y

00% 00% 00% 00% 00% 00% 00% 00% 00%

3 stories found:

-Verdict N supported by 4 evidence, with 1.00 consistency => 0.48 strength

-Verdict G supported by 2 evidence, with 1.00 consistency => 0.25 strength

-Verdict I supported by 2 evidence, with 1.00 consistency => 0.25 strength

*** Found winning story! Verdict N supported by 4 evidence, with 1.00 consistency => 0.48 strength

Page 29: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 29

Judge-System can get Stuck…Judge-System can get Stuck…results\sysout_STUCK.log

Live Evidence--Printing 17 agents:

01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17

N G G I G I I I I N I G G N N N G

x x x x x x x x x x x x x x x x x

y y y x y y y y y y y y y y y y y

100% 100% 100% 90% 100% 100% 100% 100% 100% 100% 84% 100% 79% 100% 100% 98% 100%

Dead/Rejected Evidence--Printing 10 agents:

01 02 03 04 05 06 07 08 09 10

I I I I N I I G N G

y x y y y y y y y y

y x x x x x x y x x

00% 00% 00% 00% 00% 00% 00% 00% 00% 00%

3 stories found:

-Verdict N supported by 5 evidence, with 1.00 consistency => 0.29 strength

-Verdict G supported by 6 evidence, with 1.00 consistency => 0.34 strength

-Verdict I supported by 6 evidence, with 1.00 consistency => 0.34 strength

No winning story found.

Page 30: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 30

Outputs of a Run (DVs)^Outputs of a Run (DVs)^

GGG GGGGGGGG GGGGGGG GG

5 10 15 20

04

8

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

04

8

Supporting Evidence (#)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.6

N NN

N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.6

Plausibility (%)

Time (presentation of agents)

GGGGG

GGGG

G GGGGGGG GG

5 10 15 20

0.0

0.6

N NN N N

N N N N N N N NN N

N NN N N

5 10 15 20

0.0

0.6

Consistency (%)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.4

0.8

N N NN N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.4

0.8

Strength (%)

Time (presentation of agents)

Page 31: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 31

AgendaAgenda

1. The Phenomenon

2. Agent-Based Model (ABM) Primer

3. The Model

4. Sample Run5. Experiments [1:56]

Page 32: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 32

Controlled Experiments in SimulationControlled Experiments in Simulation “control group”

vary IV => various “treatment groups”

Compare and contrast control group and treatment groups

“base case”

vary parameter => various other cases

Compare and contrast base case and other cases

Adv: Don’t have to worry about assumptions b/c study inter-group differences rather than absolute

outputs

Page 33: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 33

5 Experiments5 Experiments

Experiment 1: Emergence of consistency

Decision Speed Experiment 2: Speedup Experiment 3: Accuracy tradeoff

Decision Accuracy—Order Effects Experiment 4: Emergence of order effects Experiment 5: Extending deliberation

Page 34: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 34

Obtaining Consistent Stories – Q1Obtaining Consistent Stories – Q1

Q1: Evidence-level consistency sufficient? Which of the 3 mechanisms?

Implementation: No stopping rules

DV: Consistency of stories

Page 35: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 35

Obtaining Consistent Stories – Q1 ResultsObtaining Consistent Stories – Q1 Results

Reinforce Reinterpret Discount Median consistency

0.9

0.8

0.7

0.6

0.6

0.6

0.5

Reinterpret > Discount > Reinforcement

Page 36: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 36

Speed-Accuracy Tradeoff – Q2Speed-Accuracy Tradeoff – Q2

Q2: Reinterpretation & Discounting increase speed?

Prediction: Reinterpretation & Discounting allow faster convergence

DV: Time to Converge, Max nEvid (Max Consistency)

