Abstract
The Dark Factor of Personality (D) is a unifying theoretical model proposing that nine socially and ethically aversive personality traits — egoism, Machiavellianism, moral disengagement, narcissism, psychological entitlement, psychopathy, sadism, self-interest and spitefulness — are specific, flavoured manifestations of a single general disposition: the tendency to maximise one's own utility while disregarding, accepting, or malevolently provoking disutility for others, sustained by self-serving justifying beliefs (Moshagen, Hilbig & Zettler, 2018).
This article provides a critical appraisal of D rather than a restatement of its popular coverage. It traces the theoretical model to its founding publication and to the original measurement instruments for each of the nine component traits, establishing precisely what evidence the unifying claim rests on. It examines the psychometric case for D — the D70, D35 and D16 item sets derived from rational item-selection procedures across more than 165,000 respondents (Moshagen, Zettler & Hilbig, 2020) — and the four-year longitudinal study demonstrating D's rank-order stability exceeds that of any single dark trait it subsumes (Zettler, Moshagen & Hilbig, 2021).
The article then examines two live academic debates bearing directly on the model's validity: whether D is empirically distinguishable from low HEXACO Honesty-Humility and low Big Five Agreeableness (Horsten, Moshagen, Zettler & Hilbig, 2021; Moshagen, Zettler, Horsten & Hilbig, 2020), and whether the "Dark Triad" literature D partly absorbs has itself been shown to be substantially redundant with psychopathy alone (Glenn & Sellbom, 2015).
Finally, the article translates the evidence into workplace application, drawing on a newly published cross-national study linking D to vocational interests via both self-report and official occupational registry data (de Hesselle et al., 2026), and closes with an evidence-based caution against using D-factor assessments for candidate screening or selection — a use the measure's own developers did not design or validate it for.
Keywords: Dark Factor of Personality; D-factor; dark traits; dark core; Dark Triad; Dark Tetrad; Honesty-Humility; construct redundancy; organisational psychology; workplace personality
PART I — ONE PATTERN, NINE NAMES
Chapter 1: The Diagnostic Puzzle Behind "Difficult" Colleagues
A team leader who takes credit for a direct report's idea in a steering meeting is usually filed under narcissism. A colleague who agrees to a plan in the room and quietly undermines it afterwards is usually classified as Machiavellian. A manager who seems entirely unmoved by the distress their decisions cause is usually filed under psychopathy. A peer who goes out of their way to make someone else's failure a little worse, even when it costs them something, gets filed under spite. Workplace psychology has spent the better part of three decades building separate vocabularies for each of these patterns, and each vocabulary comes with its own questionnaire, its own case literature, and its own confident claim to have identified the mechanism.
The trouble is what happens when you actually measure these people. Study after study finds that narcissism, Machiavellianism and psychopathy scores correlate substantially with one another — high scorers on one tend to score at least moderately on the other two — and the same pattern extends, with somewhat lower but still meaningful overlap, to sadism, entitlement, spite and moral disengagement. The three colleagues in the illustrations above are not hypothetical composites drawn for effect; they are the kind of profile that recurs constantly in organisational case material, and they overlap far more often than the separate-trait framing would predict. The credit-taker in the steering meeting is often also the person who agreed to the plan and undermined it. The unmoved manager is often also the one who seems to enjoy a subordinate's discomfort just a little more than the situation calls for.
Two broad explanations are available for this overlap, and the distinction between them is not academic — it changes what you should actually watch for in a colleague, and what you should conclude once you've spotted it. The first explanation is that these are genuinely separate traits which merely happen to co-occur in the same people at above-chance rates, for reasons that might range from shared developmental origins to simple selection effects — difficult environments attracting or rewarding several dark tendencies at once. On this view, a full workplace risk assessment requires checking for narcissism, then separately for Machiavellianism, then separately for psychopathy, and so on, because each one might in principle appear without the others, and each represents a genuinely distinct hazard.
The second explanation is more radical. It proposes that co-occurrence is not a coincidence to be explained away but the central fact requiring explanation, and that a large part of what looks like nine separate traits is, in fact, one disposition, expressed with different emphasis depending on which specific behaviours and beliefs a person has settled into. On this view, asking "is this person more narcissistic or more Machiavellian" can be close to asking whether a river is wetter on the left bank or the right — a question that assumes a distinction that the underlying reality may not support to the degree the separate labels imply.
This is the question a trio of German and Danish personality psychologists set out to resolve empirically rather than philosophically. Their answer, published in 2018, gave the shared core a name — D, the Dark Factor of Personality — and a formal, testable definition. What follows traces that model from its origin to its current, contested standing: what the founding evidence actually showed, what each of the nine traits it drew on individually measures, how the model's own authors have since tried to measure D directly rather than infer it from overlap, and where independent researchers have pushed back hardest.
Why this matters beyond academic tidiness is worth stating plainly before the evidence begins. A manager who believes they are dealing with "a narcissist" and a manager who believes they are dealing with "someone high in a general antagonistic disposition that happens to be expressing itself through self-promotion and rivalry right now" will, in practice, watch for different things, document different things, and reach for different remedies. The first framing invites a search for a fixed personality type, applied from a distance and rarely revisited once assigned. The second invites attention to a pattern of behaviour that can, at least in principle, be documented, discussed and — as Chapter 13 will show — tends to soften somewhat with age, even where it does not disappear. Which framing is better supported by the evidence is not merely of academic interest to an audience that manages people for a living.
Chapter 2: The Dark Core Hypothesis: Moshagen, Hilbig and Zettler's 2018 Model
Morten Moshagen (Ulm University), Benjamin Hilbig (then University of Koblenz-Landau, now RPTU Kaiserslautern-Landau) and Ingo Zettler (University of Copenhagen) published "The Dark Core of Personality" in Psychological Review in October 2018, one of psychology's most selective and theory-focused journals (Moshagen, Hilbig & Zettler, 2018). Their proposal was that the various traits gathered under the informal umbrella term "dark traits" — a set that had grown steadily since Delroy Paulhus and Kevin Williams coined "the Dark Triad" for narcissism, Machiavellianism and psychopathy in 2002 — are not independent constructs that happen to correlate, but flavoured manifestations of a single underlying disposition. In an analogy the authors use deliberately, D is to dark personality what the general intelligence factor, g, is to cognitive ability: not a directly observable trait in its own right, but the common thread running through a family of more specific, more visible ones.
The formal definition is precise enough to be tested rather than merely asserted. D is the general tendency to maximise one's own utility — disregarding, accepting, or malevolently provoking disutility for others — accompanied by beliefs that serve as justifications (Moshagen, Hilbig & Zettler, 2018). Three separate components sit inside that sentence, and each does real theoretical work.
The first is utility maximisation. "Utility" here is deliberately broad: it covers material gains such as money or status, but it also covers purely psychological gains — a feeling of power, superiority, excitement, or straightforward pleasure. A colleague who manipulates a project timeline to make a rival look incompetent is maximising utility just as surely as one who pads an expense claim; the currency differs, the underlying mechanism does not.
