Can Implicit Association Tests Be Used To Measure Personality John Coaster and Brian P' Siers Facult

1
Can Implicit Association Tests Be Used To Measure
Personality?John Coaster and Brian P.
SiersFaculty Sponsor Dr. Neil
ChristiansenDepartment of Psychology - Central
Michigan University
Background In response to problems with the self
report method of measuring personality traits,
the Implicit Association Test (IAT) has been
adapted to measure personality traits.
Previous research indicates inconsistent
relationships between trait IATs, established
self report measures of personality, and
behavioral outcomes. The present study uses the
Multitrait-Multimethod (MTMM) approach to explore
the validity of IAT measures of personality.
About MTMM To establish validity evidence for a
measure, the MTMM matrix should meet the
following 4 criteriaCriterion 1 Validity
diagonal values should not be zero and be
practically significant. Criterion 2 Validity
diagonal values should be larger than the
adjacent heterotrait-heteromethod
values. Criterion 3 There should be higher
correlations between attempts to measure the same
trait with different methods (monotrait-heterometh
od) than between the different traits measured by
a single method (heterotrait-monomethod).
Criterion 4 Patterns of correlations should
be the same among heteromethod-monomethod values.

Results
Conclusion MTMM analyses provided little
evidence that IAT measures of personality traits
are measuring the same psychological construct
assessed by self report and peer ratings of
personality traits. More research must be done to
determine what the IAT trait measures are
actually measuring.
Criterion 1 This criterion is partially met.
Validity diagonal values involving IATs are not
0. However, only 2 of 6 values are practically
significant (extraversion IAT/self report
extraversion r -.263 stability IAT/peer
stability r .173). Criterion 2 This
criterion is not met. The average validity
diagonal values are r .104 (IAT/self report)
and r .107 (IAT/peer). The corresponding
average heterotrait-heteromethod values for IAT
and self report are r .091 and r .131
(which is larger than the average validity
diagonal value). The average heterotrait-heteromet
hod values for the IAT/peer values are r .077
and r .074.
Criterion 3 This criterion is not met. The
average heterotrait-monomethod correlation for
the IAT is r .223, which is higher than either
of the monotrait-heteromethod values (IAT/self
report r .107 IAT/peer r .104).
Criterion 4 This criterion is not met. The
pattern of correlations is uneven. For example
the largest IAT/self report correlations are the
extraversion/extraversion (r -.263) and
consciensiousness/extraversion (r -.203).
However, those are not the highest correlations
in the IAT-peer block.

• Sample
• Participants 153 Undergraduates
• Measures Administered measures of extraversion,
  conscientiousness, and emotional stability.
• Each trait measured using 3 different methods
  (a) self report (Goldbergs Adjective Scale),
  (b) IAT, and (c) peer ratings of participant
  traits.

2
Are Implicit Association Tests of Personality
More Difficult to Fake?Annalyn C. Jacob and
Brian P. SiersFaculty Sponsor Dr. Neil
ChristiansenDepartment of Psychology - Central
Michigan University

• Results
• Hypothesis Supported. Several results indicate
  self-report trait measures were impacted by
  response distortion more than IAT trait measures.
  
• Mean Shift
• Comparisons were made between test scores in the
  honest and applicant conditions. Mean scores in
  the applicant condition were higher than in the
  honest conditions. The shift (see figure) was
  much greater for self-report measures than for
  IAT measures. This indicates that participants in
  the applicant condition significantly inflated
  scores on the self-report measures, but could
  only slightly do so on the IAT measures.
• Validity Deterioration
• Correlations between peer ratings of participant
  traits and self-report measures decreased in the
  applicant condition indicating response
  distortion reduces the convergent validity of
  self report measures. IAT measures were much more
  stable for 2 of the 3 IAT trait IATs this
  correlation slightly increased in the applicant
  condition (see table).
• The average intertrait correlation among the
  three self reported traits increased in the
  applicant condition (.203 in the honest condition
  to .692 in the applicant condition). However, the
  average intertrait correlation among traits
  measured with the IAT does not increase in the
  applicant condition (.223 in the honest
  condition .220 in the applicant condition).
• Conclusions
• These results suggest IAT measures of
  personality traits are minimally impacted in
  situations that yield significant response
  distortion on self-report measures of
  personality. However, more validity data should
  be gathered before IAT trait measures are used
  for personnel selection.

