Unobtrusive Psychological Profiling for Risk Analysis
Adam Szekeres and Einar Arthur Snekkenes
Department of Information Security and Communication Technology,
Norwegian University of Science and Technology, Gjøvik, Norway
Keywords:
CEO, Psychological Profiling, Unobtrusive Measures.
Abstract:
The main objective of this exploratory study is to present how publicly observable variables reflecting indi-
vidual choice can be used to construct psychological profiles suitable for predicting behavior in the context
of risk analysis. For the purpose of demonstration, this study aimed at testing the hypothesis that there is a
selection bias among chief executive officers (CEOs), which is manifested in their personal value structures.
Values capture motivational forces that serve as guiding principles in people’s life when making decisions.
From a risk management perspective, it is crucial to understand key decision maker’s motivation in order to
be able to prepare against potentially undesirable behavior. Therefore the second objective of this study re-
lates to a detailed characterization of the observed value structures among a group of CEOs. To accomplish
these goals a non-obtrusive data collection method is utilized that requires no direct access to individuals - the
Watson Personality Insights service provided by IBM - which infers value profiles based on written or spoken
text by the subjects. Results show that CEO value profiles differ from the general population in several ways.
Furthermore, slight differences were identified between the profiles of CEOs associated with moral hazard and
CEOs not associated with it. These findings indicate that there is a meaningful selection bias and these results
contribute to the real-world applicability of the CIRA method of risk analysis.
1 INTRODUCTION
A great amount of evidence suggests that key decision
makers can have a major impact (positive or negative)
on the safety and security of organizations and infor-
mation systems spanning across the entire range of
the corporate hierarchy (Cohan, 2002; Van Peursem
et al., 2007; Soltani, 2014). Within the economics
and management literature the tension between man-
agement interests and governance objectives is rec-
ognized as the principal-agent problem within agency
theory. Agency theory addresses the situation where
one party (principal) delegates work to another party
(agent) who is responsible for performing that work
on behalf of the principal. According to Eisenhardt
the theory is concerned with resolving two problems
that may arise in any agency relationship (Eisenhardt,
1989). The first problem relates to the possibility that
the agent’s and the principal’s desires and goals are in
conflict, and it is difficult or expensive for the princi-
pal to verify what the agent is actually doing (i.e. it is
difficult to verify that the agent’s behavior is appropri-
ate). The second problem arises from the difference
between the parties’ attitude towards risk, where the
principal and the agent might prefer different actions
due to different risk preferences.
As more and more critical infrastructures are un-
dergoing a radical change by the introduction of the
Internet of Things (IoT), social and economic stability
is increasingly dependent upon the decisions that peo-
ple in key positions make (Fosso et al., 2014). These
aspects of risks are often neglected within informa-
tion security risk analyses, as they mostly focus on
the technical aspects, while largely ignoring the cru-
cial human influence. It is suggested by Anderson
and Moore that information security problems should
be addressed from a broad range of perspectives
(e.g. economics, psychology, etc.), since stakehold-
ers might face various misaligned incentives, while
various psychological factors can be utilized to reveal
the ways in which people pose threat to information
systems (Anderson and Moore, 2009).
The Conflicting Incentives Risk Analysis (CIRA)
developed by Rajbandhari and Snekkenes aims to
bridge the existing gap by focusing on decision mak-
ers’ motivation when addressing risks (Rajbhandari
and Snekkenes, 2012). In order to enhance the exist-
ing method and make it applicable to real-world cases
it is necessary to update the method with relevant in-
sight about human behavior. To this end this study
210
Szekeres, A. and Snekkenes, E.
Unobtrusive Psychological Profiling for Risk Analysis.
DOI: 10.5220/0006858802100220
In Proceedings of the 15th International Joint Conference on e-Business and Telecommunications (ICETE 2018) - Volume 2: SECRYPT, pages 210-220
ISBN: 978-989-758-319-3
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
focuses on leader characteristics within an organiza-
tion and its consequences as the CEO role offers the
biggest potential to exert influence over the entire or-
ganization. However, the principles of the analysis
are applicable for the study of any type of stakehold-
ers (i.e. not limited to the CEO role). It is acknowl-
edged that people might be reluctant to be explicitly
subjected to risk analysis activities, which poses a ma-
jor obstacle in conducting the work, therefore in such
highly data-sparse environments, the analysis must
rely on publicly observable features as much as pos-
sible.
1.1 Problem Statement
The CIRA method investigates how a misalignment
between stakeholders’ motivation can result in var-
ious forms of risk. For the development of the
method, it is necessary to quantify and characterize
the strength of an individual’s motivation without di-
rect access to the subject, since subjects might be re-
luctant to reveal their real motivations or would be
tempted to mislead an analyst. Therefore for the en-
hancement of the method the objective is:
to select a suitable framework that captures a com-
prehensive set of human motivations,
to derive the motivational structure without direct
access to the individuals,
to use the knowledge derived from groups of in-
dividuals in settings where data about an individ-
ual’s past choices is scarce.
1.2 Research Questions
Based on the aforementioned requirements, the pri-
mary research question that is addressed in this study
was formulated as: can publicly observable variables
reflecting individual choice be used to construct psy-
chological profiles suitable for predicting behavior in
the context of risk analysis?