The second is disutility for others, and this is where the model earns its explanatory range. The theory proposes that high-D individuals sit at different points along a spectrum of how they relate to the harm they cause. At one end, harm is simply disregarded — the person does not register, or does not weight, the cost their actions impose on someone else. Psychopathy, with its documented deficits in affective empathy, sits closer to this end. Further along, harm is accepted as a foreseen, tolerated cost of getting what you want — a calculated trade-off rather than an oversight. Machiavellianism, built around cold strategic calculation, sits closer to this position. At the far end, harm stops being a cost to be minimised and becomes part of the utility itself — the person derives satisfaction from the other person's disutility, not merely alongside it. Sadism sits here. The same underlying disposition, differently expressed, produces three recognisably different colleagues.
The third component is justifying beliefs — the cognitive scaffolding that lets a person behave this way while maintaining a broadly positive self-image. The authors describe this scaffolding as flexible rather than fixed: it can take the form of viewing the world as a competitive jungle where ruthlessness is simply realism, seeing oneself or one's in-group as superior and therefore differently entitled, or regarding other people as sufficiently foolish or culpable that exploiting them barely counts as wrongdoing. Critically, the theory does not claim that every high-D person holds the same belief. It claims only that they hold some belief serving this justifying function — which is one reason two high-D colleagues can present very differently while sharing the same underlying disposition.
The 2018 paper tested this framework against nine established dark traits with independent, well-validated measurement histories: egoism, Machiavellianism, moral disengagement, narcissism, psychological entitlement, psychopathy, sadism, self-interest, and spitefulness. Using bifactor modelling — a statistical technique that separates a general factor common to all indicators from the specific factors unique to each trait — the authors found consistent support for a single dominant general factor running through all nine indicators, alongside much smaller trait-specific factors capturing the residual variance once the shared variance was accounted for. The model's practical bite came from what that general factor predicted: across the studies, D was substantially related to aggression, dominance, a lack of empathy, and a diminished sense of moral identity, and the nine individual traits rarely explained additional variance in these outcomes once D itself was accounted for. In plain terms, once you know someone's D score, knowing their specific narcissism or psychopathy score on top of it adds comparatively little further predictive information about how they actually behave.
That last finding is the crux of the whole model, and it is also the finding critics have returned to most often, as later chapters in this article examine directly. Before assessing the criticism fairly, however, it is worth understanding exactly what the nine component traits are — because the strength of a unifying claim depends entirely on the quality and independence of the things being unified, and the traits themselves are worth examining on their own terms.
PART II — THE NINE FACES OF D
Each of the nine traits below entered the D-factor model with its own independent research history and its own validated measurement instrument, predating the unifying theory by anywhere from a few years to five decades. Understanding each on its own terms is what allows the unifying claim in Chapters 1 and 2 to be evaluated on its merits rather than taken on trust. Two of these traits — psychopathy and everyday sadism — are covered in more evidential depth elsewhere in this series (respectively, this series' articles on the Dark Empath profile and on the bad apple effect); they are given full treatment here rather than a brief cross-reference, because the D-factor model asks a different question of them than those earlier treatments did, namely how much of what makes each trait "dark" is shared rather than unique.
Chapter 3: Egoism — Self-Interest at the Community's Expense
Egoism is defined in the personality literature as an excessive concern with one's own pleasure or advantage, pursued at the expense of the community's well-being (Weigel, Hessing & Elffers, 1999). The construct originates not in clinical or organisational psychology but in criminological research into tax evasion and everyday rule-breaking: Russell Weigel, Dick Hessing and Henk Elffers were interested in why some people are willing to free-ride on collective goods — public services, shared infrastructure, mutual trust — that they benefit from without being willing to fund or protect. Their 20-item Egoism Scale, developed for that purpose, measures the general disposition rather than any single act of rule-breaking.
What makes egoism theoretically distinct from simple selfishness, and therefore a useful component of the D-factor model, is its explicit relational structure: the definition does not merely describe wanting things for oneself, but wanting them at a cost to a collective one is embedded in and benefits from. In an organisational setting, this is the colleague who treats shared resources — a team's goodwill, a department's budget flexibility, a manager's trust — as an unlimited personal account, without the Machiavellian's specific manipulative strategy or the narcissist's grandiosity. It is a quieter, more diffuse pattern: a consistent willingness to draw down collective goods without corresponding investment in replenishing them, present in behaviours as mundane as chronic under-contribution to shared administrative work and as consequential as treating a team's political capital with senior leadership as an asset to be spent solely on personal advancement.
Egoism was one of the two traits Moshagen, Hilbig and Zettler's original modelling identified as loading exceptionally strongly onto the general D factor once measured through the finalised item sets discussed in Chapter 12, alongside psychopathy — meaning that, empirically, egoism items carry relatively little variance that is not already captured by D itself. In practice, this makes egoism one of the traits for which the unifying claim is best supported: comparatively little about egoism, once measured properly, seems to do independent work beyond the general disposition it shares with the other eight.
Chapter 4: Machiavellianism — The Cynical Strategist
Machiavellianism is the oldest of the nine component constructs by a wide margin, tracing back to Richard Christie and Florence Geis's 1970 book Studies in Machiavellianism, itself named for the sixteenth-century political theorist whose counsel to rulers assumed that appearing virtuous mattered more than being virtuous. The trait is defined as a duplicitous interpersonal style, built on a broader network of cynical beliefs about human nature and a pragmatic disregard for conventional morality (Jones & Paulhus, 2014). Where psychopathy centres on a deficit — reduced affective response to others' distress — Machiavellianism centres on a strategy: a considered, long-horizon calculation about when honesty, loyalty and fair dealing are worth the cost and when they are not.
The construct has been measured across several instruments with slightly different emphases: the original Mach-IV, Christie and Geis's 20-item scale probing cynical worldview and manipulative tactics; the Machiavellianism subscale of Daniel Jones and Delroy Paulhus's Short Dark Triad, which foregrounds strategic and controlled manipulation specifically; and the Machiavellian Personality Scale developed by Jason Dahling, Brian Whitaker and Paul Levy, which splits the construct into distinct amorality and desire-for-control components to address limitations identified in the older Mach-IV.
In the workplace, high Machiavellianism is the profile of the colleague who agrees enthusiastically in the room and works against the plan the moment it leaves the room — not from confusion or a genuine change of mind, but as a calculated position taken in advance. Distinct from the psychopathic profile's more blunted register of harm, the Machiavellian typically registers the cost their actions impose on others accurately; they simply weigh it against their own goals and, often, conclude it is an acceptable price. This is the harm-is-accepted position described in Chapter 2's account of D's second component, and it is why Machiavellianism reads, on the page, as more calculating and less impulsive than psychopathy, even though both draw heavily on the same underlying disposition.
