Psychological profiling of world leaders

“Magical models” are hard to find, but mathematical modeling shows potential to help predict behavior in various scenarios.

By Dean Hartley and Ken Jobson

What would Putin do?

What would Putin do?

According to recent news reports, the United States has created a psychological profile of Russia’s Vladimir Putin [10, 18]. Whether the claim relating to Putin is true or not, the idea has merit. Jerrold Post edited a book [32] describing this type of profiling, including leaders who have had, for better or worse, extraordinary influence on world events. Recent examples include Saddam Hussein of Iraq, Muammar Gaddafi of Libya, Mahmoud Ahmadinejad of Iran, Bashar al-Assad of Syria and Kim Jong-un of North Korea.

The current annexation of the Crimean peninsula by Russia and the claims that Eastern Ukraine is populated by ethnic Russians who should be part of Russia strongly resemble Hitler’s claims and actions with regard to the Sudetenland of Czechoslovakia prior to World War II. In each case it would be nice to have a magic model of the leader that would foretell what the person would do in any given situation. Such predictions would be valuable whether the situation was diplomatic, full-scale war or something in between.

Not surprisingly, different cultures tend to produce different leaders.

Not surprisingly, different cultures tend to produce different leaders.

Unfortunately, magic models are hard to find. The question is, can we build something that is good enough to use and can we define its shortcomings so that we don’t believe improper predictions and end up worse than if we had no model? Post’s work describes profiles that were used as stand-alone text models. This type of model will continue to be needed; however, recent work in creating ontologies of irregular warfare [16] supports integrating these profiles into computer models that include significant individual actors, group actors and population models. Computers require more completely defined models (as they are too simple-minded to handle hints and allusions) and variable scenario factors.

In the following sections, we will discuss the current and future states of profiling (modeling) individuals, in particular world leaders. We begin with some sample expectations of profile results. In order to produce profiles, we should understand what things are significant in modeling a person. We start with a discussion of human nature, something that must underlie any model of a particular person. We follow this with a discussion of the influences on an individual’s “personal nature.” Having laid the groundwork, we then discuss the current state of profiling and new additions that are now or will soon be parts of profiling. We end with sections on caveats and conclusions.

Expectations

Magic models aside, what might we expect to find? First, we would not be surprised to find differences among leaders from different cultures. Consider democracies with validly elected leaders, autocracies with leaders chosen from a power elite by the power elite, countries with inherited leadership and terrorist groups.

In democracies, it would not be surprising to find that most leaders are competitors, think well of themselves, are gregarious, are compromisers and are related to used car salesmen in matters of truth. Truly pathological candidates would stand a fair chance of being weeded out in the election process. However, moderate pathologies might include excessive narcissism and episodic mood disorders.

In autocracies, it would not be surprising to find that leaders are callous and ruthless, are self-interested, distrustful and uncompromising, and regard truth as a malleable tool. Pathologies could be expected to include sociopathic traits, excessive narcissism and paranoia.

In countries with inherited leadership, virtually any natural personality might pertain to a new leader. However, depending on the personality of the predecessor, the new leader’s upbringing might produce a deep sense of entitlement, a slavish devotion to the predecessor or reactions against the predecessor. Pathologies could include paranoia, pathologic narcissism or fixity (deficiencies in seeing social complexity, harboring fixed preconceptions and world view).

In terrorist groups, one would expect the leaders to be pathological as classified by traditional categories; however, that appears not to be the case. Rather, characteristics of these leaders appear to involve stimulus seeking, extraversion, severe action orientation and fixity.

These characterizations illustrate some of the potential range of differences among leaders. However, even if the characterizations were right as generalizations, each leader is a unique person, requiring evaluation. In particular, producing a characterization of a particular person requires attempting to find neutral ground that avoids preconceptions. Practically speaking, approaching this requires multiple viewpoints in the model building process.