The following sub-questions are addressed in or-
der to answer the main research question:
1. Is it feasible to derive motivational characteristics
of CEOs using unobtrusive measures?
2. Is there a significant difference between the basic
human value structure of CEOs in comparison to
the general population?
3. Is there any significant difference between the
value profiles of CEOs associated with moral haz-
ard and the profiles who have no association with
it?
This work contributes to the literature of risk anal-
ysis by presenting how publicly observable decisions
from individuals can be utilized for the purpose of
information security risk analysis. The presented
method builds on an existing application, while the
purpose of the analysis differs significantly from cur-
rent domains of application. In order to illustrate
the method’s practicality, the study focuses on orga-
nizational leaders due to the fact that other classes
of stakeholders might not be allowed to interact offi-
cially with the public, due to company policies, how-
ever the approach can be applicable to any class of
key stakeholders (e.g. CFO, COO, CIO, CISO).
2 RELATED WORK
This section provides an overview about the psy-
chological theories, constructs and applications that
served as the foundations of this study.
2.1 Sources of Bias
There are several research perspectives that aim to
provide an explanation about how people with certain
traits or characteristics are self-selected to leading po-
sitions, how these characteristics are desirable on one
hand and how they might have a negative impact on
the organization’s objectives, and can have greater so-
cietal impacts. This section introduces two mecha-
nisms that could contribute to a selection bias in ex-
ecutive roles (i.e. personal attraction to a specific role
and selection of candidates by relevant stakeholders).
2.1.1 Bias by Personal Motivations
Boddy defines Corporate Psychopaths “as people
working in corporations who are self-serving, oppor-
tunistic, ego-centric, ruthless and shameless but who
can be charming, manipulative and ambitious” who
are drawn to corporations as they provide sources of
power, prestige and money (Boddy, 2005). There is
a distinction between the popular image of criminal
psychopaths, and corporate psychopaths. While the
former group is pictured as insane, suffering from
mental delusions, Corporate Psychopaths are out-
wardly charming, and engaging, skillful at manipu-
lating others to their own advantage, with a lack of
concern for the consequences of their actions, and
give a high priority for the pursuit of their own goals
and ambitions. The prospects of power, prestige and
money are assumed to be the main motivators that
draw individuals with psychopathic traits to the cor-
porate world. The ability to demonstrate desirable
Unobtrusive Psychological Profiling for Risk Analysis
211
traits that the organization values for a certain position
is easily exploited by such individuals by presenting
a charming facade and appear to be an ideal leader.
The risks that such individuals pose to the corpora-
tions they work for can take various forms e.g. acting
on the basis of pure self-interest, that may hamper or-
ganizational goal achievement, when self-interest and
greater corporate interests are misaligned. The link
between an organization’s questionable practices (e.g.
exploiting sweatshop labor, environmental pollution,
etc.) in pursuit of profit is the decision-maker, who
authorizes such activities. According to the argument,
these leader characteristics contribute to a less-than
desirable social responsibility by the organization.
Babiak, Neumann and Hare empirically investi-
gated a sample of 203 corporate professionals who
were selected by their companies to take part in a
management development program for indications of
psychopathic tendencies (Babiak et al., 2010). Ac-
cording to the authors, the lack of available coop-
erative subjects is a major obstacle when the aim
is to understand how a key decision maker’s per-
sonal characteristics can negatively influence others.
However there is a “growing public and media in-
terest in learning more about the types of person
who violate their positions of influence and trust, de-
fraud customers, investors, friends, and family, suc-
cessfully elude regulators, and appear indifferent to
the financial chaos and personal suffering they cre-
ate”. Large-scale Ponzi schemes, embezzlement, in-
sider trading, mortgage fraud, and internet frauds and
schemes, are some of the activities where psychopa-
thy was invoked as one explanation for such socially
destructive behavior. The study aimed at investigat-
ing the prevalence, strategies and consequences of
psychopathy in the corporate world. According to
the results very high psychopathy scores were ob-
tained from high potential candidates who held senior
management positions. An interesting finding of the
study had to do with how the corporation viewed in-
dividuals with many psychopathic traits. High psy-
chopathy scores were associated with perceptions of
good communication skills, strategic thinking, and
creative/innovative ability and, at the same time, with
poor management style, failure to act as a team player,
and poor performance appraisals (as rated by their
immediate bosses). The findings shed some light on
the complex association between situation-congruent
self-presentation and how psychopathic traits (al-
though not classified as Antisocial Personality Disor-
der) can be adaptive in corporate environments.
An empirical study investigated the link between
the Dark Triad personality traits and the Schwartz ba-
sic human values (Kajonius et al., 2015). The Dark
Triad (Machiavellianism, Narcissism, and Psychopa-
thy) is a popular grouping of individual differences
that represent antisocial personality traits below clini-
cal threshold. The antisocial aspect of the triad comes
from the shared underlying attitudes and modes of
behavior that characterize these traits. Entitlement,
superiority, dominance, manipulativeness, lack of re-
morse, impulsivity are among the key features of the
triad. The study found in two different cultures (i.e.
Swedish and American) that Hedonism, Stimulation,
Achievement and Power values were the most impor-
tant values held by individuals high on Dark Triad
traits. The authors conclude that those character-
ized by high scores on the Dark Triad traits, hold
values that imply the exclusion of others and self-
enhancement, viewing others as means toward self-
ish gains. The connection between Self-enhancement
values and the Dark Triad traits is referred to as dark
value system that has further moral implications.