The specific organisational cost of high Machiavellianism is less the individual harmful act — most such acts, taken in isolation, are small and deniable — than the erosion of trust that accumulates once a pattern becomes visible to others. A team that has learned, even informally, that a particular colleague's public agreement predicts nothing about their private conduct pays a continuous tax on every subsequent negotiation: proposals get relitigated privately after they should have been settled, coalitions form defensively rather than productively, and the ordinary efficiency of simply taking a colleague's word for something disappears, often well before anyone can point to a single incident serious enough to raise formally.
Chapter 5: Narcissism — Grandiosity Aimed at Others
Narcissism is characterised by an exaggerated sense of self-importance, a persistent need for admiration, and — in the specific slice of the construct most relevant to D — a willingness to exploit others in the service of that self-image (Back et al., 2013). This last qualifier matters, because narcissism research has increasingly distinguished between different expressions of the trait, and not all of them carry the same relevance to workplace harm.
Contemporary narcissism research typically separates an agentic dimension — confidence, self-promotion, charm, the tendency to seek out leadership and visibility — from an antagonistic dimension, sometimes called narcissistic rivalry, which involves devaluing others to protect and elevate the self, treating colleagues' successes as threats and their failures as opportunities. Moshagen, Hilbig and Zettler's model draws specifically on this antagonistic dimension: their measurement approach uses the Narcissistic Rivalry subscale of Mitja Back and colleagues' Narcissistic Admiration and Rivalry Questionnaire, alongside grandiosity items from the Short Dark Triad and leadership, authority and exploitativeness items from Robert Raskin and Calvin Hall's Narcissistic Personality Inventory. The agentic, purely self-promotional side of narcissism — the side most compatible with confident, effective leadership and least associated with interpersonal harm — is deliberately underweighted in the D-factor version of the construct, because it is the antagonistic side that overlaps most strongly with the shared disposition the model is trying to isolate.
This distinction has real practical value for anyone trying to read a colleague accurately. Self-promotion, visibility-seeking and even a degree of self-importance are not, on their own, strong evidence of the pattern this article is about. What distinguishes the D-relevant expression of narcissism is what happens when someone else's success threatens the narcissist's own standing: a shift from simple self-promotion into active devaluation of the rival, often disguised as constructive criticism or concern.
The tell, in practice, is rarely the self-promotion itself but its selectivity: enthusiastic credit-claiming for shared successes paired with a conspicuous absence when a shared project runs into difficulty, and a pattern of praise for subordinates that tracks how much reflected credit the praise generates for the person giving it, rather than tracking the subordinate's actual contribution.
Chapter 6: Psychopathy — Callousness Without the Criminal Record
Psychopathy, in its sub-clinical, workplace-relevant form, is characterised by impulsive hostility: deficits in affective response and in self-control, expressed through callousness, a striking absence of guilt or remorse, and reduced sensitivity to punishment (Levenson, Kiehl & Fitzpatrick, 1995). This article's companion piece on the bad apple effect covers the workplace evidence for corporate psychopathy in more depth, including prevalence estimates and the primary/secondary distinction; the treatment here focuses specifically on psychopathy's role within the D-factor model rather than repeating that ground.
Within the nine-trait framework, psychopathy is measured through items drawn from the Levenson Self-Report Psychopathy scale — covering egocentricity, callousness, and antisocial tendencies — combined with the psychopathy subscale of the Short Dark Triad. Notably, the model's authors deliberately excluded items describing psychopathy's disinhibition and impulsivity facets, such as items about avoiding dangerous situations or a general lack of planfulness, on the grounds that these facets, while genuinely part of clinical conceptions of psychopathy, lie outside the theoretical scope of D itself: a lack of impulse control is not, on its own, malevolent, even though it may make existing malevolence more visible in its consequences.
What emerges from this deliberately narrowed measurement is psychopathy in its purest D-relevant form: not the reckless, poorly self-controlled stereotype of popular imagination, but callousness — an accurate registration of harm that simply fails to generate the emotional weight that would normally restrain it. Of the nine traits in the model, psychopathy loads most heavily on the general D factor: in the founding studies, nearly all of the common variance among psychopathy items was explained by D itself, leaving little that was uniquely "psychopathy" once the shared disposition was accounted for. This finding is worth holding onto, because it becomes the central piece of evidence in the construct-redundancy debate examined in Chapter 15.
Chapter 7: Everyday Sadism — Pleasure in Others' Discomfort
Everyday sadism is the tendency to engage in cruel, demeaning or aggressive behaviour specifically to derive pleasure from it, distinguishing it from behaviour that merely happens to cause harm as a side effect of pursuing something else. This series' article on the bad apple effect covers sadism's emergence as the "fourth" trait, extending the Dark Triad into the Dark Tetrad, following Erin Buckels, Daniel Jones, and Delroy Paulhus's 2013 laboratory demonstrations that sadism predicts cruelty even after accounting for psychopathy, narcissism, and Machiavellianism scores.
Within the D-factor model specifically, sadism was measured using items from Rebecca Plouffe, Donald Saklofske, and Martin Smith's Assessment of Sadistic Personality Inventory, combined with items from Abigail O'Meara, Jayne Davies, and Sheila Hammond's Short Sadistic Impulse Scale. Sadism occupies a distinctive position within the model's own theoretical architecture, described in Chapter 2: where psychopathy sits at the "disregard" end of D's harm spectrum and Machiavellianism sits at the "accept" end, sadism sits at the far end, where inflicting disutility on another person stops being a cost of pursuing one's goals and becomes part of the goal itself. This is the theoretical reason sadism, despite being the most recently formalised of the nine traits, integrates so cleanly into the model: it fills a position on the spectrum the other eight traits leave comparatively empty.
In workplace terms, everyday sadism rarely looks like anything so obvious as cruelty for its own sake; it is usually smuggled inside behaviour that has a defensible surface justification. The manager who lingers a beat too long on a colleague's mistake in a public meeting, past the point the correction itself required, or the peer who volunteers to deliver difficult feedback with unusual enthusiasm, is not necessarily doing anything a bystander could object to on procedural grounds. What distinguishes the pattern is the surplus — the visible extra satisfaction that exceeds what the situation itself calls for, present in tone and timing rather than in the content of what was said.
Chapter 8: Psychological Entitlement — Deserving More Than Others
Psychological entitlement is a stable, pervasive belief that one deserves more, and is entitled to more, than other people — not tied to any specific achievement or claim, but operating as a general expectation applied across situations (Campbell, Bonacci, Shelton, Exline & Bushman, 2004). W. Keith Campbell and colleagues developed the 9-item Psychological Entitlement Scale to capture this as a stable individual difference rather than a situational reaction, validating it across nine separate studies and showing it correlated negatively with two Big Five traits in particular — agreeableness and emotional stability — while remaining unrelated to socially desirable responding, meaning high scorers were not simply concealing the trait when given the chance.