• Background
• Many organizations (13 in a recent survey) use
  personality tests to select employees for jobs.
• However, the use of self report measures of
  personality traits for personnel selection is
  questionable because applicants often distort (or
  fake) their responses to the tests.
• These response distortions degrade the tests
  validity and impact hiring decisions made based
  on test scores.
• Recently, the Implicit Association Test (IAT)
  has been adapted to measure personality traits.
  The IAT uses response latencies as an index of
  the association strength between self concept and
  trait concepts.
• The present study compares test scores between
  two groups (honest and applicant conditions) to
  determine if IAT trait measures show more
  resistance to the effects of response distortion
  than self-report trait measures.
• Hypothesis
• In the applicant condition, important
  psychometric properties of self-report
  personality measures will be negatively impacted
  more than the psychometric properties of IAT
  trait measures.

Honest vs. Applicant Condition Standardized Mean
Differences on Self Report and IAT Trait Measures
Standardized Mean Differences
Extraversion
Conscientiousness
Emotional Stability

• Method
• Participants 298 Undergraduates
• Procedure 2 conditions
• Honest Condition (n 153) Participants
  completed all measures honestly.
• Applicant Condition (n 145) Participants given
  a job description for a customer service position
  and told the testing session was an employment
  testing simulation for the position. Told if
  hired, would receive a cash prize.
• Measures Self report (NEO, Goldbergs Adjective
  Scale), IAT, and peer ratings of participant
  traits measured extraversion, conscientiousness,
  and emotional stability.

3
Do Implicit Association Test (IAT) Measures of
Personality Traits Predict Job Performance Beyond
Self Report Trait Measures?Maria Jacome and
Brian P. SiersFaculty Sponsor Dr. Neil
ChristiansenDepartment of Psychology - Central
Michigan University

• Results
• (H1) Partially Supported
• Self-report measures of personality predicted
  scores on the job performance scale (see Figure
  1). Self report measures of extraversion and
  conscientiousness were weakly correlated with the
  overall performance item.
• (H2) Partially Supported.
• IAT extraversion and stability measures were
  related the overall performance item. IAT
  conscientiousness was less related to the overall
  job performance item, but still to a greater
  extent than the self report scale.
• (H3) Partially Supported.
• IAT trait measures added a practically
  significant degree of predictive utility beyond
  self report measures in 3 of the 6 tests (see
  Table 1).
• Conclusions
• IAT trait measures consistently add prediction
  of the overall job performance measure beyond
  self report measures of personality traits.
• The observed relationships were inconsistent.
  More research needs to be done to identify which
  criterion measures consistently relate to self
  report predictors, which criterion measures
  relate to IAT predictors, and the reasons for the
  discrepancies.
• More data needs to be collected. It is difficult
  to make meaningful inferences using such a small
  (N 30) sample.

Figure 1 Correlations Between Personality
Measures (Self Report and IAT) and Two Measures
of Job Performance
Background Personality tests are an important
component of many selection processes. The
use of self report personality measures is
questionable because response distortion (e.g.
faking) can impact hiring decisions made from
test scores, and degrade the tests construct and
criterion-related validity. Recently, the
Implicit Association Test (IAT) has been adapted
to measure personality traits. The IAT uses
response latencies as an index of the association
strength between self concept and trait concepts.
This study explored the criterion-related
validity of IAT trait measures by relating scores
on IAT trait measures to job performance.
Hypotheses (H1) Explicit trait measures
(extraversion, conscientiousness, and emotional
stability) will predict job performance. (H2)
Implicit trait measures (extraversion,
conscientiousness, and emotional stability) will
predict job performance. (H3) Implicit trait
measures will predict job performance beyond the
increment of performance predicted by explicit
measures.
Job Performance Scale (10 item)
Overall Performance (1 item)
Extraversion
Extraversion
Stability
Stability
Conscientiousness
Conscientiousness