2.1.2 Selection Bias by Requirements of the Role
Person-organization fit is a specialized area of inquiry
within the broader Person-environment fit studies that
aims to investigate how certain personality character-
istics influence the fit of the individual within organi-
zational settings. Morley (Morley, 2007) discusses a
recent shift in the recruitment process where the tra-
ditional focus on knowledge, skills, abilities (KSAs),
has shifted toward seeking an optimal fit between the
candidate’s personality, beliefs and values and the or-
ganization’s espoused culture, norms and values. Fur-
thermore, others suggest that work values are a core
means by which individuals judge their fit, and candi-
dates are attracted to organizations that exhibit char-
acteristics similar to their own, and in turn organiza-
tions tend to select employees who are similar to the
organization, which is a similar idea to Schneider’s
Attraction-Selection-Attrition (ASA) framework, that
identifies a similar fit at the personal level between
the candidate and the organization (Schneider et al.,
1995). Value congruence has become widely ac-
cepted as the defining operationalization of P-O fit
(Kristof-Brown et al., 2005).
On a more fine-grained level however the specific
roles within an organization pose a variety of specific
requirements (e.g. managerial role requirements are
very different from the requirements of a production
line worker). Using a large sample Lounsbury et al.
looked for a distinctive managerial profile that differ-
entiated them from workers in other occupations. The
investigation revealed a distinctive managerial per-
sonality profile in terms of the Big Five and other
measures of personality. In the following 9 person-
ality trait facets managers reached higher scores than
SECRYPT 2018 - International Conference on Security and Cryptography
212
non-managers: Extraversion, Assertiveness, Consci-
entiousness, Emotional Stability, Agreeableness, Op-
timism, Work Drive, Customer Service Orientation,
Openness. These results have practical implications
from the personnel selection perspective to guide the
search for candidates who possess the necessary traits
to increase the person-organization fit required for
specific job types (Lounsbury et al., 2016).
Knafo and Sagiv examined the relationship be-
tween several occupations and the value profiles of
the individuals working in the respective roles. From
the different work environments, the enterprising en-
vironment is the one that is mostly characterized by
material and concrete goals, and requires one to lead,
convince or manipulate others in order to achieve de-
sired organizational and financial goals. According
to the hypothesis Power and Achievement values are
most compatible with these requirements, while the
enterprising environment would inhibit the expres-
sion of Benevolence and Universalism values. The
occupations that were examined and most closely re-
sembled the enterprising environment were: manager,
banker, financial advisor. The results showed a strong
positive correlation between the enterprising occupa-
tion and both Power and Achievement values, while a
negative correlation was observed in relation to Uni-
versalism values. The study successfully differenti-
ated occupations based on the dominant values that
are required in each particular field, thus providing
further evidence about a selection bias in place.
In summary these research results suggest some
of the mechanisms by which individuals with certain
traits or characteristics are selected for specific jobs,
first by their own attraction to these positions, and fur-
thermore by the active involvement of the recruiters.
2.2 Conflicting Incentives Risk Analysis
The relevance of focusing on the motivation of
stakeholders is recognized in the Conflicting Incen-
tives Risk Analysis (CIRA) method (Rajbhandari and
Snekkenes, 2013). The method identifies stakehold-
ers (i.e. individuals), the actions that can be taken
by these stakeholders and the consequences of the ac-
tions. A stakeholder is always an individual who has
interest in taking a certain action within the scope of
the analysis. The procedure distinguishes between
two types of stakeholders: Strategy owner (the per-
son who is capable of executing an action) and the
Risk owner (whose perspective is taken - the person at
risk). At the core of the method is the economic con-
cept of utility, which captures the benefit by imple-
menting a strategy for each stakeholder. The cumula-
tive utility encompasses several utility factors, each
representing valuable aspects for the corresponding
stakeholders, thus modelling an individual’s motiva-
tion. Two types of risks are identified in the method:
Threat risk relates to the perceived decrease in the to-
tal utility for the risk owner, and Opportunity Risk re-
lates to the lack of potential increase of utility because
the strategy owner is not motivated to take actions that
would be beneficial for the Risk owner. Therefore
risk is conceptualized as a misalignment of incen-
tives between these two classes of stakeholders, and
risk identification is concerned with uncovering activ-
ities that would be beneficial for the Strategy owner
while being potentially harmful for the Risk owner
(Snekkenes, 2013). Therefore, threat risk closely re-
sembles the concept of moral hazard as it captures
a wide range of behaviors that are beneficial for one
party and detrimental for another (i.e. strategy owner
inflicting negative externalities on the risk owner - in-
creasing one’s utility, while causing a decrease in the
utility of someone else) (Dembe and Boden, 2000).
This study focuses on Threat risks that can be at-
tributed to the motivation of key decision makers.
2.3 Theory of Basic Human Values
The theory of basic human values by Schwartz
(Schwartz, 1994) identifies 10 distinct values that
are universally recognized across various cultures and
provide a unified and comprehensive view on the mo-
tivation of individuals. Values both represent desir-
able end goals and prescribe desirable ways of acting.
Schwartz summarizes the six core features that char-
acterize values:
Values are beliefs linked to affect.
Values refer to desirable goals that motivate ac-
tion.
Values transcend specific actions and situations.