The version of entitlement built into the D-factor model adds three further items from a large, 28-country entitlement measure developed by Magdalena Żemojtel-Piotrowska and colleagues, specifically the items addressing revenge entitlement — the belief that one is entitled not merely to more, but to retaliation when a perceived slight occurs. This addition matters for the workplace reading of the trait: pure entitlement without a revenge component might produce a colleague who simply expects disproportionate recognition; entitlement with a revenge component produces someone who treats being overlooked, criticised or merely treated as an equal as a grievance requiring redress, often disguised as a fairness complaint or an appeal to being undervalued.
Entitlement's distinguishing feature within the model is that it supplies much of D's third component directly — the justifying belief itself. Where sadism supplies motive and psychopathy supplies the absent brake, entitlement supplies the internal narrative that makes disproportionate behaviour feel, to the person doing it, like simple fairness.
The revenge-entitlement addition described above is what most reliably separates an ordinarily confident, perhaps even irritatingly self-assured colleague from one whose entitlement is doing genuine organisational damage. Confidence without the revenge component might produce someone who over-negotiates their own compensation; confidence with it produces someone who treats a manager's routine, evenly applied decision — a promotion given to someone else, a project reassigned — as a personal affront requiring a response, often pursued through channels (a complaint, a quiet campaign against the beneficiary, a withdrawal of previously freely given cooperation) disproportionate to the actual decision.
Chapter 9: Spitefulness — Harm at a Cost to Yourself
Spitefulness is the willingness to harm another person even when doing so also carries a cost to oneself (Marcus, Zeigler-Hill, Mercer & Norris, 2014). David Marcus and colleagues developed the first dedicated self-report measure of the trait, the 17-item Spitefulness Scale, having noted that spite — despite being instantly recognisable in everyday life, from the driver who brakes sharply to punish tailgating to the colleague who blocks a proposal they privately agree with simply because a rival proposed it — had received almost no dedicated attention in the personality literature up to that point. Their validation studies, run across large student and adult samples, found spitefulness positively associated with aggression, psychopathy, Machiavellianism and narcissism, and negatively associated with self-esteem, guilt-proneness, agreeableness and conscientiousness — a correlational profile that placed it squarely within the same territory as the other dark traits years before the unifying D-factor model existed to explain why.
Spitefulness is the clearest embodiment of D's theoretical claim that utility maximisation and harming others are not always separate acts. Every other trait examined so far involves harm as either an oversight (psychopathy), a calculated trade-off (Machiavellianism), or a source of pleasure that costs the perpetrator nothing (sadism, in its purest form). Spitefulness is different: it is defined specifically by the perpetrator's willingness to accept a personal cost purely to ensure someone else bears one too, which is difficult to explain under a simple rational self-interest model of behaviour and correspondingly easy to explain under a model, like D, that treats inflicting disutility on others as capable of generating its own utility.
This is what makes spitefulness one of the more organisationally expensive traits on the list, despite its comparatively low public profile relative to narcissism or psychopathy: a purely self-interested actor can, in principle, be managed through incentives, because their behaviour tracks their own advantage. A spiteful actor's behaviour does not reliably track their own advantage at all, which is precisely what makes it harder to predict or deter through the usual organisational levers — the colleague who withholds a piece of information that would have cost them nothing to share, purely because sharing it would help someone they resent, is not responding to an incentive structure in any straightforward sense.
Chapter 10: Moral Disengagement — The Permission Structure
Moral disengagement is not itself a form of harmful action but a set of cognitive processing styles that make harmful action feel permissible — mechanisms such as dehumanising a target, diffusing or displacing responsibility for an outcome, minimising the harm caused, or reframing a harmful act in more palatable, even virtuous-sounding, terms (Bandura, 1999). Albert Bandura's original theoretical account described these as the specific psychological manoeuvres by which ordinarily conscientious people become capable of cruelty without experiencing the self-censure that would normally follow. Christopher Moore, James Detert, Linda Treviño, Vicky Baker and David Mayer operationalised the construct for organisational research with their 16-item Propensity to Morally Disengage Scale, published in Personnel Psychology specifically to study why otherwise capable employees participate in unethical organisational behaviour.
Moral disengagement's role in the D-factor model is structurally distinct from the other eight traits, and understanding that difference clarifies what D's "justifying beliefs" component, introduced in Chapter 2, actually describes. Where egoism, entitlement and spitefulness are dispositions toward specific kinds of harmful motivation, moral disengagement is closer to a permission structure that can be applied to almost any of them: it is the mental move that allows the egoist to reframe free-riding as "everyone does it," the entitled colleague to reframe disproportionate reward-seeking as simple fairness, and the Machiavellian to reframe calculated harm as an unfortunate but necessary business decision. Its presence in the nine-trait model is a large part of why D's definition explicitly includes justifying beliefs as a necessary third component alongside utility maximisation and disregard for others — without some form of moral disengagement, sustained dark behaviour would generate a level of guilt most people cannot tolerate indefinitely.
This is also why moral disengagement is one of the more organisationally revealing traits to listen for directly, because unlike most of the others it shows up in language rather than only in outcomes. Moore and colleagues' organisational research found the pattern especially visible in how people account for unethical acts after the fact: diffusion of responsibility ("it wasn't just me, everyone signed off on it"), euphemistic relabelling ("aggressive," not "misleading"), and advantageous comparison ("it's nothing compared to what happens elsewhere in the industry") recur across very different organisational contexts as the standard vocabulary of self-permission.
Chapter 11: Self-Interest — The Trait That Almost Didn't Make the Cut
Self-interest is defined as the motivation to pursue personal gains in socially valued domains — material goods, social status, recognition, academic or occupational achievement, and general happiness (Gerbasi & Prentice, 2013). Melissa Gerbasi and Deborah Prentice developed the Self- and Other-Interest Inventory to measure this alongside a parallel other-interest dimension, treating the two not as opposite poles of a single scale but as largely independent orientations — a person can be high in both, low in both, or any combination, because wanting good things for yourself and wanting good things for others turn out, empirically, not to be well described as trade-offs against each other.
Self-interest earns a chapter of its own in this article not because it is the most dramatic of the nine traits, but because of an unusually candid piece of scholarly housekeeping in the model's own development. When Moshagen, Zettler and Hilbig built the concise D70, D35 and D16 measurement instruments discussed in Chapter 12, they explicitly excluded self-interest from the item pool — the only one of the original nine traits dropped at that stage. Their stated reason, given in a footnote to the measurement paper, is worth taking seriously precisely because it is a self-correction rather than a defence: self-interest, they noted, describes a form of utility maximisation that rarely implies or requires disutility for others, and measures of it often correlate positively with the parallel other-interest dimension rather than trading off against it. In plain terms, wanting career success, recognition or happiness is, for most people most of the time, not something pursued at anyone else's expense — and a trait that doesn't reliably involve harming others sits awkwardly inside a model explicitly defined by the disregard, acceptance, or provocation of harm to others.