Human Nature

Throughout history, human nature has been constant. However, there is reason to believe that the (pre-historical) invention of agriculture changed human societies, introducing the possibilities of cities and civilization and inducing changes in human nature through the necessity of dealing with denser populations [39]. Humans are complex adaptive systems within complex adaptive systems. Since that time human nature has been regarded as consisting of three parts: the human as a biological being, overlaid with the human as a social being, and overlaid with the human as a psychological being, all within the context of the external world.

According to very recent indications, human nature may be undergoing another change. The creation of computers, smart phones and other information services appears to be having a profound effect on at least some humans. These digitally immersed people have different initial reactions to stimuli and different modes of cognition and decision-making from those of their non-digitally enhanced peers [13, 35]. Further, advances in medicine and neuroscience are creating the fields of cognitive optimization/cognitive enhancement and nootropics, pharmacological agents for cognitive enhancement [26, 33]. Mind- and mood-altering chemical agents have been known at least as far back as the Pythian oracles of ancient Greece [40]. However, these agents had uncertain effects and were not used generally. As suggested in Figure 1, this would mean that models of humans as bio-psycho-social beings would be inadequate. We should be modeled as bio-psycho-social-techno/info beings.

Influences on ‘Personal Nature’

Figure 2: Influences on a person’s psychology.

Figure 2: Influences on a person’s psychology.

If “human nature” describes those things that are common to all of us, “personal nature” describes those things that make us different. The relevant differences for profiling are the psychology and behavior of the individual. Figure 2 shows the influences on a person’s psychology (using the term broadly).

A person’s cognitive state and inherent traits lead to biases that are now better understood. For example, the incomplete or attenuated manifestation of depression brings a negative bias. Even the person’s chronotype (morning or evening-person) may advise when in the day the person is more likely to say “yes” [9]. A person’s genetics certainly influences his psychology. For example, genes explain approximately half of the variance in human aggression [25, 20], and there is research indicating a specific MAOA gene allele influences anger control [8]. The physical environment can even influence the expression of genes (epigenetics), which influences psychology.

A person’s history and social and physical environment influence his psychology. For example, these combine in the potential for hubris in political leaders, often increasing with the duration of power [28]. Recent progress in understanding the complexity of human communication in general and within the digital matrix is helpful in prediction and influence. Alex Pentland’s work at MIT on the dialectic of social signals, discussed in the books “Honest Signals” [29] and “Social Physics” [30], addresses this.

Psychological traits are also subject to drug influence. Even casual THC (the active ingredient in marijuana) use influences types of impulsivity [24]. There are also medications designed for such influence, such as agents with serentic (anti-aggression) effects and others that influence impulsivity.

The last influence in the diagram is “Technology Filters.” The digitally immersed are subject to influence, search for and evaluate data, and make decisions differently from non-digitally immersed people. Increasingly, they even connect with others differently. The human-computer interfaces and the digital world inhabited are expanding parts of the human model.

Purposed Modification

Up to this point, we have considered a person as relatively static, i.e., changes took place in the past. The techno/info parts of the model permit new ways for a person to change or be changed.
Computer-assisted persuasion (captology), social media with its potential for MIP (massive interpersonal persuasion), mobile persuasion and expanded digital and social information sources [12, 13] now influence people and are methods for externally manipulating people through information technology. Mimicry of human behavior (acting similarly to the behaviors of the other person) and mood contagion lead to more successful negotiations [2, 21]. The available information matrix and the individual’s facility to use it lead to self-modification, such as affecting the individual’s openness to iterative inductive discovery versus fixity of mind, even fixity of world view.

The “pharmaceuticals” influence supplies another avenue for changing psychology. Treatments of many general medical, neurologic and psychiatric conditions influence cognitive capital. Illness can produce mild cognitive decline, attention deficits and mood and anxiety disorders, influencing cognitive capital. The basic use of digital enhancement, exercise, chronotherapeutic optimization, the current, even if temporary, influence of affect, positive expectation and the field of nootropics, are areas of interest [33, 26, 37]. Biologic cognitive enhancement effects are (currently) largely experimental. Cognitive enhancement effects range from genetic considerations to Sahakian’s recent selected use of modafinil at the Cambridge Brain Institute [35] and by extension the prospect of the use of D-cycloserine to promote neuro-plasticity (new learning) [19].