Values serve as standards or criteria.
Values are ordered by importance.
The relative importance of multiple values guide
action.
Furthermore, all 10 distinct values in the theory
capture one of the three key motivational aspects that
are grounded in universal requirements of human ex-
istence: needs of individuals as biological organisms,
requisites of coordinated social interaction, and sur-
vival and welfare needs of groups. Values serve as
an internal compass guiding behavior, given that the
decision context or situation activates the relevant val-
ues. The values form a circular structure which cap-
ture a motivational continuum, where adjacent values
Unobtrusive Psychological Profiling for Risk Analysis
213
Figure 1: Circular value structure, with 4 higher dimensions
comprising of the 10 basic human values.
are compatible with each other, while opposing val-
ues are in conflict. The ten values are grouped under 4
higher dimensions as represented by Fig 1 (Schwartz,
2012).
The link between values and specific individ-
ual behaviors is a surprisingly neglected area of in-
quiry, and researchers show little agreement about
the strength of the value-behavior relationship. On
one end of the spectrum are researchers claiming that
values guide behavior (Rokeach, 1973), while others
claim that values rarely guide behaviors and not for
most people (McClelland, 1987). The behavior guid-
ing aspect of values was investigated by Bardi and
Schwartz in order to clarify the role of values in the
expression of behaviors (Bardi and Schwartz, 2003).
Their results suggest that values guide a range of be-
haviors, when the frequency of value expressive be-
haviors are investigated. A detailed analysis showed
that Stimulation and Tradition values correlate highly
with corresponding behaviors, while Hedonism, Self-
direction, Universalism, and Power values showed di-
minishing associations. Security, Conformity, Benev-
olence and Achievement values showed only weak
correlations with the behaviors that are expressive of
them. Based on these findings the predictive utility of
values is limited in a few ways: values are useful for
predicting the behavior when value-expressive behav-
iors are clearly defined and a selection is made from
a predefined set of options (e.g political party prefer-
ences and choosing university courses (Schwartz and
Bardi, 2001)). However, it is acknowledged in the
theory that most actions are expressive of more than
one value, and that the value structure that an indi-
vidual holds modifies his/her perception giving rise
to ambiguous interpretations of the same situation.
Therefore, the correspondence between values and
actions is expected to be highest in case of single de-
cisions expressing Tradition and Stimulation values,
and lower in other cases.
2.4 IBM Watson Personality Insights
Personality Insights (PI) is part of IBM’s artificial in-
telligence platform called Watson. Previously known
for defeating the top human players in Jeopardy, the
service these days is a comprehensive set of artificial
intelligence solutions available for the consumer mar-
ket. The service is utilized in a wide range of fields
including health care, weather forecast, electric load
optimization, etc. The PI utilizes advanced machine
learning solutions to uncover an individual’s psycho-
logical characteristics based on texts produced by the
person. The PI service’s main use cases involve tar-
geted marketing, customer care services, automated
personalized interactions, among several others. The
service produces profiles based on four different mod-
els of individual differences:
1. Big five personality model - one of the most
widely investigated and accepted model of per-
sonality that captures ve major dimensions about
one’s personality. These characteristics describe
relatively stable behavioral tendencies and modes
of experiences.
2. Needs - based on the earliest investigations into
human motivation capturing an individual’s high-
level desires.
3. Basic human values - values capture both desir-
able goals that people pursue and standards of act-
ing, thus providing a summary about what are the
underlying motivations behind one’s actions.
4. Consumption preferences - optimized for predict-
ing the user’s likelihood for buying a certain prod-
uct or engaging in different activities.
In terms of the Value profiles, the service cal-
culates the four higher-level profiles: Conserva-
tion, Openness to change, Self-enhancement, Self-
transcendence and Hedonism as separate values,
whereas the original formulation by Schwartz iden-
tifies four major ones, with Hedonism being part of
either Openness to change or Self-enhancement. For
each personality characteristic the PI computes two
scores: percentile scores and raw scores. “To compute
the percentile scores, IBM collected a very large data
set of Twitter users (one million users for English,
200,000 users for Korean, 100,000 users for each of
Arabic and Japanese, and 80,000 users for Spanish)
and computed their personality portraits. IBM then
SECRYPT 2018 - International Conference on Security and Cryptography
214
compared the raw scores of each computed profile to
the distribution of profiles from those data sets to de-
termine the percentiles. The service computes nor-
malized scores by comparing the raw score for the
author’s text with results from a sample population”
(IBM, 2017). While the percentile scores can provide
insights about the value distributions of a sample in
relation to the original sample it is not well-suited to
characterize an individual’s profile for the purpose of
choice predictions, because the value structure rela-
tive to a sample population does not necessarily cor-
respond to the individual’s own value priorities. To
allow comparison between different populations and
scenarios the service also provides raw scores that re-
flect a score the person would get when completing a
personality test. Therefore, raw scores are more suit-
able to compare with populations with existing means
and standard deviations, and to analyze value priori-
ties within and among individuals.