This is a genuinely important qualification for anyone applying this material practically, and it is one that popular coverage of the D-factor model frequently glosses over. Self-interest remains part of the theoretical nine-trait taxonomy set out in the model's founding 2018 paper, and, in principle, it can shade into egoism or entitlement when it crosses into harming others. But as an operational matter, the model's own authors do not treat raw ambition, career-mindedness, or a straightforward desire to succeed as evidence of D — a distinction worth holding on to before reflexively filing an ambitious colleague under "dark personality" simply because they are unapologetically pursuing their own advancement.
PART III — MEASURING AND TESTING THE MODEL
Chapter 12: How Do You Measure a Common Core? The D70, D35 and D16
The 2018 founding paper established D theoretically and demonstrated it statistically, using bifactor models fitted to 93 items drawn from measures of the nine component traits. What it did not provide was a practical instrument: a researcher, or a curious individual, wanting to know their own level of D had no direct way to find out short of completing all nine separate trait questionnaires and reconstructing the shared factor after the fact. Moshagen, Zettler, and Hilbig addressed this directly in a 2020 paper in Psychological Assessment, "Measuring the Dark Core of Personality," which served as the methodological backbone of everything that followed, including the popular online D-score test hosted on the model's own research site (Moshagen, Zettler & Hilbig, 2020).
The scale of the underlying research is worth stating plainly, because it is unusually large even by the standards of personality psychology. The authors drew on seven separate samples totalling more than 165,000 participants, recruited partly through international media coverage of the original 2018 paper and partly through professionally managed research panels, spanning more than 150 countries. From an initial pool of 184 candidate items drawn from twelve established dark-trait measures — the original nine plus three additional constructs considered for inclusion — the authors used a genetic-algorithm-based item-selection procedure to identify progressively shorter sets of items that retained the strongest possible measurement of the shared D factor while dropping content that was specific to individual traits rather than the common core.
Three item sets emerged from this process. The D70 comprises 70 items, evenly balanced between positively and negatively worded statements to guard against simple acquiescence bias — the tendency of some respondents to agree with statements regardless of content. The D35 and D16 are 35- and 16-item distillations of the same underlying construct, built using the same selection logic applied recursively to the longer set, intended for contexts where a full 70-item assessment is impractical. All three showed strong internal consistency in the validation samples, and all three correlated highly with one another — above .95 in every case — indicating that even the briefest 16-item version captures most of what the full instrument measures.
The validation evidence goes beyond internal consistency. The D70 was tested against three genuinely behavioural outcomes, not just further self-report questionnaires: a probabilistic cheating task in which participants could dishonestly report a favourable coin-toss outcome to avoid a tedious follow-up task; an incentivised allocation measure of prosocial versus self-interested choice, paid out in real money; and a purpose-built behavioural measure of sadistic tendencies, again with real financial stakes. D correlated meaningfully with all three — with dishonest behaviour in the cheating task, with a preference for self-interested over joint-maximising allocations, and with the willingness to sacrifice one's own payoff purely to reduce another participant's. This matters because self-report personality measures are frequently, and fairly, criticised for only ever predicting other self-reports; a measure that also predicts what people actually do with real money on the line, without knowing they are being watched for exactly this pattern, is making a stronger claim to validity.
One further, technically important feature of the model deserves mention because it explains why the specific item content of any one measure matters less than critics sometimes assume. D is explicitly conceptualised as what personality researchers, following the early-twentieth-century psychometrician Charles Spearman, call a fluid construct: one that does not depend critically on any particular set of indicator items, provided that a sufficiently broad and diverse set of dark-trait indicators is included. Later work by Luisa Horsten, Morten Moshagen and Benjamin Hilbig tested this directly by randomly removing between 50% and 75% of the items from the full item pool and recomputing D scores from what remained; the resulting reduced-item scores still correlated at .90 or above with scores from the complete item set (Horsten, Moshagen & Hilbig, 2023). In practice, this means the model is making a claim about a genuine underlying statistical regularity in how these traits cluster together across people, not merely about the specific wording of any one questionnaire — a meaningfully stronger claim than critics sometimes credit it with.
Chapter 13: Is D Stable? The Four-Year Test
A single-timepoint correlation between traits proves that they overlap; it does not prove that the shared factor behind them is a genuine, temporally stable disposition rather than a snapshot artefact of how people happen to be feeling, or behaving, at the moment they completed a questionnaire. The strongest single piece of evidence addressing that gap is Ingo Zettler, Morten Moshagen and Benjamin Hilbig's 2021 longitudinal study in Social Psychological and Personality Science, which followed 1,261 adults from the general population — with 470 of the original sample retained at a second measurement occasion four years later — using full-information maximum-likelihood estimation to handle the resulting attrition appropriately rather than simply discarding incomplete cases (Zettler, Moshagen & Hilbig, 2021).
The study tested three specific predictions the D-factor model makes about how a genuine basic disposition, as opposed to a measurement artefact, ought to behave over time. First, D itself should show higher rank-order stability over the four-year gap than any of the nine individual dark traits it subsumes — because a basic disposition, on this account, is more fundamental and therefore more resistant to change than any one of its more specific expressions. This prediction was confirmed: D's rank-order stability exceeded that of every individual trait measured at both timepoints. Second, D measured at the first timepoint should predict individual differences in each of the nine specific traits measured four years later, not merely predict itself — a test of whether the shared factor is doing genuine causal or structural work rather than just describing a correlation that happens to hold at any one moment. This too was confirmed: D longitudinally predicted later individual differences in all nine traits, in several cases nearly as strongly as each trait predicted its own later self. Third, if a person's specific dark traits changed measurably over the four years — becoming more or less narcissistic, more or less Machiavellian — that change should itself be traceable back to a corresponding change in their underlying D score, rather than occurring independently of it. The data supported this as well.
The study also examined mean-level change across the sample as a whole, independent of individual rank-order stability, and found a pattern consistent with what personality psychologists call the maturity principle — a broad, well-replicated finding across many personality traits that most people become somewhat more socially adapted, less impulsive and less antagonistic as they age. D, and most of the individual dark traits alongside it, decreased with age across the sample, by roughly a third of a standard deviation over the range studied. This is a genuinely reassuring finding for anyone managing a workplace pattern consistent with high D in a younger employee: while the underlying disposition is stable relative to other traits at any given career stage, it is not fixed for life, and the population-level trend runs in the direction of moderation rather than escalation.
Taken together, the stability evidence is arguably the strongest single pillar supporting the D-factor model's central claim. Cross-sectional overlap between dark traits could, in principle, be explained by several less interesting mechanisms — shared method variance from similar-sounding questionnaire items, or a temporary situational state affecting several trait measures at once. A four-year prediction of trait-specific change, from a general factor measured years earlier, is much harder to explain away on those grounds. It does not settle every question about what D actually is — the debates in the next two chapters concern exactly that — but it substantially raises the burden of proof for anyone arguing the shared factor is merely a statistical convenience.