Modeling a Person

What exactly needs to be evaluated and how can this be done? If we were to try to model a person, we would certainly want to gather all the information about him that we could find. We would also want to identify the scenarios into which we would place the person/model for which we would require predicted actions. We would have to analyze the information and then construct the model.

Historically, profiling has been based on personality and intellectual assessment, integrated with a life narrative. We can understand this by starting with a simple form of profiling. Immelman [18] administered a diagnostic test of personality variables (in this case at a distance) of Vladimir Putin and performed an assessment of his personality. Immelman used the Millon Inventory of Diagnostic Criteria (MIDC), “which yields 34 normal and maladaptive personality classifications.”

Figure 3: Required disciplines.

Figure 2: Influences on a person’s psychology.

Two issues need to be addressed: personality assessments in general and “at a distance” assessments. When the authors were in college decades ago there were already several personality assessment instruments in use, such as the Minnesota Multiphasic Personality Inventory (MMPI), first published in 1943. Many such instruments are administered only by professional psychologists and psychiatrists. Other instruments such as the Myers-Briggs Type Indicator assessment are popularly administered at sales force meetings and even self-administered. This long history of personality assessment has allowed for considerable research into the repeatability of the assessments, both longitudinally for a particular person and across administrators.

Further, considerable research has been performed concerning the nature of the personality variables assessed by the instruments, i.e., what are the underlying factors that are being observed and how do they correspond to the real-world actions of people. As might be expected, there are differences among the various instruments in reliability and correspondence to reality. However, at this point, there are several very good instruments available.

Given that one or more good instruments are chosen, the question arises concerning “at a distance” assessments. As Post [32, page 37] notes, in some cases researchers have been able to administer tests to the leaders directly; however, often the tests must be filled out by the clinician or researchers based on documents and speeches given by the leaders. Post comments that, despite the fact that most of these documents are prepared by speechwriters, in general these speechwriters know the clients and have been chosen by the leaders. He says that research has generally validated the scores.
In contrast to Immelman’s one-test example, Post [32] presents a more complex set of examples of how profiling has been done. Several of the chapters in Post’s book describe different approaches to psychological profiling. These go beyond the use of a single instrument and include several types of analysis with a pre-defined set of issues that need to be addressed. The details are omitted here; however, there are some common elements of profiling.

  1. Attributes: A portion of the psychological profile consists of evaluating standardized attributes (model inputs).
  2. Output shaping: Another element of the profile defines expected behaviors (model outputs).
  3. Thinking: The inputs and outputs are connected by theories (creating the model logic).
  4. Finally, part of the profiling process involves methods for analyzing data to define the model.

In current (and future) profiling, narrative histories are constructed from in-person interviews that include developmental history, family, school and culture. One looks for mentors, talents, source of influence (now digital as well as traditional), role models, traumas, when did the person feel “in place” and when an “outsider.” In the personality/cognitive assessment, one evaluates empathy, hubris, gratitude, beliefs, evidence of charisma, values and goals. One also wants to evaluate intellectual, emotional and social capital, current (internal) models, metaphors and preconceptions, the person’s adaptive learning style, leadership methods and flexibility versus fixity. The medical and psychiatric history should look for any substance abuse and “process addictions.”

One wishes for personal interviews, neurocognitive testing, psychological evaluation and neuropsychologic tests. Formalized testing, if available, would include intellectual assessment such as the WAIS (Wechsler Adult Intelligence Scale) and an evaluation of personality utilizing the Five Factor Model, e.g., Costa-McCrae and its derivatives. For many, the evaluation must be at-a-distance by gathering “tiles” to construct an incomplete mosaic. The matrix of life and narrative life history are critical.

The narrative is then integrated with the personality/cognitive assessment and the medical and psychiatric history. The goal is to capture the multi-dimensional nature of the person. The approach involves multiple scientific disciplines and multiple independent measures.