3 METHODS
3.1 Participants
The convenience sampling method produced the fi-
nal sample that consisted of 116 CEOs (105 male, 11
female), aged between 34-95 years (M = 59.41, SD =
9.23) with sufficient amount of texts for running accu-
rate analysis by the IBM Watson service. The amount
of text available for the individuals ranged between
264-11384 words (M = 3830.98, SD = 1672.28). The
majority of the subjects were born in the USA (N
= 52.6%), followed by India (N = 12.9%), United
Kingdom (N = 6.9%) and 21 other countries (N =
27.6%). 84.4% of the sample had at least bache-
lor or equivalent level degrees. The total compen-
sation for the CEOs in year 2016 ranged between
$45,936 - $46,968,924 (M = $15,988,276.78, SD
= $10,600,982.56) according to publicly available
sources (Salary.com, 2004).
3.2 Data Collection
In order to answer the Research Questions it was nec-
essary to run an initial pilot study to assess the feasi-
bility of the data collection activity. During the pilot
study the first step involved the identification of rele-
vant sources of data. To this end the Wikipedia article
on the List of chief executive officers of notable com-
panies was used that contains CEOs with diverse na-
tional and industrial backgrounds (Wikipedia, 2004).
At the time of the start of the data collection the list
consisted of 174 subjects. The second step involved
the identification of suitable sources of information
that could be linked to the individual and provided
sufficient input to the Watson service for achieving
it’s maximum precision (3000 words/subject is rec-
ommended by the service description). In this phase
we relied on video interviews, interviews published in
online newspapers, news articles, company commu-
nications and social media profiles. Although it was
possible to collect the necessary amount of data from
the individuals, the procedure was not feasible due to
high diversity of contexts, the uncertainty related to
the actual author of the texts and the time needed to
collect the data, so in the final data collection phase
this procedure was modified in the following way:
The search was restricted to videos published on
YouTube that (a) were in English, (b) the sub-
ject could be clearly identified while providing
his thoughts, and (c) were supplemented with cap-
tions.
The search then was executed by using the sub-
ject’s name with the following additional terms (in
the same order): - interview, talk, presentation. In
case the first search term did not provide sufficient
amount of text the next one was used.
In order to achieve as high validity as possible for
the analysis we aimed at collecting mainly inter-
views and discussions that are more spontaneous
and reflective in content (thus we aimed at min-
imizing the reliance on well-rehearsed communi-
cations or texts written by other parties for presen-
tation purposes).
Each video was carefully observed in real time
to check the accuracy of the captions and to en-
sure that only the subject’s utterances are ex-
tracted for analysis, while omitting any noise (in-
terviewer/audience questions, false transcriptions,
etc.)
A fresh install of Google Chrome was utilized in
incognito mode, to keep personalized search re-
sults to a minimum and to maximize the repro-
ducibility of the search results.
After a sufficient amount of text was collected
from the subjects, the texts were submitted to the Wat-
son PI service, that produced a profile for each indi-
vidual.
For the purpose of a more fine grained analysis,
CEOs that have been associated with various deci-
sions leading to moral hazard have been identified
in the current sample. To this end extensive web
searches were conducted with the name of the individ-
ual and the additional search term (e.g. fraud, scandal,
corruption). The first 20 search results were screened
Unobtrusive Psychological Profiling for Risk Analysis
215
for each subject in order to identify possible associa-
tions with moral hazard. Using a broad sense of the
moral hazard concept, any behavior was eligible for
inclusion when it had a negative effect on the repu-
tation of the organization by drawing public attention
to the underlying misconduct (irrespective of the na-
ture of the misconduct) and the behaviors were con-
ducted under the administration of the CEO in focus.
The included activities covered a wide range of be-
haviors (e.g. bribery of public officials, tax evasion,
accounting fraud, insider deals, ethical misconduct,
etc.). The procedure resulted in the identification of
31 CEOs (26.7% of the sample) associated with un-
desirable behavior, and enabled the value profile com-
parisons between the two classes of CEOs.
4 RESULTS
4.1 A Note on the Concept of Difference
It is important to note that the term ‘difference‘ can
have several meanings. In order to characterize the
differences we utilized several approaches. In the first
approach the percentile scores derived from the Wat-
son PI service were used, that inherently contains a
comparison between the subject’s results and the orig-
inal sample’s distribution, on which the service was
validated (N = 1 million users) (IBM, 2017). This
view provides an understanding about the CEO sam-
ple’s overall position for each value dimension. Due
to the meaning of percentile scores certain, expecta-
tions can be calculated on the number of value profiles
that are expected to fall within 1 SD from the means.
These assumptions were tested in the first procedure.
A second approach utilizes the raw scores derived
from the PI service, which are suggested to be equiv-
alent to the scores one would get when completing an
actual psychometric test (in this case any variant of
the several existing Schwartz value surveys (Schwartz
et al., 2001)). These scores can be compared to re-
sults established in different populations, therefore
are more suitable for comparing results obtained by
other researchers. The second procedure followed
this line of reasoning, and was mainly concerned in
identifying a difference in the rank ordering of val-
ues between CEOs and the general population. In this
sense, any difference in the ordering of the values (to
most important to least important) would be indicative
of a marked difference between the group of CEOs
and the general population.