A related, more recent strand of evidence extends the stability question from the individual to the society in which a person lives and is worth including here because it is among the most current research this article draws on. Ingo Zettler, Lau Lilleholt, Martina Bader, Benjamin Hilbig and Morten Moshagen published a large cross-national study in the Proceedings of the National Academy of Sciences in 2025, examining whether average D levels vary systematically across countries and, within the United States, across individual states, and whether that variation tracks measurable societal conditions — inequality, corruption, weak governance — rather than appearing essentially random (Zettler, Lilleholt, Bader, Hilbig & Moshagen, 2025). The item sets underpinning this research have themselves now been translated and separately validated in more than thirty languages, from German and Portuguese to Chinese and Norwegian, which is itself a meaningful form of evidence: a construct that only holds together in the original English-language, largely Western samples it was built from would be a much weaker candidate for a genuine basic disposition than one that replicates its factor structure across linguistically and culturally distant populations. This article has not independently verified the specific societal-condition findings of the 2025 study beyond confirming its existence and publication venue, and readers with a specific interest in the cross-national evidence should consult the primary paper directly; it is flagged here as a live, current research direction rather than a settled finding this article is relying on for its own argument.
PART IV — CONTESTED GROUND
Chapter 14: Is D Just Low Honesty-Humility in a New Coat?
Long before Moshagen, Hilbig and Zettler proposed D, other personality researchers had already identified strong candidates for the "common core" of dark traits, using frameworks that predate the D-factor model by years. The most serious of these rival candidates is Honesty-Humility, the sixth factor in the HEXACO model of personality — an alternative to the more familiar Big Five, developed by Kibeom Lee and Michael Ashton, which adds a dimension that captures sincerity, fairness, greed-avoidance, and modesty. Several researchers had already proposed before 2018 that the common core of dark traits was essentially equivalent to low Honesty-Humility, and that the correlation between the two was genuinely substantial. If this equivalence claim is correct, D would be less a new discovery than a relabelling of an already-known trait — a serious challenge, because it would mean the entire nine-trait unifying apparatus, and the years of research built on it, add nothing that a much older, simpler measure did not already provide.
The D-factor research group has treated this challenge as serious enough to warrant dedicated empirical study rather than a rhetorical dismissal, which is itself worth noting: a genuinely weak unifying claim would be an easy target to abandon, and the authors have instead published multiple papers directly testing the overlap. Luisa Horsten, Morten Moshagen, Ingo Zettler and Benjamin Hilbig's 2021 paper in the Journal of Research in Personality tested the two constructs directly against each other and reported that D incrementally predicts a range of outcomes — including aggression, insensitivity, and a lack of nurturance — over and above what low Honesty-Humility predicts on its own (Horsten, Moshagen, Zettler & Hilbig, 2021). The theoretical case for why the two should be distinguishable, rather than simply asserting that they are, rests on three points. First, low Honesty-Humility characteristically involves disutility to others as an incidental by-product of pursuing material or status goals — the exploitation is a side effect of greed, not the point. D, by contrast, explicitly accommodates cases where inflicting harm is itself the source of utility, which is the mechanism sadism and spite depend on and which sits awkwardly inside a Honesty-Humility framing built primarily around dishonesty and unfairness. Second, the justifying beliefs D specifies are broader than the modesty-and-fairness violations central to low Honesty-Humility, extending to cynical worldviews and beliefs favouring group-based dominance that have no obvious home in the HEXACO framework. Third, D's conception of utility explicitly includes intangible psychological rewards — excitement, a sense of power — alongside the money and status that are more central to Honesty-Humility.
A parallel challenge comes from the Big Five's Agreeableness dimension, or more precisely its negative pole, sometimes labelled antagonism, which several researchers have proposed as an equally strong candidate for the shared dark core. Morten Moshagen, Ingo Zettler, Luisa Horsten and Benjamin Hilbig addressed this directly in a 2020 paper in the same journal, arguing that Agreeableness is a broad, primarily descriptive trait characterising general interpersonal style and cooperative tendency, whereas D is a narrower, more specifically motivational construct organised around utility maximisation at others' expense (Moshagen, Zettler, Horsten & Hilbig, 2020). This particular exchange has not gone unanswered: a subsequent published reply from critics Colin Vize and Donald Lynam pushed back on the distinctiveness claim, prompting a direct rejoinder from Hilbig, Moshagen, Horsten and Zettler defending the original position (Hilbig, Moshagen, Horsten & Zettler, 2021). This article has not independently reviewed the full technical substance of that exchange, and readers with a specific interest in the Agreeableness-overlap question should treat it as a live, published academic debate rather than a settled matter in either direction.
What can be said with reasonable confidence is this: D correlates substantially and negatively with both Honesty-Humility and Agreeableness, as any genuine measure of a dark core should, given both existing frameworks were themselves developed partly to capture antagonistic, uncooperative personality variance. The dispute is not about whether the correlation exists — it plainly does — but about whether, once that overlap is accounted for statistically, D is doing enough independent predictive work to justify treating it as a genuinely new construct rather than old wine in a new, more dramatically named bottle. The D-factor research programme has produced evidence that it does; the debate over how much independent work is "enough" to justify a new construct runs far more broadly through personality psychology than this specific case and is unlikely to be resolved by any single study.
For a reader without a stake in the theoretical dispute, the practical upshot is more reassuring than the academic back-and-forth might suggest. Whichever label ultimately proves most precise — D, low Honesty-Humility, low Agreeableness, or some later refinement combining insights from all three research traditions — the underlying behavioural territory these frameworks are all attempting to describe is the same one this article has been mapping throughout: a consistent disregard for the cost one's actions impose on others, sustained by beliefs that make that disregard feel justified. The theoretical argument over which framework owns that territory most precisely does not change what the territory itself looks like from the outside.
Chapter 15: The Construct-Proliferation Problem, and Whether Psychopathy Is Doing All the Work
A second and, in some respects, more damaging line of criticism does not challenge D directly at all. It challenges the raw material D was built from — specifically, whether the Dark Triad, the three-trait framework (narcissism, Machiavellianism, psychopathy) that supplied the historical starting point for the whole dark-personality research area, was ever genuinely three distinct constructs to begin with.
The most direct version of this challenge comes from Andrea Glenn and Martin Sellbom's 2015 paper in the Journal of Personality Disorders. Using a large correctional sample of 972 individuals, the authors attempted to fit a confirmatory factor analysis model in which narcissism, Machiavellianism, and psychopathy were treated as three indicators of a single latent Dark Triad factor — a structurally similar exercise, on a smaller scale, to what the D-factor model later attempted across nine traits. The model would not converge on a sensible solution: more than 100% of the variance in the proposed latent Dark Triad factor was attributable to psychopathy alone, a statistical result indicating that Machiavellianism and narcissism, once psychopathy was accounted for, were contributing essentially nothing distinct to the shared factor (Glenn & Sellbom, 2015). Their theoretical conclusion was blunt: rather than three siblings, Machiavellianism and narcissism might be better understood as features or facets of psychopathy specifically, not co-equal partners in a genuine triad. A separate, independently conducted meta-analysis by Colin Vize, Donald Lynam, Courtland Collison and Joshua Miller reported substantial overlap specifically between Machiavellianism and psychopathy as conventionally measured, again raising the question of whether existing Machiavellianism scales measure something genuinely separate from psychopathy or a differently worded version of the same construct.