Profiling in the Future

From our discussion of possible alterations of human nature, we would expect that there should also be changes in profiling. Considerations of digital enhancement and physiologic and pharmacological cognitive enhancements would require special attention. If computer modeling is a consideration, the profile will need to be constructed with explicit attention to alternatives in personality expression, together with the cues for selecting alternatives and the associated probabilities.

Further, new data collection and analysis methodologies can be brought to bear. Future profiling adds new and emergent sources of information to enhance these assessments. The science of “social physics” is a quantitative assessment of interpersonal idea- and information-flow and spread dynamics within traditional and digital systems. Social physics uses real-time audio-visual monitoring and the techniques of mining big data, and it has evidence-based predictive values, views of individuals, networks and ideas and combination of ideas for emergent understanding [30]. It enriches signal analysis of speech and other social signals and the pattern of interactions [2, 7, 27].

Figure 3: Required disciplines.

Figure 3: Required disciplines.

The blend of advances in reading the dialectic of social signals and social physics has advanced toward artificial social intelligence [38]. Truth detection, both human and machine augmented, has also advanced [11, 22, 23, 17, 4, 14]. Similarly, the use of emotional social signal analysis of anger, contempt and disgust have been reported to be predictive of near-term aggression by political leaders.

Access to the individual genome is rapidly becoming possible with greater ease. Genetic information does not determine individual behavior. However, knowledge of genetic influences is emerging and, with the expanded advance of genetics and big data, should increase. Currently we have gene-environment associations with some predictive power. One example is the increased risk for psychosis conveyed by even mild cannabis use with a history of child abuse and a certain COMT gene allele [1]. There are already published papers of successful genome editing in primates [20], which is being used to create twin monkeys that vary only in known genetic locations. The concept is to use the technique to study brain disorders. Behavioral genetics is a field that holds promise for understanding some of these biological influences.

The cultural/legal milieu information would require sociological, cultural anthropological and legal evaluation. If the person were not a complete dictator, information would be required on the governmental, quasi-governmental and the important social and commercial organizations. Together these would be informative on the types of reactions that could be expected. These would require sociological, political science and economic analysis. Scenario identification would require an historian with military and diplomatic expertise. In addition, selective crowd sourcing (an example of collective learning) may be useful in forecasting future events to support scenario development (e.g., The Good Judgment Project led by Philip Tetlock at the University of Pennsylvania [36]). Model construction would require operations research, big data analytics and computer science expertise.

Those familiar with the history of operations research will recognize the need for a classical multidisciplinary O.R. team:

  • psychiatric, psychological and neurocognitive science expertise;
  • sociological, cultural anthropology and legal expertise;
  • political science and economics expertise;
  • historical, military and diplomatic expertise;
  • social physics, human communications, truth detection and selective crowd sourcing; and
  • operations research, big data analytics computer science expertise.

Let us presume such a team was put together. Some details will help identify what can and cannot be expected from the output of the team. The first three elements from Post’s work (with new additions in red) are illustrated in Figure 4 as a conceptual model of a leader. The fourth element (with new additions) comprises the methodology for creating the model. We should expect that this structure would be found in any modeling effort: model inputs, model outputs, model logic and methodologies for defining the input values. In addition, any such model should be accompanied by an assessment of the degree and range of validity of the model. That is, we should know how much and under what circumstances the model can be trusted – and how much and under what circumstances it cannot be trusted. Practically speaking, completing such an assessment with a high level of accuracy and precision is impossible; however, it must be attempted.

Figure 4: Conceptual model of a leader.

Figure 4: Conceptual model of a leader.

Caveats

One much quoted aphorism, attributed to George E. P. Box [5], is appropriate: “Essentially, all models are wrong, but some are useful.” For a model to be useful, its creators should develop an understanding of its problems and clearly explain them to all users. We will call these explanations “caveats.” Caveats articulate the flaws and weaknesses of the profile to the users of the model.