Implementation: Both rules; all cases have reinforce

Page 37: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 37

Speed-Accuracy Tradeoff – Q2 ResultsSpeed-Accuracy Tradeoff – Q2 Results

Results of 10 Runs—Time to First Convergence, Maximum Number of Evidence,

Maximum Story Consistency

Run1 2 3 4 5 6 7 8 9 10

W/o Reinterp, W/o Discount results/sysout_081215_1938.log

NoC160.5

NoC130.5

NoC120.5

NoC120.5

NoC120.5

NoC140.5

NoC150.5

NoC120.5

NoC120.5

NoC130.5

W/o Reinterp, W/ Discountresults/sysout_081215_1937.log

NoC9

0.8

64

1.0

18110.6

NoC9

0.6

NoC100.5

NoC140.5

NoC110.6

NoC110.7

178

1.0

NoC9

0.5

W/ Reinterp, W/o Discountresults/sysout_081215_1935.log

NoC131.0

NoC120.8

NoC140.6

54

1.0

86

1.0

NoC120.6

NoC130.8

NoC120.9

21140.9

54

1.0

W/ Reinterp, W/ Discountresults/sysout_081215_1946.log

NoC100.8

NoC131.0

43

1.0

NoC8

0.8

126

1.0

43

1.0

NoC8

0.8

86

1.0

NoC9

0.8

19101.0

Page 38: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 38

Speed-Accuracy Tradeoff – Q2 ResultsSpeed-Accuracy Tradeoff – Q2 Results

Medians of 10 Runs—Time to First Convergence, Maximum Number of Evidence,

Maximum Story Consistency

Reinterpret > Discount > Reinforcement only

W/o Discounting W/ Discounting

W/o Re-interpretation Converged 0% of runs nEvid = 12.5

(0.5)

Converged 30% of runs, time = 17,

8(1.0)

W/ Re-interpretation Converged 40% of runs, time = 6.5, nEvid = 5

(1.0)

Converged 50% of runs, time = 8

6 (1.0)

Page 39: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 39

Speed-Accuracy Tradeoff – Q3Speed-Accuracy Tradeoff – Q3 Q3: What would happen if allow process to continue even after having

found winner? Any point to "holding off judgment" until all evidence presented?

DV: Which story wins? (Strength)

Prediction: Leader will only be strengthened; competing stories never get a foothold.

Implementation: Allow system to continue running even if found winner

Run 1 2 3 4 5 6 7 8 9 10

Winner once conditions met (G/N)

Winner after additional evidence presented (G/N)

Changed winner? (Y/N)

Page 40: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 40

Speed-Accuracy Tradeoff – Q3 Results^Speed-Accuracy Tradeoff – Q3 Results^

Over 20 runs, # runs same winner:# runs different winner = 15:1

=> Good heuristic to stop deliberation, for less time & effort

GGG GGGGGGGG GGGGGGG GG

5 10 15 20

04

8

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

04

8

Supporting Evidence (#)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.6

N NN

N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.6

Plausibility (%)

Time (presentation of agents)

GGGGG

GGGG

G GGGGGGG GG

5 10 15 20

0.0

0.6

N NN N N

N N N N N N N NN N

N NN N N

5 10 15 20

0.0

0.6

Consistency (%)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.4

0.8

N N NN N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.4

0.8

Strength (%)

Time (presentation of agents)

Figure 2. Example run where winner switches

Page 41: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 41

Order Effects – Q4Order Effects – Q4

Heuristic may be ok only if randomized evidence…what if biased evidence?

Q4: Is there an Order Effect?

Took {20 randomly-generated evidence} and then “doctored” it; % D win = “accuracy”

IV: Presentation order--P goes first, followed by D vs. interwoven evidence

Prediction: earlier, weaker side (e.g., P) beats out later, stronger side (e.g., D); "Accuracy" of D…P… > PDPDPD… > P…D…

DV: Time to Converge, (Projected) Winner

Page 42: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 42

Order Effects – Q4 ResultsOrder Effects – Q4 Results

GGG GGGGGGGG GGGGGGG GG

5 10 15 20

04

8

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

04

8

Supporting Evidence (#)

Time (presentation of agents)

GGGGGGG G

GGGGGG GGGGGG

5 10 15 20

0.0

0.6

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.6

Plausibility (%)

Time (presentation of agents)

G

GG GGGGGG

GG GGGGGGG GG

5 10 15 20

0.0

0.6 N N N

N N N NN N N N N N N N N N N

N N

5 10 15 20

0.0

0.6

Consistency (%)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.4

0.8

N NN N N N N N N N N N N N

N N N N N N

5 10 15 20

0.0

0.4

0.8

Strength (%)

Time (presentation of agents)

Figure 3. Sample random presentation order run.

