Bandura's (1999) analysis of moral disengagement begins from a premise that distinguishes it from character-based accounts of misconduct: most people hold moral standards they do not wish to violate, and they regulate their conduct through anticipated self-sanction, the discomfort of acting against their own standards. Conduct that would normally trigger that sanction becomes possible when the self-regulatory process is selectively deactivated. Moral disengagement names the cognitive mechanisms that accomplish this deactivation, and its central claim is that they operate on ordinary people in ordinary circumstances rather than only on the unusually callous.
Bandura organized the mechanisms across four points at which self-sanction can be disengaged. Regarding the conduct itself, people can portray harmful behavior as serving a worthy purpose (moral justification), relabel it in sanitized language (euphemistic labeling), or compare it favorably with something worse (advantageous comparison). Regarding agency, they can attribute responsibility to an authority (displacement of responsibility) or dilute it across a group (diffusion of responsibility). Regarding effects, they can minimize, disregard, or distort the harm that results. Regarding the people affected, they can deny their humanity (dehumanization) or assign them blame for what happened to them (attribution of blame). Each mechanism changes how the act is understood rather than the act itself, which is why a person using them can sincerely continue to regard themselves as someone of good character.
Organizational research has translated this framework into measurement. Bandura, Barbaranelli, Caprara, and Pastorelli (1996) developed a measure of the mechanisms and linked individual differences in disengagement to detrimental conduct. Moore, Detert, Treviño, Baker, and Mayer (2012) adapted the approach for working adults and reported, across a validation study and four further laboratory and field studies, that propensity to morally disengage predicted self-reported unethical behavior, fraud decisions, self-serving workplace decisions, and unethical work behavior reported by supervisors and coworkers, in each case beyond established individual-difference predictors and the closest existing measure of the construct. Detert, Treviño, and Sweitzer (2008) examined what predicts the propensity itself, finding in three waves of survey data from undergraduates that empathy and moral identity were associated with lower disengagement, while trait cynicism and a belief that outcomes are determined by chance were associated with higher disengagement, and that disengagement was in turn positively associated with unethical decision making.
Taken together, these findings shift the useful question away from whether a given person is ethical and toward which conditions allow ethical people to stop applying their own standards. A propensity is expressed or suppressed by circumstance, and the remainder of this monograph examines the circumstances.
Two implications of this framing are worth stating before turning to organizations. First, the mechanisms are continuous rather than categorical: everyone uses some of them some of the time, as when a tactful phrase softens an unwelcome message or a shared decision feels less personally owned. Propensity scores in the studies above describe differences in degree, not a division between the ethical and the unethical. Second, because the mechanisms work by changing interpretation, they are difficult to detect from the inside. A person using them does not experience a violation being excused; they experience an accurate understanding of the situation. This is why external structures, rather than introspection alone, carry so much of the burden of keeping standards in place.
Vaughan's (1996) study of the decision to launch the space shuttle Challenger remains the most thoroughly documented account of how a standard can erode without any single decision appearing to violate it. She described what she termed the normalization of deviance: engineering anomalies that were initially treated as unexpected were, flight after flight, reinterpreted as within the range of acceptable performance, because earlier flights with the same anomaly had ended without catastrophe. The standard was never formally abandoned. The reference point against which new evidence was judged moved with each accepted exception, until a decision that would once have been unthinkable was assessed as consistent with established practice.
The same logic appears in the organizational corruption literature as incrementalism. Ashforth and Anand (2003) identified it as one of three paths through which newcomers are socialized into corrupt practice, in which small steps that each differ only marginally from the last accumulate toward conduct the newcomer would have rejected at the outset. Experimental evidence is consistent with this account. Welsh, Ordóñez, Snyder, and Christian (2015), across four multiround studies, found that small ethical transgressions made larger later transgressions more likely, extending Bandura's framework by showing that moral disengagement can erode ethical behavior across a series of escalating indiscretions, and that inducing a prevention focus weakened the effect.