This is a serious problem for the D-factor model to sit alongside, because Chapter 6 of this article already flagged the same pattern independently: in the founding D-factor studies, psychopathy items showed the highest loading on the general D factor of any of the nine traits, with almost none of the variance in psychopathy items left over once D was accounted for. If Glenn and Sellbom are broadly right that much of what looks like "narcissism" and "Machiavellianism" in existing measures is actually psychopathy wearing a different label, and D-factor research independently finds that psychopathy is the trait contributing the least unique variance beyond D, a sceptical reading becomes available: perhaps the "unifying" general factor is not unifying nine genuinely separate traits so much as absorbing several measures that were already substantially redundant with one another before the D-factor model arrived. On this reading, D's strong statistical performance would be, in part, an artefact of measuring overlapping constructs rather than evidence of a single, deep underlying disposition.
The D-factor research programme's strongest counter to this reading, examined in Chapter 12, is the "indifference of the indicator" evidence: that D scores remain highly stable even when the specific items and traits feeding into the measurement are substantially altered, including when up to three-quarters of items are randomly dropped. If D were simply an artefact of Machiavellianism-and-narcissism-measures secretly being psychopathy measures, removing most of the psychopathy-adjacent content should meaningfully change the resulting factor; the evidence suggests it largely does not. There is also a more recent, partially countervailing line of evidence worth noting for balance: Christopher Kowalski and colleagues' 2024 multi-study investigation in Scientific Reports, examining Dark Tetrad measures directly across four separate studies, found that Machiavellianism and psychopathy, and separately sadism and psychopathy, showed empirically distinguishable patterns of correlation with a range of narrowband personality and behavioural measures (Kowalski et al., 2024) — evidence against, rather than for, the strong redundancy claim, at least for the specific instruments that study examined. The dispute has not been resolved by consensus.
A related, more recent development worth flagging for completeness is the 2025 publication of the Faceted Dark Triad Scale by John-Paul Martindale, David Hughes, Paul Irwing and Leigha Rose, in the journal Assessment. Their project addressed a related but distinct problem — not whether the Dark Triad traits are redundant with one another, but whether existing facet-level measures of psychopathy, narcissism and Machiavellianism were internally redundant with each other, describing what they term a "jingle-jangle" problem in which near-identical facets carrying different labels, or genuinely identical facets measured by different scales, had proliferated across the literature without theoretical integration (Martindale, Hughes, Irwing & Rose, 2025). Their solution runs in something like the opposite direction to D: rather than finding a single general factor beneath the noise, they built a single non-redundant set of facets that preserves the three traits as distinguishable while removing duplication across the instruments used to measure them. That two entirely different, contemporaneous research programmes independently concluded the existing dark-trait literature suffers from serious measurement redundancy — while disagreeing sharply about whether the correct fix is to find the one thing underneath the noise or to clean up the noise itself — is perhaps the clearest evidence available that this is a genuinely open methodological problem for the field, not a settled question either side of the D-factor debate can claim to have closed.
For a reader trying to reach a practical verdict rather than simply cataloguing the dispute: the weight of evidence supports treating D as capturing something real and stable, not a pure statistical artefact — the longitudinal evidence in Chapter 13 is difficult to explain away on artefact grounds alone. What remains genuinely unsettled is how much of that "something real" is D specifically, as opposed to a psychopathy-adjacent general antagonism factor that would emerge from almost any sufficiently broad collection of dark-trait measures, D-factor branded or otherwise. Both this article and, more importantly, the primary literature it draws on treat that as an open question rather than a closed one.
PART V — APPLICATION AND CONCLUSION
Chapter 16: Recognising D at Work — and Why Testing for It Is a Trap
Everything in this article to this point has been about establishing what D is and how well the underlying evidence holds up. This chapter turns to what a manager, HR professional, or colleague might reasonably do with that evidence — and, just as importantly, what the evidence does not support doing with it.
Start with recognition, because it is the genuinely useful, low-risk application of this material. The pattern this article describes is not primarily a vocabulary problem — swapping "narcissist" for "high in D" changes very little on its own — but a noticing problem. The practical value of the unifying framework in Chapter 2 is that it redirects attention away from diagnosing which specific label fits a difficult colleague, a task few managers are qualified to do reliably and fewer still should attempt, and toward a smaller, more tractable set of behavioural questions: does this person's account of events consistently minimise the cost their choices impose on others? Does their justification for a decision, when scrutinised, rely on a worldview where ruthlessness is simply realism, or where they are straightforwardly exempt from standards applied to everyone else? Do they seem to derive some extra, disproportionate satisfaction from a colleague's setback that the situation itself doesn't explain? None of these questions require a clinical qualification to ask, and none of them require settling whether a specific person is "more Machiavellian or more narcissistic" — a question Chapter 1 suggested may not even be well formed.
The most recent evidence on where this pattern is concentrated in the workforce comes from a study published in early 2026, after most existing commentary on this topic had been written. Lea de Hesselle and colleagues, including two of the model's founding authors, linked D directly to Holland's RIASEC framework of vocational interests — the six-category system (realistic, investigative, artistic, social, enterprising, conventional) widely used in careers guidance and occupational psychology — using six separate samples totalling 8,409 participants across Germany, the United States and Denmark, and combining self-reported interests with, notably, officially registered occupational data rather than relying on self-report alone (de Hesselle et al., 2026). The clearest and most consistent finding across every sample was a negative relationship between D and both social interests — work built around helping, teaching, or caring for others — and artistic interests, involving creative or expressive work. This held up whether the researchers used what people said about their own interests or the job people were officially registered as doing, which strengthens confidence that the pattern reflects something real about occupational sorting rather than a quirk of self-report.
The picture for enterprising interests — roles built around persuasion, leadership and competitive advancement, the category popular commentary most often assumes should be D's natural home — was considerably messier than a simple "dark personalities gravitate toward business and leadership roles" narrative would predict. The relationship was positive in the German samples, statistically indistinguishable from zero in the US-based samples and mixed in the Danish occupational registry data. Realistic, investigative and conventional interests showed no consistent, substantial relationship to D in any sample. The honest reading of this new evidence is not "dark personalities are drawn to management" — a claim the data do not clearly support — but something narrower and, in a way, more useful: high-D individuals appear reliably under-represented in interests and roles centred on helping and creative expression, while their relationship to competitive, advancement-oriented roles depends heavily on the specific labour market and culture being studied, a genuinely open empirical question rather than a confirmed pattern.