Page 43: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 43

GGG GGGGGGGG GGGGGGG GG

5 10 15 20

04

8

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

04

8

Supporting Evidence (#)

Time (presentation of agents)

GGGGGGG GGGGGG

G GGGGGG

5 10 15 20

0.0

0.6

N N N N N N N N N N N NN N N N N N N N

5 10 15 20

0.0

0.6

Plausibility (%)

Time (presentation of agents)

GGG GGGGGGGG GGGGG

GG GG

5 10 15 20

0.0

0.6

N N

N N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.6

Consistency (%)

Time (presentation of agents)

GGGGGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.4

0.8

NN

N N N N N N N N N NN N N N N N N N

5 10 15 20

0.0

0.4

0.8

Strength (%)

Time (presentation of agents)

Order Effects – Q4 ResultsOrder Effects – Q4 Results

Figure 4. Sample D…P… biased presentation order run.

Page 44: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 44

Order Effects – Q4 ResultsOrder Effects – Q4 Results

Figure 5. Sample P…D… biased presentation order run.

GGG GGGGGGGG GGGGGGG GG

5 10 15 20

06

12

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

06

12

Supporting Evidence (#)

Time (presentation of agents)

GGGG

GGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.6

NN N N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.6

Plausibility (%)

Time (presentation of agents)

GGGG

GGGGGGG GGGGGGG GG

5 10 15 20

0.0

0.6 N

N N N N N N

N NN N N N N N N N

N N N

5 10 15 20

0.0

0.6

Consistency (%)

Time (presentation of agents)

GGG

GGGG GGGGGGG GGGGGG

5 10 15 20

0.0

0.4

0.8

N N N N N N N N N N N N N N N N N N N N

5 10 15 20

0.0

0.4

0.8

Strength (%)

Time (presentation of agents)

Page 45: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 45

Order Effects – Q4 ResultsOrder Effects – Q4 Results

Table 4. Results of 10 Runs Varying Presentation Order of Evidence

Presentation Order

Median Time to Converge% Runs that Defense Won

% Runs Resulting in Switch in Winner

PDPDPD… 7D won 90%, NoC 10%

0%

D… P… 5D won 90%, NoC 10%

0%

P... D… 8P won 80%, NoC 20%

30%

• Strong primacy effect

• Exper3 conclusion no longer holds; longer deliberation DOES help!

Page 46: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 46

(Increasing Deliberation – Q5)(Increasing Deliberation – Q5)

Q5: Can deliberating more often between births reduce order effects (i.e., increase "accuracy")?

Implementation: Use P…D… model from Exper4

IV: Varied I (I = 0 => wait till end to deliberate)

DV: % runs that P wins

Page 47: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 47

(Increasing Deliberation – Q5 Results)(Increasing Deliberation – Q5 Results)

Run 1 2 3 4 5 6 7 8 9 10 % P Wins

I = 0.8 * (# agents – 1)

60%

I = 1.0 * 80%

I = 1.2 * 100%

Too much lag time during trials can be detrimental!

Page 48: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 48

Summary of Key FindingsSummary of Key Findings Why reinterpret and discount evidence?

1. Maximizes consistency (Experiment 1)2. Hastens convergence on decision (Experiment 2)

Reinterpret > Discount > Reinforcement

Speed-accuracy tradeoff? Yes… Accuracy ok if evidence balanced (Experiment 3) Not ok/primacy effect if biased (Experiment 4) => Important to interweave evidence, like in real trials!

Can reduce primacy effect by decreasing premature deliberation (Experiment 5)

All achieved by modeling consistency @ evidence level, not story level! more parsimonious & realistic(?)

Page 49: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 49

Other Simplifying AssumptionsOther Simplifying Assumptions Features

Any combination of abstract features can be framed (e.g., by the lawyers) to support a verdict. All evidence are automatically categorizable into a verdict.

Interactions Interactions only take place between the newest agent and another agent.

Computing Consistency Verdict feature neither considered nor included in interactions

Comparing Stories judge forms 1 story / verdict

If only 1 story is found at time t, use the "sufficiently strong" rule as opposed to the "sufficiently stronger" rule. (i.e., no “holding out” by judge)

Which layer drives consistency-seeking? consistency at the individual agent interaction level, AND/OR consistency at the story level

Page 50: Seeking Consistent Stories

04/21/23 Lydia L. Chen <[email protected]>“Seeking Consistency” - 50

(Future Expansions)(Future Expansions) Q6: What happens when introduce bias toward

certain verdicts? Prediction: Verdict-driven process takes less time to

converge DV: Time to Converge (Consistency of Stories) Implementation: Add favoredVerdict

Q7: In general, what conditions lead to indecision?