However, rank orders in isolation do not provide
all the necessary information about and individual’s
trade-off decisions, since a preference reversal (i.e.
choice of different strategies with the same value or-
ders among individuals) is possible, while maintain-
ing the same value order. In light of this fact and in
accordance with the theory’s concepts it is the relative
importance of values that should be analyzed when
certain decisions are weighed against each other. Fur-
thermore, since many studies use different instru-
ments and methodologies for assessing value profiles
or use different levels of analysis, it was necessary
to enhance the compatibility and comparability of re-
search findings (Lindeman and Verkasalo, 2005). To
this end, in the third procedure the raw scores were
summed across all values, and each score was multi-
plied by the Sum
1
, to quantify each value’s contri-
bution to the overall utility (=1). The same procedure
was carried out for research results that served as ref-
erence for the comparisons. This approach provides
the assessment of an individual’s value structure in-
dependent of the instrument used for conducting the
profiling.
4.2 Comparison with Watson PI Sample
The first procedures aimed at detecting the existence
of a bias among the percentile scores among the five
higher level values. The percentile scores were trans-
formed by mapping them to a standard normal dis-
tribution, then for each value Kolmogorov-Smirnov
tests were conducted with a reference standard nor-
mal distribution (M = 0, SD = 1) to assess whether
the value score would be drawn from the same distri-
butions. The results indicate that all distributions are
significantly different from the reference normal dis-
tribution. All five, one-sample Kolmogorov-Smirnov
tests rejected the null hypothesis that the data fol-
lowed the normal distribution for variables: Conser-
vation (D = .741), Openness to change (D = .194),
Hedonism (D = .916), Self-enhancement (D = .657)
and Self-transcendence (D = .534), N = 116 each, and
p > .05 each). Fig 2 shows the distribution of all the
values based on the transformed percentile scores.
4.3 Comparison with National
Representative Samples
In the following procedure the raw scores have been
transformed to match with the original scale’s scoring
system used in the studies conducted by Schwartz and
Bardi (Schwartz and Bardi, 2001). The representative
or near-representative samples provide the necessary
comparison that allows for a more detailed descrip-
tion of the value profiles. Fig 3 shows the general
population’s value priorities compared with the CEO
value priorities based on the raw scores.
SECRYPT 2018 - International Conference on Security and Cryptography
216
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
-5 -4 -3 -2 -1 0 1 2 3 4 5
Standard normal distribution
Hedonism
Conservation
Openness to change
Self-enhancement
Self-transcendence
Figure 2: Distribution of CEO value percentiles relative to
reference standard distribution.
5.55
5.29
4.22
3.91
3.76
4.57
3.75
3.10
3.73
3.81
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
Self-transcendence
Openness to change
Self-enhancementHedonism
Conservation
Adjusted CEO mean raw scores
Mean scores based on 13 national
representative samples
Figure 3: Value profile comparison between sample of
CEOs and representative samples covering 13 nations.
4.4 Analysis of CEO Value Profile
For the purpose of individual level choice prediction
it is the relative importance among the values guiding
the actions, therefore for each behavior the relation of
all the values has to be considered. To this end the
CEO group’s average profile is analyzed in terms of
the relative importances among values based on the
procedure previously described. Additionally, based
on the previous classification of individuals with their
association with moral hazard, five independent sam-
ples t-tests were conducted on the raw scores to com-
pare the importance of each value for the two classes
of CEOs to detect any difference that might be in-
dicative of negative organizational outcomes. Table
1 shows the results of the performed t-tests. Open-
ness to change and Conservation values were signifi-
cantly different between the two groups. Specifically,
a slight decrease in both of these two values is associ-
ated with a value profile corresponding to undesirable
behaviors. Fig 4 illustrates the relative importance
scores among the CEO sample, the CEO sub-sample
associated with moral hazards and the general popu-
lation. Rank order of the values is marked above the
bars where the CEO sample’s ranking is followed by
the general population’s rank for each value. The *
symbol marks the values which have been identified
to be significantly different among the CEO groups,
based on the previous analysis.
5 DISCUSSION
This study aimed at exploring the basic human value
structure of CEOs by using text-based personality in-
ferences using the IBM Watson PI service. Our re-
sults suggest that there is a selection bias that mani-
fests itself through the individual value profiles. Ac-
cording to the results there are clearly identifiable
differences among the universally established value
structures in the general population and the sample of
CEOs. This marked difference is interpreted as an ev-
idence of the selection bias within leading positions
and the consequences of this distinct value structure
are discussed in this section, additionally identifying
further research directions.
The first remarkable difference among the value
profiles is manifested in the difference between the
rank order of values among CEOs and the general
population. While Self-transcendence values (i.e.
care for the welfare of closely related others, as well
as care for all the people and for nature) are most im-
portant for both groups the similarities between CEOs
and non-CEOs end at this point.
Openness to change (i.e. self-direction, indepen-
dence, creating, stimulation and seeking out chal-
lenges) ranks as the second most important value
in case of corporate leaders, while it is the sec-
ond least important motivational factor for the pop-
ulation. Openness to change and Conservation val-
ues can be found at opposing sides of the moti-
vational circumplex, which reflects that decisions
that promote the obtaining of a particular value in-
hibit the simultaneous fulfillment of the competing
value. Therefore a high priority given to Openness
to change values would result in choices increas-
ing novelty and chances for expressions of indepen-
dent action at the expense of maintaining stability
and stability. Self-enhancement values (i.e. expres-
sion of competence, achievement of status and con-
trol over others) rank at the third position for CEOs,
while it is the least important motivational force in
the general population. Although one might expect
that leaders of world-leading organizations (express-
Unobtrusive Psychological Profiling for Risk Analysis
217
Table 1: Results of the independent samples t-tests among two CEO groups.
