Situational and Psychological FactorsPredicting Deception and its Detection:
Implications for Non-Cognitive Assessment
Jeff Hancock
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
motivations
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
motivations
detection
any intentional control of information in a message to create a false belief in the receiver of the message
Deception Defined
1. Deception production
a successful or unsuccessful deliberate attempt, without forewarning, to create in another a belief which the communicator considers to be untrue
--Burgoon
--Vrij
1. Deception production
How frequently does lying occur?
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
1.75 lies identified in a 10 minute exchangeRange from 0 lies to 14 liesSelf-presentation goal (‘likeable’) increases deception
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
• type message• rate deceptiveness of message• message and rating is sent to our corpus
Lie-M
• type message• rate deceptiveness of message• message and rating is sent to our corpus
Lie-M
6% of all messageswere deceptive
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
How do different media affect lying and honesty?
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
Three ways to catch a liar
nonverbalphysiologicalverbal
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
Three ways to catch a liar
nonverbalphysiologicalverbal
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
Three ways to catch a liar
nonverbalphysiologicalverbal
DePaulo et al (2003) meta-analysis
• more tense• higher vocal pitch• fidgeting
1. basic facts, examples, principles
How do different media affect lying and honesty?
“Electronic mail is a godsend. With e-mail we needn’t worry about so much as a quiver in our voice or a tremor in our pinkie when telling a lie. Email is a first rate deception-enabler.”
~Keyes (2004) The Post-Truth Era
Three ways to catch a liar
nonverbalphysiologicalverbal
DePaulo et al (2003) meta-analysis
• more tense• higher vocal pitch• fidgeting
eye gaze: unreliable
1. basic facts, examples, principles
0
5
10
15
20
25
30
35
40
45
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Frequencyof Lies perInteraction
How do different media affect lying and honesty?
0
5
10
15
20
25
30
35
40
45
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Nonverbal prediction
Frequencyof Lies perInteraction
0
5
10
15
20
25
30
35
40
45
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Social Distance Theory
< < <
(DePaulo et al, 1996)
Social Distance
Frequencyof Lies perInteraction
Nonverbal prediction
0
5
10
15
20
25
30
35
40
45
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Social Distance Theory(DePaulo et al, 1996)
Frequencyof Lies perInteraction Media Richness Theory
(Daft & Lengel, 1984; 1986)
Nonverbal prediction
0
5
10
15
20
25
30
35
40
45
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Social Distance Theory
Media Richness Theory
Richness > > >
(Daft & Lengel, 1984; 1986)
Frequencyof Lies perInteraction
Nonverbal prediction
0
5
10
15
20
25
30
35
40
45
Line 1
SocialDistanceTheory
HIGH
LOW
FtF Phone EmailInstant
Message
Social Distance Theory
Media Richness Theory
Richness > > >
(Daft & Lengel, 1984; 1986)
Frequencyof Lies perInteraction
Nonverbal prediction
0
5
10
15
20
25
30
35
40
45
Line 1
SocialDistanceTheory
HIGH
LOW
Frequencyof Lies perInteraction
FtF Phone EmailInstant
Message
Social Distance Theory
Social Distance
Media Richness Theory
Richness
FtF Phone IM Email
Media Features
Synchronous X X n
Recordless X X n
Distributed X X X
Lying predictions
Feature-based 2 1 2 3
Media Richness 1 2 3 4
Social Distance 4 3 2 1
Feature Based Approach
PDA-based journal
0
10
20
30
40
50
60
70
Line 1
SocialDistanceTheory
Social Distance Theory
Media Richness Theory
% of Lies per Interaction
Nonverbal prediction
0
10
20
30
40
50
60
70
Line 1
SocialDistanceTheory
Line 3
Social Distance Theory
Media Richness Theory
% of Lies per Interaction
27%
37%
21%
14%Data
FtF Phone EmailInstant
Message
Nonverbal prediction
10
15
20
25
30
35
40
Line 1
Line 2
Line 3
DistributedSimultaneityRecordless
***
** n* * n
% of Lies per Interaction
FtF Phone EmailInstantMessage
27%
37%
21%
14%
Features Model
10
15
20
25
30
35
40
Line 1
Line 2
Line 3
DistributedSimultaneityRecordless
***
** n* * n
% of Lies per Interaction
FtF Phone EmailInstantMessage
27%
37%
21%
14%
Features Model
• more symptoms & undesirable behaviors reported (Griest, Klein & VanCura, 1973)•more sexual partners and symptoms reported (Robinson & West, 1992)• more honest, candid answers in pre-clinical psychiatric interviews (Ferriter, 1993)• 20% of telephone callers vs. 50% of email contacts report suicidal feelings (The Scotsman, 1999)
when interviewed by computer compared to face-to-face:
High levels of self-disclosure and honesty in text-based contexts
1. Deception production
Visual Anonymity
Private Self-Awareness
Public Self-Awareness
Self-Disclosure
self-disclosure and honesty in mediated contexts
Joinson (2001)
1. Deception production
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
Why do people lie?
- Situational factors- Self-presentation goals
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
Why do people lie?
- Situational factors- Self-presentation goals
NOTMONOTLITHIC
1. Deception production
How frequently does lying occur?
