Date post: | 13-Dec-2015 |
Category: |
Documents |
Upload: | anthony-eaton |
View: | 225 times |
Download: | 1 times |
Hidden Patterns of Nonverbal Behavior Associated with
Truth and DeceptionSPSP Data Blitz
Dr. Judee Burgoon Jeff Proudfoot David Wilson
Ryan Schuetzler
SPSP 2
Introduction• Communication tends to be highly
patterned– Including subtle, perhaps
imperceptible patterns– Communication patterns are complex
• Deception– Much research focused on very brief
segments– Ignores patterns among behaviors and
dynamic changes
• Does deception affect patterning and temporal changes?1/17/2013
SPSP 3
Behavior Pattern Analysis• Bottom-up search of time-coded event data• Identifies behaviors occurring sequentially within a critical,
statistically significant time interval• Patterns may be combined to form multi-level, nested patterns
(Magnusson 2005, 2006)More information available at www.noldus.com
1/17/2013
SPSP 4
Behavior Pattern Analysis
1/17/2013
SPSP 5
Behavior Pattern Analysis
1/17/2013
SPSP 6
Behavior Pattern Analysis
t-pattern
1/17/2013
SPSP 7
Sample (Nested) Pattern17 elements, 6 levels, 4 occurrences
1/17/2013
SPSP 8
Experiment 1: Mock Theft Method
• Participants randomly assigned to “steal” a wallet from a classroom
• Both guilty and innocent participants interviewed– Innocent participants told the truth– Guilty participants lied
• Video-recorded interviews included baseline and theft-relevant questions
• Nonverbal behaviors manually coded with timestamps, submitted to Theme Analysis– e.g., illustrative gestures, adaptor behaviors, etc.
1/17/2013
SPSP 9
Mock Theft Results
• Truth tellers averaged longer patterns (M = 6.55, SD = 1.95) than did deceivers (M = 5.17, SD = 2.16)
• Deceivers repeated patterns (M = 9.48, SD = 2.95) more than truth tellers (M = 7.93, SD = 2.48), i.e., more redundancy
• During baseline questions, truth tellers had more patterns (M = 247, SD = 336) than deceivers (M = 98, SD = 395)
• During theft questions, truth tellers introduced far more new patterns (M = 23.6, SD = 32.2) than deceivers (M = 3.6, SD = 9.95)
1/17/2013
SPSP 10
Experiment 2: Cheating Method
• Participants played a trivia game with a partner (confederate)– Randomly induced to cheat (or not)– High-stakes academic consequences if
caught cheating– Some refused to cheat and some cheaters
confessed• All participants interviewed• Video-recorded interviews included
baseline, suspicion and direct accusation questions
• Nonverbal behaviors manually coded with timestamps, submitted to Theme Analysis
1/17/2013
SPSP 11
Cheating Results• Cheaters had fewer total patterns and fewer
unique patterns (due to inactivity, redundancy?)
Statistic
Condition
Honest (n = 25)
Cheaters (n = 6)
Confessors (n = 4)
Induced, but didn’t cheat
(n = 18)
Mean Unique Patterns 293.9 195.3 280.3 339.8
StDev Unique Patterns 337.5 98.7 134.9 290.8
Mean Total Patterns 1478.4 1055.3 1574.5 1678.1
StDev Total Patterns 1303.8 371.3 651.1 1199.8
1/17/2013
SPSP12
Comparison
Cheaters vs. Honest Liars vs. ConfessorsCheaters vs. Induced
but didn’t cheatdf = 29 df = 8 df = 22
t p-value t p-value t p-value
Num Unique Patterns 1.105 0.281 1.042 0.328 2.028 0.052
Total Num Patterns 1.235 0.230 1.341 0.217 2.185 0.037
Cheating Results--Group Comparisons
1/17/2013
SPSP 13
Experiment 3: StrikeComMethod
• 3-person groups (N = 14 triads) in a mock military command scenario (search & destroy enemy missile sites)
• Completed 5 search turns + 1 strike turn• 1 deceptive, 1 suspicious, and 1 naïve player• Nonverbal behaviors manually coded with
timestamps, submitted to Theme Analysis
1/17/2013
SPSP 14
StrikeCom Results
• Number of unique patterns ranged from 48 to over 1,600
• Deceivers exhibited strategic, manipulative patterning behavior
• Suspicious players showed investigatory probing patterning behavior
1/17/2013
SPSP 15
StrikeCom Results
VariableAir Intel Space
M SD M SD M SD% Illustrator gestures 20.6 (15.6) 32.6 (12.0) 46.9 (23.1)% Adaptor gestures 27.4 (15.4) 43.9 (16.1) 28.7 (12.7)% Lip adaptors 36.0 (20.6) 29.0 (12.0) 35.0 (17.5)% Speaker head movements 34.2 (21.4) 30.4 (16.0) 35.4 (19.5)% Listener head movements 42.4 (29.7) 26.6 (17.4) 31.0 (18.5)% of speaking activities 29.8 (16.9) 31.2 (11.4) 39.1 (19.4)% Total patterns 46.2 (21.2) 41.3 (23.2) 54.1 (19.3)% Total patterns solo 21.3 (13.5) 25.8 (18.6) 27.9 (17.5)% Total patterns initiating 34.5 (17.0) 26.7 (13.4) 38.8 (19.0)
% Total patterns with switches 25.0 (16.8) 25.9 (22.8) 28.2 (16.4)Pattern length (complexity) 1.42 (0.7) 1.31 (0.8) 1.67 (0.6)
Means and Standard Deviations of Behaviors and General Pattern Statistics (N = 14)
1/17/2013
SPSP 16
Analysis and Results
Percent of session patterns
Speaking behavior Interactivity Specific roles
Started by Air
talking
Started by Intel talking
Started by
Space talking
With at least one
switch
With one
actor
With two
actors
With three actors
No Air
No Intel
No Space
With Air
With Intel
With Space
With Air .55 .34 -.63 .03 -.19 .04 .45 -.56 .37 .58 — .01 -.74
With Intel -.08 .26 -.51 .96 -.84 .81 .62 .66 -.59 .59 — — -.11
With Space -.52 -.58 .74 .03 -.09 .15 -.08 .45 .11 -.53 — — —
Select Intercorrelations of Session-Level Patterning Behaviors (N = 14)
1/17/2013
SPSP 17
Discussion
• Deceptive behavior is highly patterned• Pattern analysis reveals many relationships that would
otherwise go unnoticed– Tendency of deceiver to initiate patterns with manipulative
behaviors– Higher frequency of interaction between deceiver and
suspector (excluding the third group member)• Structure of and relatedness among interactive
behaviors only available through pattern analysis
1/17/2013
SPSP18
Questions?
1/17/2013