CEO raw scores associ-
ated with moral hazard
(n = 31)
CEO raw scores not as-
sociated with moral haz-
ard (n = 85)
M SD M SD t-test
Values
Self-transcendence 0.82 0.01 0.82 0.01 ns
Openness to change 0.78 0.02 0.79 0.02 2.20*
Self-enhancement 0.65 0.02 0.65 0.02 ns
Hedonism 0.61 0.01 0.61 0.02 ns
Conservation 0.59 0.02 0.60 0.03 2.07*
*p < .05; two-tailed.
Note. M = Mean. SD = Standard Deviation
0.236
0.227
0.188
0.177
0.172
0.237
0.226
0.188
0.178
0.170
0.241
0.198
0.164
0.197
0.201
0.000
0.050
0.100
0.150
0.200
0.250
0.300
Self-transcendence Openness to change* Self-enhancement Hedonism Conservation*
CEOs not associated with moral hazard CEOs associated with moral hazard Cross cultural group
Value priorities: (1) CEO - (2) General population
1. 1. 2. 4.
3. 5. 4. 3. 5. 2.
* significant difference between the two CEO groups
Figure 4: Comparison between the relative importance of values among two groups of CEOs and general population.
ing power and achievement values) would be mainly
motivated by Self-enhancement values at the expense
of Self-transcendence values, these results contradict
this expectation. The rank order difference of Self-
enhancement values between non-CEOs (5.) and
CEOs (3.) however clearly expresses their preference
for high social status and prestige.
While for non-CEOs, the second most important
motivational tendencies relate to Conservation values
(i.e. security, safety of self and of society, restraint
of actions likely to harm others, respect for customs),
these aspects of behavior seem much less important
to leaders, as it ranks the lowest on the their mo-
tivational hierarchy, indicating that actions promot-
ing Conservation values have a much lower intrinsic
motivational effect (e.g. in order to make an action
appear at least as rewarding as an action expressing
Openness to change values it has to be incentivized
much more externally).
The relative importance of values matches closely
with the various Enterprising value profiles identified
by Knafo and Sagiv, placing CEOs close to other
occupations characterized by material and concrete
goals, leading and manipulating people (occupations
within the category that have similar value profiles:
financial advisor, banker, manager) (Knafo and Sagiv,
2004).
The final analysis uncovered differences in the
SECRYPT 2018 - International Conference on Security and Cryptography
218
value profiles of two CEO groups when their associa-
tion with moral hazard is taken into account. Specif-
ically, a slight, but significantly lower relative impor-
tance attributed to Openness to change and Conser-
vation values was associated with various undesirable
behaviors that can be detrimental to the reputation of
the organization lead by the particular CEOs.
A limitation of the present study is the small sam-
ple size, which can be extended in further studies,
since the method of analyzing value profiles by using
the Watson PI service is a feasible method for gather-
ing information about the motivation of decision mak-
ers for the purpose of risk analysis.
Furthermore, a more detailed description and clas-
sification of the various forms of moral hazard would
have the potential to elaborate on the connection be-
tween the particular value profile displayed by a strat-
egy owner and the level of impact that was inflicted
upon the organization, to have a better assessment of
the risks relating to particular individuals.
Future work will focus on the issue of how other
observable features (e.g. gender, age, occupational
choice (Dohmen et al., 2011), consumer preferences
(Kassarjian, 1971) or various forms of online behav-
ior with digital footprints (Kosinski et al., 2013), etc.)
can be utilized for the construction of psychological
profiles suitable for predicting behavior in the context
of risk analysis. In particular, it is crucial to identify
observable features that can significantly reduce the
uncertainty associated with an observable’s ability to
convey information about a specific motivational trait.
6 CONCLUSION
In summary, this exploratory study is the first one pre-
senting how publicly observable traces of individuals
can be used for constructing a psychological profile
suitable for risk analysis. The study presented a de-
tailed description of the procedures necessary for un-
covering the motivational structure of leaders of vari-
ous organizations utilizing an unobtrusive psycholog-
ical profiling method. The procedure was conducted
by means of textual analysis based on publicly avail-
able written or spoken statements by the subject, and
the results supported the hypothesis that key decision
makers’ motivation is significantly different from that
of the general population that is interpreted as evi-
dence of a meaningful selection bias. Presenting the
motivational structure in terms of basic human values
in a form that is independent of the instruments uti-
lized provides an useful input for comparing different
stakeholder’s motivation, and for analyzing a poten-
tial candidate’s similarity with the profiles established
here. Finally, the distinct value profile associated with
undesirable behaviors can be helpful during the selec-
tion of candidates for a leading position by the board
of directors in determining the potential risks result-
ing from the employment of a specific individual.
ACKNOWLEDGEMENTS
This work was partially supported by the project IoT-
Sec Security in IoT for Smart Grids, with number
248113/O70 part of the IKTPLUSS program funded
by the Norwegian Research Council.
REFERENCES
Anderson, R. and Moore, T. (2009). Information security:
where computer science, economics and psychology
meet. Philosophical Transactions of the Royal Soci-
ety of London A: Mathematical, Physical and Engi-
neering Sciences, 367(1898):2717–2727.
Babiak, P., Neumann, C. S., and Hare, R. D. (2010). Corpo-
rate psychopathy: Talking the walk. Behavioral sci-
ences & the law, 28(2):174–193.