• retrospective identification• message-by-message identification• diary studies• ground truth based
Why do people lie?
- Situational factors- Self-presentation goals
NOTMONOTLITHIC
GOALTENSIONS
Female Male
Female Male
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
motivations
- self-presentation goals fundamental- self-presentation goals are tension-based- self-presentation goals can be primed
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
motivations
detection
2. Detecting deception
acoustic profiles•Judee Burgoon’s group• pitch profile changes• large effects for energy and f0 features
2. Detecting deception
New, computer-assisted methods
acoustic profiles•Judee Burgoon’s group• pitch profile changes• large effects for energy and f0 features
facial features• micro-facial expressions (FACS), Mark Frank
2. Detecting deception
New, computer-assisted methods
acoustic profiles•Judee Burgoon’s group• pitch profile changes• large effects for energy and f0 features
facial features• micro-facial expressions (FACS), Mark Frank
linguistic footprints – text-based• fewer 1st person, more 3rd person references• fewer exclusive words• more negative emotion terms• changes in detail level
2. Detecting deception
New, computer-assisted methods
Sender Receiver
“to not tell the truth,the whole truth, and nothing but the truth”on two topics.
Discuss 4 topics
“Maintain the conversation”
Sender Receiver
“to not tell the truth,the whole truth, and nothing but the truth”on two topics.
Discuss 4 topics
“Maintain the conversation”
• transcripts were analyzed with Pennebaker’s Linguistic Inquiry and Word Count (LIWC) program
• LIWC analyzes transcripts on a word-by-word basis and compares words against a dictionary of words divided into 74 psychologically relevant linguistic dimensions
100
110
120
130
140
150
160
170
Sender Receiver
Deception
Truth
Word Count
100
110
120
130
140
150
160
170
Sender Receiver
Deception
Truth
Word Count
28% increase
0
1
2
3
4
5
6
7
8
9
Sender Receiver
deceptiontruth
%words
1st person singular
0
1
2
3
4
5
6
7
8
9
Sender Receiver
deceptiontruth
%words
1st person singular
0
1
2
3
4
5
6
7
8
9
Sender Receiver Sender Receiver
%words
deceptiontruth
2nd person1st person singular
0
1
2
3
4
5
6
7
8
9
Sender Receiver Sender Receiver Sender Receiver
2nd person 3rd person
%words
deceptiontruth
1st person singular
0
1
2
3
4
5
6
7
8
9
Sender Receiver Sender Receiver Sender Receiver
1st person singular 2nd person 3rd person
%words
deceptiontruth
Fewer 1st person singularMore 3rd person
Psychological effect Language process NLP approach/tool
Distancing the speaker from the lie
Non-immediate language - reduced 1st person singular- increased use of passive voice- reduced transitivity- semantic roles (agent v. patient)
Syntactic parser and semantic role identifier
Increased levels of negative affect
Changes in affect terms- increased negative affect valence- attitude type- contextual disambiguation
Sentence- and phrase-level sentiment analysis
Attempt to convey a convincing story
Changes in detail level- noun phrase complexity- dependent/relative clausesChanges in evidentiality - subjective vs. factual presentation- changes in reporting verbs (e.g., saw, hear)
Syntactic parserSentence-level subjective/objective classifier; Reporting verb analyzer
Increased cognitive load Reduced coherenceReduced use of exclusive terms (e.g., never)
CohmetrixLIWC
Collaborative processes Linguistic style matchingQuestion – answer patternsSequential discourse analysis
Auto-correlationSequence prediction
Keila & Skillicorn (2005)
Keila & Skillicorn (2005)
More deceptive
• 105 subjects generating two email texts each• Each completed the Eysenck Personality Questionnaire:
– Extraversion: outgoing - shy– Neuroticism: worrying - relaxed– Psychoticism: toughminded - sympathetic– Lie Scale - measures social desirabity
• Each then composed two emails:– “To a good friend whom they hadn’t seen for quite some time”– One concerned past activities over the previous week– The other concerned planned activities over the next week.
• Each message took around 10 minutes to compose and submit by HTML form.
• The resulting 210 texts contain 65,000 words.
• Texts split by level of Social Desirability– measured by EPQ-R Lie Scale– Split scores by greater +/- 1 Std Dev of mean
• Resulted in two groups– High SDR Authors (N=21)– Low SDR Authors (N=22)
• Corpus comparison of these two groups using Wmatrix software (Rayson 2003, 2005; cf. Oberlander & Gill, 2006)– Identified features significantly over-used or under-used
by each group (using log-likelihood)– All features reported p<0.001
• Hi SDR scorers over-used:– You– ‘Personal names’
• (Richard, Kathy, London)
– words related to ‘Business: Selling’• (shopping, buy, sales, bought)
• Low SDR scorers over-used:– ‘Mental object: Means, method’ words
• (way, system, method, tactical, pattern, set-up)
Some questions about faking
1. Can people fake when instructed?2. What is prevalence of faking?3. What is the nature of faking?4. Can faking be prevented or reduced?5. Can faking be detected?6. Can people avoid detection?
Deception Research
production
motivations
detection
Situational and Psychological FactorsPredicting Deception and its Detection:
Implications for Non-Cognitive Assessment
Jeff Hancock
Top Related