The process of finding such problems is called verification and validation (V&V). Adding the determination of the usefulness for a particular purpose (accreditation) produces VV&A. VV&A can be difficult for models of well-understood systems; however, when human, social and cultural behaviors (HSBC) are involved, the systems are poorly understood, and the VV&A process becomes extraordinarily difficult. The Defense Advanced Research Projects Agency (DARPA) report describes how VV&A should be conducted in such cases [6]. Because profiles of world leaders would be expected to be used for policy decisions, extreme care in developing and explaining their caveats is most definitely required. Saltelli and Funtowicz [34] wrote on this and developed seven rules for the process they call sensitivity auditing. The first three are particularly appropriate here.

  1. Avoid using the model to advance an agenda.
  2. Find and explain both the explicit and tacit assumptions.
  3. Do not ignore or hide uncertainties in the model inputs that drive the outputs to preferred policy choices.

For profiles, we have developed some particular caveats and categories of caveats.

  • The anchoring bias of initial opinion and the first point above echo the warning in a synthesis of some of Cicero’s writing, “Tell me what you want to believe and I will show you the evidence.”
  • Behavioral predictions that are “point estimates” are problematic. “Worst case/best case” type predictions may be more useful than point estimates. Even range predictions of behaviors should be considered uncertain.
  • All predictions should be scenario constrained. That is, they should be formatted as “within this situation, this is the prediction.”
  • Weaknesses in the data should be explicitly specified. For example, “this type of data is weak for all profiles” and “this type of data is weak for this particular person.”
  • Weaknesses in the analysis should be explicitly specified in a similar manner.
  • Mitigations for investigator biases should be explicitly explained.

Conclusions

Humans are now augmented digitally, and efforts to augment biologically are being studied. We are now augmented bio-psycho-socio-techno/info beings. This means profiling must change. However, we still have aspects that are predicatively irrational [2], with preconceptions and biases, generally using frugal heuristics instead of exhaustive search to make decisions [15]. Traditionally and still centrally, the subject’s contextualized life narrative is integrated with an intellectual and personality/behavioral assessment to produce a profile.

Creating a profile or model of a person can now take advantage of recent advances in neuro-cognitive science and digital monitoring, communications and information-processing, armed with selective crowd sourcing data mining and using the analytics of big data. These developments along with improved trans-domain mixing bring improved granularity to the profile.

All modelers bring their frames of reference, talents, limitations, biases and preconceptions. The anchoring bias of initial opinion, discussed in “Predictably Irrational” [2], underlines that neutral ground is elusive, cautions against a single vision and advises an iteratively inductive approach. This suggests the value of multi-disciplinary modelers looking for areas of consensus. We now need teams of augmented humans evaluating augmented humans.

People are building models of – profiling – world leaders. And other people are making decisions based on these profiles. It behooves us to ensure that the models are as accurate as possible and that the decision-makers have a very good understanding of their shortcomings. ORMS

Dr. Dean Hartley is the principal of Hartley Consulting. Previously he was a senior member of the Research Staff at the Department of Energy Oak Ridge Facilities. He received his Ph.D. in piecewise linear topology from the University of Georgia in 1973. Hartley is a past vice president of INFORMS, a past director of the Military Operations Research Society, a past president of the Military Applications Society and a member of the College on Simulation of INFORMS. Hartley’s interests include modeling of irregular warfare, verification and validation of models, psychopharmacology modeling and simulation. His website is http://dshartley3.home.comcast.net.

Dr. Ken Jobson has a clinical practice in psychiatry and psychopharmacology, Psychiatry and Psychopharmacology Services PC in Knoxville, Tenn. He is the founder and is chairman of the board of the International Psychopharmacology Algorithm Project (www.ipap.org), on the clinical faculty at the University of Tennessee, Department of Psychiatry, and co-editor of a textbook, “Treatment Algorithms and Psychopharmacology.” He has facilitated the establishment of algorithm projects in Europe and Asia.