Bardi, A. and Schwartz, S. H. (2003). Values and behavior:
Strength and structure of relations. Personality and
social psychology bulletin, 29(10):1207–1220.
Boddy, C. R. (2005). The implications of corporate psy-
chopaths for business and society: An initial exam-
ination and a call to arms. Australasian Journal of
Business and Behavioural Sciences, 1(2):30–40.
Cohan, J. A. (2002). " i didn’t know" and" i was only doing
my job": Has corporate governance careened out of
control? a case study of enron’s information myopia.
Journal of Business Ethics, 40(3):275–299.
Dembe, A. E. and Boden, L. I. (2000). Moral hazard:
a question of morality? New Solutions: A Jour-
nal of Environmental and Occupational Health Policy,
10(3):257–279.
Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp,
J., and Wagner, G. G. (2011). Individual risk at-
titudes: Measurement, determinants, and behavioral
consequences. Journal of the European Economic As-
sociation, 9(3):522–550.
Eisenhardt, K. M. (1989). Agency theory: An assess-
ment and review. Academy of management review,
14(1):57–74.
Fosso, O. B., Molinas, M., Sand, K., and Coldevin,
G. H. (2014). Moving towards the smart grid: The
norwegian case. In Power Electronics Conference
(IPEC-Hiroshima 2014-ECCE-ASIA), 2014 Interna-
tional, pages 1861–1867. IEEE.
IBM (2017). The science behind the service. [Online; ac-
cessed 14-February-2018].
Kajonius, P. J., Persson, B. N., and Jonason, P. K. (2015).
Hedonism, achievement, and power: Universal values
Unobtrusive Psychological Profiling for Risk Analysis
219
that characterize the dark triad. Personality and Indi-
vidual Differences, 77:173–178.
Kassarjian, H. H. (1971). Personality and consumer behav-
ior: A review. Journal of marketing Research, pages
409–418.
Knafo, A. and Sagiv, L. (2004). Values and work environ-
ment: Mapping 32 occupations. European Journal of
Psychology of Education, 19(3):255–273.
Kosinski, M., Stillwell, D., and Graepel, T. (2013). Pri-
vate traits and attributes are predictable from digital
records of human behavior. Proceedings of the Na-
tional Academy of Sciences, 110(15):5802–5805.
Kristof-Brown, A. L., Zimmerman, R. D., and Johnson,
E. C. (2005). Consequences of individuals’fit at work:
A meta-analysis of person–job, person–organization,
person–group, and person–supervisor fit. Personnel
psychology, 58(2):281–342.
Lindeman, M. and Verkasalo, M. (2005). Measuring val-
ues with the short schwartz’s value survey. Journal of
personality assessment, 85(2):170–178.
Lounsbury, J. W., Sundstrom, E. D., Gibson, L. W., Love-
land, J. M., and Drost, A. W. (2016). Core personality
traits of managers. Journal of Managerial Psychol-
ogy, 31(2):434–450.
McClelland, D. C. (1987). Human motivation. CUP
Archive.
Morley, M. J. (2007). Person-organization fit. Journal of
Managerial Psychology, 22(2):109–117.
Rajbhandari, L. and Snekkenes, E. (2012). Intended ac-
tions: Risk is conflicting incentives. Information Se-
curity, pages 370–386.
Rajbhandari, L. and Snekkenes, E. (2013). Using the con-
flicting incentives risk analysis method. In IFIP Inter-
national Information Security Conference, pages 315–
329. Springer.
Rokeach, M. (1973). The nature of human values. Free
press.
Salary.com (2004). Executive compensation it starts with
the ceo. [Online; accessed 14-February-2018].
Schneider, B., GOLDSTIEIN, H. W., and Smith, D. B.
(1995). The asa framework: An update. Personnel
psychology, 48(4):747–773.
Schwartz, S. H. (1994). Are there universal aspects in the
structure and contents of human values? Journal of
social issues, 50(4):19–45.
Schwartz, S. H. (2012). An overview of the schwartz theory
of basic values. Online readings in Psychology and
Culture, 2(1):11.
Schwartz, S. H. and Bardi, A. (2001). Value hierar-
chies across cultures: Taking a similarities perspec-
tive. Journal of cross-cultural psychology, 32(3):268–
290.
Schwartz, S. H., Melech, G., Lehmann, A., Burgess, S.,
Harris, M., and Owens, V. (2001). Extending the
cross-cultural validity of the theory of basic human
values with a different method of measurement. Jour-
nal of cross-cultural psychology, 32(5):519–542.
Snekkenes, E. (2013). Position paper: Privacy risk analy-
sis is about understanding conflicting incentives. In
IFIP Working Conference on Policies and Research in
Identity Management, pages 100–103. Springer.
Soltani, B. (2014). The anatomy of corporate fraud: A com-
parative analysis of high profile american and euro-
pean corporate scandals. Journal of business ethics,
120(2):251–274.
Van Peursem, K. A., Zhou, M., Flood, T., and Buttimore,
J. (2007). Three cases of corporate fraud: an audit
perspective.
Wikipedia (2004). List of chief executive officers
Wikipedia, the free encyclopedia. [Online; accessed
06-December-2017].
SECRYPT 2018 - International Conference on Security and Cryptography
220