• Method
• Participants 30 Undergraduates
• Procedure
• Participants completed a variety of measures
  in a lab setting.
• Participants took surveys home to their
  friends, family, etc. Completed surveys and
  returned them.
• Supervisors mailed job performance measures,
  which they completed and returned.
• Measures
• Self report measures Goldbergs Adjectives
  Scale measuring extraversion, conscientiousness,
  and emotional stability.
• Peer measures IPIP mimics of NEO-PI-R
  extraversion, conscientiousness, and emotional
  stability scales.
• IAT measures IAT measures of extraversion,
  conscientiousness, and emotional stability
  administered.
• Job performance measures
• Job performance scale (10 item scale) Overall
  job performance (1 item)

JobPerformance
Overall Performance
Table 1 Incremental Utility of Self Report and
IAT Trait Measures in the Prediction of Job
Performance
4
Do Multidimensional Implicit Association Tests
Work? On the Discriminant Validity of Different
IAT Formats Brian P. Siers, Maria Jacome, Bryan
Borenitsch, and Joel LaneFaculty Sponsor Dr.
Neil ChristiansenDepartment of Psychology -
Central Michigan University
Sample Participants 215 Undergraduates Procedur
e Participants were randomly assigned to 2
different conditions. - Group 1 3 Separate
IATs Participants completed 3 different trait
IATs (extraversion, conscientiousness,
stability), and a separate Flower-Insect
control IAT. - Group 2 Multidimensional IAT
Participants completed a positive/negative IAT
containing high trait and low trait adjectives in
related to 3 traits (extraversion,
conscientiousness, stability). Participants also
completed a Flower-Insect control IAT.
Research Questions 1) Will multiple D scores
computed on subsets of IAT stimuli be
discriminable? 2) Will multidimensional IAT
scores reflect more method variance than IATs
administered separately?
About the Implicit Association Test
Since its introduction in 1998, the Implicit
Association Test (IAT) paradigm has been adapted
to measure a variety of attitudinal (e.g. racism)
and dispositional (e.g. self esteem) constructs.
The IAT uses response latency as an index of
association strength between two concepts. An
example of the IAT is provided below.
The IAT has been adapted to measure personality
traits. Instead of using generally positive
words (see example below), the trait IATs use
high-trait (e.g. talkative, bold) and low-trait
adjectives (e.g. quiet, shy) related to a
specific trait. Personnel selection
applications often require assessment of multiple
traits. From an applied standpoint, it would be
advantageous to measure multiple traits using the
same IAT, as multiple IATs can be fatiguing for
the test taker.
Figure 1 Average Correlations Among IAT Trait
Scores and Relationships Between Trait Scores and
Control IAT
Results and Conclusions 3 Separate IATs (See
Figure) Average correlation between traits r
.174 Average correlation with control IAT r
.040 Multidimensional IAT (See Figure) Average
correlation between traits r .306 Average
correlation with control IAT r .140
Average Correlation with Control IAT
Average Intertrait Correlation
IAT EXAMPLE In Block 1 a test taker sees
screens like this Here Thoughtful
is a positive stimulus being sorted into either
the self/positive (press e) or
other/negative (press i) category. In this
step, the participant sorts through a series of
stimuli related to self (I, me, my), other (they,
them) and positive and negative adjectives. An
average response latency is computed for this
step.
In Block 2 the categories switch and test
takers see screens like this Here
Forgetful is a negative stimulus that is being
sorted. If participants are quicker associating
self-related words with negative adjectives (they
sort stimuli quicker in block 2), it is inferred
that they have a negative implicit self concept
or a strong association between self and negative
concepts or traits.
A difference score (block 1 latency block 2
latency) is computed. This D score serves as
an index of whether it was easier/quicker for a
test taker to associate their self concept with
positive or negative exemplars.
Self
Other Or
Or Positive
Negative Thoughtful
Self
Other Or
Or Negative
Positive Forgetful
In this case, LOW correlations are desirable, as
they are evidence that the dimensions being
measured by the IAT are different from one
another. Separate IATs demonstrated better
discrimination (lower correlations) among traits,
and reflected less method variance (lower
correlations with control IAT) than the combined
IAT. However, the combined IAT also
demonstrated good discriminability commensurate
with established (e.g. self report) measures of
personality traits. The observed values (both
intertrait correlations and control IAT
correlations) are within the acceptable range.
These promising results encourage further
research into the development of multidimensional
IATs.