References

  1. Alemany, S., B. Avias, M. Fatjo-Vilas, H. Villa, J. Moya, M. I. Ibanez, G. Ortet, C. Gasto, and L. Fananas, 2014, “Psychosis-inducing efforts of cannabis are related to both childhood abuse and COMP genotypes,” Acta Psychiatric Scandinavica, Vol. 54, No. 1, pp. 54-62.
  2. Aviely, D., 2009, “Predictably Irrational,” Harper Collins Publishers, London.
  3. Bai, Shoutian, Tinqshaw Zhu and Li Chang, 2012, “Big Five Personality Prediction Based on User Behavior at Social Networking Sites,” arXiv:1204.4809 [cs.CY].
  4. Bhaskanan, N., I. Nwogu, M. G. Frank, and V. Govindaraju, 2011, “Lie to Me: Deceit Detection via Online Behavioral Learning,” Conference presentation 2011, Centers for United Biometrics and Sensors, SUNY Buffalo.
  5. Box, George E. P. and Norman R. Draper, 1987, “Empirical Model-Building and Response Surfaces, p. 424, Wiley. ISBN 0-471-81033-9.
  6. Clemence, Robert D., et al., 2007, “Verification, Validation and Accreditation,” (VV&A), Evidence Based Research, Vienna, Va.
  7. Correa, T., A. W. Housley and H.G. deZuniga, 2009, “Who interacts on the web?: The intersection of users personality and social media use,” Computers in Human Behavior, Vol. 26, No. 2, March 2010, pp. 247-253.
  8. Denson, T.F., C. Dobson-Stone, R. Ronay, W. von Hippel, and M. M. Schira, 2014, “A Functional Polymorphism of the MAOA Gene is Associated with Response to Induced Anger Control,” Journal of Cognitive Neuroscience, Vol. 26, No. 7, February 2014, pp. 1,418-27.
  9. Diaz-Morales, J. F., 2007, “Morning and evening-types: Exploring their personality styles,” Personality and Individual Differences, Vol. 43, No. 4, pp. 769-778. Science Direct.
  10. Entous, Adam, Greg White, and Peter Nicholas, 2013, “U.S. Decided Not to Horse-Trade With Russia on Assad,” Wall Street Journal, Sept. 5, 2013.
  11. Faro, S.H., F. B. Mohamed, L. Gordon, and S. M. Platek, 2013, “Detection of deception and truthtelling using FMRI,” US 857 1629 B2 Patent Application Grant.
  12. Fogg, B.J., 2002, “Persuasive Technology: Using Computers to Change What People Think (Interactive technologies),” Morgan Kaufman Publishers, Amsterdam.
  13. Fogg, B. J. and D. Eckles, 2014, “Mobile Persuasion,” Stanford Captology Media, Stanford, Calif.
  14. Frank, M.G. and E. Svetieva, 2012, “Lies worth catching involve both emotion and cognition,” Journal of Applied Research in Memory and Cognition, Vol. 1, No. 2, June 2012, pp. 131-133. Elsevier, Amsterdam.
  15. Gigerenzer, G. and P.M. Todd, 1999, “Simple Heuristics That Make Us Smart,” Oxford University Press, Oxford.
  16. Hartley, Dean S. III and Lee W. Lacy, 2013, “Irregular Warfare (IW) Ontology Final Report,” TRADOC Analysis Center, Fort Leavenworth, Kan.
  17. Hurley, C. M. and M. G. Frank, 2011, “Executing Facial Control During Deception Situations,” Nonverbal Behavior, Vol. 35, pp. 119-131.
  18. Immelman, Aubrey, 2014, “The Personality Profile of Russian President Vladimir Vladimirovich Putin,” The Immelman Turn, http://www.immelman.us/news/russia-threat-assessment-psychological-profile-of-vladimir-putin/, March 27, 2014, modified April 25, 2014 and July 30, 2014. Accessed Aug. 19, 2014.