This finding sits alongside older, well-established evidence specifically on workplace consequences. Ernest O'Boyle, Donelson Forsyth, George Banks and Michael McDaniel's 2012 meta-analysis in the Journal of Applied Psychology, drawing on the pre-D-factor Dark Triad literature, found the three traits associated with counterproductive work behaviour and negatively associated with both task performance and organisational citizenship behaviour — the discretionary, above-and-beyond conduct that keeps teams functioning smoothly (O'Boyle, Forsyth, Banks & McDaniel, 2012). Why a given dark-leaning employee's behaviour actually surfaces in visible, actionable ways — rather than staying latent, masked by a permissive environment, or channelled into competent rather than destructive behaviour — is a separate question with its own dedicated literature, examined in this series' companion article on the bad apple effect through the lens of trait activation theory: the finding that personality is expressed as behaviour only when situational cues make it relevant to do so, meaning the same underlying disposition can look very different in a well-governed team than in a poorly governed one. This is the single most important qualifier on everything else in this chapter: a high-D disposition is not, on its own, destiny. Two colleagues with a similar underlying disposition can produce very different organisational outcomes depending on whether their environment consistently rewards, tolerates, or actively constrains the behaviour that disposition inclines them toward — which is also the reason environmental and governance fixes, examined in this series' companion articles, are never optional extras alongside individual-level recognition, but the other half of any serious response.
A further related caution concerns leadership specifically, because it is the context in which this material is most often misapplied toward overconfidence. It is tempting to read the enterprising-interests finding above as confirmation that dark-leaning individuals systematically rise to senior positions, a claim that circulates widely in popular commentary on corporate psychopathy. The de Hesselle et al. (2026) findings, taken at face value, do not support that claim cleanly — the enterprising-interest relationship was inconsistent across countries and non-significant in the US samples specifically. What the older O'Boyle et al. (2012) meta-analysis does more directly support is a narrower and more defensible claim: once dark-leaning individuals are in a role, their behaviour there tends toward the counterproductive and away from the discretionary, cooperative conduct that organisations depend on informally. Whether they are more likely to arrive in senior roles in the first place is a separate, less settled empirical question, and this article declines to overstate what the current evidence shows.
Now for what the evidence does not support: using a D-factor assessment, or any of its component measures, as a screening or selection tool for hiring or promotion decisions. This caution rests on more than general prudence, and it is worth stating the specific reasons plainly. First, the D70, D35 and D16 item sets described in Chapter 12 were developed and validated using anonymous, low-stakes research samples, where participants had no incentive to present themselves favourably. Personality measures — and dark-trait measures in particular, given the transparency of what a high score would communicate — are well documented to be vulnerable to faking under evaluative conditions, such as a job application, where a motivated, often reasonably socially skilled respondent can shift their answers substantially toward a more desirable profile. A measure's strong validity in an anonymous research context provides no guarantee it retains that validity once a hiring decision depends on the answer. Second, the model's own developers have not published or endorsed the item sets for selection use; on the contrary, the explicit purpose stated throughout their published work is research measurement of a general disposition, not individual-level clinical or employment decision-making, and the authors are direct that even their most granular models cannot cleanly separate an individual's standing on any single specific trait from their general D score. Third, and most seriously from a legal and ethical standpoint, personality-based screening for constructs this broad and this value-laden carries real adverse-impact and discrimination risk, and the appropriate response to a suspected pattern in an existing employee is a documented, behaviourally specific performance or conduct conversation — the kind of framework developed in this series' companion article on the bad apple effect — not a psychometric test applied at the hiring stage to people who have not yet done anything.
The distinction worth holding on to is this: the article's evidence base is strong enough to justify taking the underlying pattern seriously when it shows up in documented behaviour. It is not strong enough, and was never intended, to justify testing for the pattern in people before it has shown up at all.
Chapter 17: Conclusion
11.1 What the Evidence Actually Reframes
The core empirical claim examined in this article holds up well under scrutiny. Nine dark traits with genuinely independent measurement histories — some dating back over fifty years, developed by entirely separate research groups for entirely separate purposes, from tax-evasion research to clinical psychopathy assessment — correlate with one another far more than independence would predict, and a substantial proportion of that correlation is explained by a single general factor. That general factor, D, shows rank-order stability exceeding any of the nine specific traits it subsumes over a four-year gap, predicts later change in those specific traits, and correlates with genuine behavioural outcomes — dishonesty in an incentivised task, self-interested allocation of real money, willingness to reduce a stranger's payoff for no material gain — not merely with further rounds of self-report. This is a materially stronger evidence base than exists for many long-accepted constructs in personality psychology, and it justifies treating "difficult" workplace behaviour as more coherently patterned, and more traceable to a single underlying disposition, than the older separate-trait vocabulary implied.
11.2 What Remains Unknown
Two genuinely open questions sit alongside that evidence, and this article has treated both honestly rather than resolving them by omission. The first is whether D is empirically distinguishable, in practice rather than only in theory, from the strongest pre-existing candidates for a "dark core" — low HEXACO Honesty-Humility and low Big Five Agreeableness — a live, published debate involving direct exchanges between the D-factor research group and its critics that this article has not attempted to adjudicate. The second, and arguably more serious, question is whether D is unifying nine genuinely independent traits or substantially absorbing measures that were already redundant with psychopathy specifically before the model existed — a concern this article's own chapter-by-chapter treatment of the nine traits surfaced independently, well before Chapter 15 addressed the academic literature making the same point directly. Readers should treat any confident claim, from any source, that either of these debates is settled with real scepticism; the primary literature itself does not make that claim.
11.3 A More Accurate Way to Think About "Dark Personality" at Work
The everyday intuition this article set out to examine — that difficult colleagues who look very different on the surface often turn out, on closer acquaintance, to share something underneath — survives contact with the evidence considerably better than most folk psychological intuitions do when tested this rigorously. What the evidence does not support is the version of that intuition that reaches for a single diagnostic label, applies it from a distance, and treats the label itself as the useful output. The more defensible version, and the one this article's evidence base actually supports, is behavioural rather than diagnostic: watching for a consistent pattern of disregarding, accepting, or actively provoking cost to others, watching for the self-justifying story that accompanies it, and treating that pattern — not a borrowed clinical label, and certainly not a test score obtained without the person's knowledge — as the thing worth documenting and addressing.
A final note on how this article has tried to handle the evidence, offered because it applies equally to every other article in this series. A unifying theory with a memorable name, a free online test, and international media coverage carries an obvious risk: the temptation to treat popularity as a proxy for certainty. This article has tried to resist that temptation throughout, by giving the model's own strongest evidence its full due — the longitudinal stability findings in Chapter 13 are, on their own terms, genuinely impressive — while also giving its most serious unresolved critics, in Chapters 14 and 15, the same fair hearing rather than a token mention. Readers who came to this article already convinced that "dark personality" explains a specific colleague should leave it with a more precise, more falsifiable version of that conviction; readers who came to it sceptical of the entire dark-traits literature should leave it aware that the strongest version of this particular model has cleared empirical hurdles most personality constructs never face at all. Both positions can be held responsibly. What this article has tried to make harder to hold responsibly is a confident verdict, in either direction, that the question is closed.