  19. Kuriyama, K., M. Honma, S. Koyama, and Y. Kim, 2011, “D-Cycloserine facilitates procedural learning but not declarative learning in humans: a randomized controlled trial of the effect of D-cycloserine and valproic acid on overnight properties of performance of non-emotional memory tasks,” Neurobiology of Learning and Memory, Vol. 95, No. 4, pp. 505-509.
  20. Larson, C., 2014, “Genome Editing,” MIT Technology Review, May/June 2014, pp. 26-29, Boston.
  21. Maddux, W. W., E. G. Mullen, and A. D. Galinsky, 2008, “Chameleons Bake Bigger Pies and Take Bigger Pieces: Strategic Behavioral Mimicry Facilitates Negotiation Outcomes,” Journal of Experimental Social Psychology, Vol. 44, No. 2, pp. 461-468.
  22. Matsumoto, D., H. S. Hwang, L. Skinner, and M. G. Frank, 2011, FBI Law Enforcement Bulletin: “Detecting Deception.”
  23. Matsumoto, D., M. G. Fran and H. S. Hwang, 2013, “Nonverbal Communication,” Science and Application, Sage Publications, Los Angeles.
  24. McDonald, J., L. Schleifer, J. B. Richards, and H. DeWitt, 2008, “Effects of THC on behavioral measures of impulsivity in humans,” Neuropsychopharmacology, Vol. 28, No. 7, July 2003, pp. 1,356-1,365.
  25. Moffitt, T.E., 2005, “Genetic and Environmental Influences on Antisocial Behaviors: Evidence from Behavioral-Genetic Research,” Advances in Genetics, Vol. 55, p. 41-104.
  26. Mohammed, A. D. and B. J. Sahakian, 2011, “The Ethics of Elective psychopharmacology,” International Journal of Neuropsychopharmacology, Vol. 15, No. 4, pp. 559-571.
  27. Muchnik, L., S. Aval and S. J. Taylor, 2013, August 2013, “Social Influence Bias: A Randomized Experiment,” Science, Aug 2013, pp. 647-651.
  28. Owen, D. and J. Davidson, 2009, “Hubris Syndrome; an acquired personality disorder? A study of U.S. Presidents and U.K. Prime Ministers over the last 100 Years,” Brain, Feb. 12, 2009, Oxford University Press.
  29. Pentland, A., 2008, “Honest Signals, How They Shape Our World,” MIT Press, Boston.
  30. Pentland, A., 2014, “Social Physics, How Good Ideas Spread: The Lessons From A New Science,” The Penguin Press, New York.
  31. Post, Jerrold M., 2004, “Leaders and their Followers in a Dangerous World, the psychology of political behavior,” Cornell University Press, Ithaca.
  32. Post, Jerrold M., 2006, “The Psychological Assessment of Political Leaders,” The University of Michigan Press, Ann Arbor.
  33. Sahakian, Barbara and Sharon Morein-Zamir, 2007, “Professor’s little helper,” Nature, Vol. 450, 20/27, December 2007, pp. 1,157-1,159.
  34. Saltelli, Andrea and Silvio Funtowicz, 2014, “When All Models Are Wrong,” Issues in Science and Technology, Winter 2014, pp. 79-85.
  35. Seife, Charles, 2014, “Virtual Unreality,” Viking Penguin Press, New York.
  36. Tetlock, Philip, 2014, “The Good Judgment Project,” http://goodjudgmentproject.com/. Accessed Oct. 22, 2014.
  37. Turner, D. C., T. W. Robbins, L. Clark, A. R. Avon, J. Dawson, and B. J. Sahakian, 2003, “Cognitive Enhancing Effect of Modafinil in Healthy Volunteers,” Psychopharmacology, Vol. 165, No. 3, pp. 260-269.
  38. Vinciarelli, A., M. Pantic and H. Bourland, 2009, “Social signal processing: Survey of an emerging domain,” Image and Vision Computing, Vol. 27 No. 12, November 2009, pp. 1,743-1,759.
  39. Wikipedia, 2014, “Neolithic Revolution,” http://en.wikipedia.org/wiki/Neolithic_Revolution. Accessed Oct. 3, 2014.
  40. Wikipedia, 2014, “Pythia,” http://en.wikipedia.org/wiki/Pythia. Accessed Oct. 3, 2014.