Tenbrunsel and Messick's (2004) concept of ethical fading names the perceptual side of the same process. Ethical fading describes how the ethical dimensions of a decision drop out of view when attention is absorbed by other features of the situation, with self-deception allowing the decision maker to proceed without experiencing the decision as an ethical one at all. Where moral disengagement reconstrues conduct that is recognized as morally relevant, ethical fading removes the recognition itself, and the two can operate together: a decision framed from the outset as a technical or commercial matter requires less disengagement because it triggers less self-sanction.
The organizational consequence is that a standard can be abandoned without any decision visibly abandoning it. An organization that evaluates conduct only by asking whether the most recent decision was acceptable relative to recent practice will approve each step of a trajectory it would reject if shown the endpoint beside the starting point.
Three features of review practice make this more likely. Reviews that are conducted by the same people who approved earlier exceptions inherit the reference point those exceptions created. Reviews that are scheduled only after an adverse event arrive after the trajectory has already been completed. And reviews that compare a proposal only with the preceding quarter or project give a series of small changes no opportunity to be seen as one large change. None of these features requires bad intent from anyone involved, which is precisely why they are easy to leave unexamined.
Individual-level accounts leave open why misconduct so often outlasts the people who began it. Ashforth and Anand's (2003) analysis of the normalization of corruption answers by identifying three mutually reinforcing processes. Institutionalization embeds an initial corrupt act in structures, routines, and organizational memory, so that it is repeated as precedent and performed without question, while specialization and interdependence spread the activity across so many people that none sees the whole. Rationalization supplies the shared ideologies that make the conduct acceptable to those performing it, such as the claims that nothing illegal occurred, that no one was harmed, or that loyalty to the group takes priority; because these accounts are shared and repeated, they acquire the feel of established fact. Socialization brings in newcomers, who usually arrive with ordinary ethical standards, through rewards, modeling by veterans, and the treatment of doubt as a personal shortcoming.
The correspondence with Bandura's mechanisms is close. The denial of harm resembles the distortion of consequences, the denial that victims deserve concern resembles attribution of blame, and the appeal to loyalty resembles moral justification. Moore (2008) examined moral disengagement as a component of the processes through which organizational corruption develops, an approach that places the individual cognitive mechanism inside the collective processes by which wrongdoing becomes routine, rather than beside them.
The role of specialization deserves separate attention because it connects the organizational and individual levels directly. When work is divided so that each person contributes a small, routine portion of an activity, the diffusion and displacement of responsibility that Bandura (1999) describes become structurally available rather than merely psychologically convenient. Each participant can accurately say that their own part was minor, that others reviewed it, or that it conformed to instructions. The organization has then produced the conditions for disengagement without anyone intending it, and no individual intervention will change that arrangement until the division of responsibility itself is examined.
Kish-Gephart, Harrison, and Treviño's (2010) meta-analysis supplies the broader evidentiary frame. Synthesizing more than thirty years of research, the authors organized the sources of unethical decisions into individual characteristics (bad apples), features of the moral issue (bad cases), and the organizational environment (bad barrels). Their structural equation modeling indicated that unethical choices have multiple causes, and they argued that research should examine several sets of antecedents together. An organization that attributes a failure entirely to individual character is therefore making a claim the accumulated evidence does not support.
If disengagement is sustained by organizational processes, the conditions that interrupt those processes become the more useful object of study. Kaptein's (2008) corporate ethical virtues model specifies eight such conditions, developed and tested across four empirical studies into a 58-item questionnaire: clarity, congruency of supervisors, congruency of management, feasibility, supportability, transparency, discussability, and sanctionability. The model is normative in construction, since each dimension describes a condition an ethical culture should satisfy, and its value for the present argument is that the dimensions correspond to points at which the disengagement process can be interrupted.
Clarity and the two forms of congruency address the ambiguity that rationalization requires. When expected conduct is stated concretely and supervisors and senior managers visibly behave consistently with it, there is less room for the claim that a questionable act was acceptable or customary. Feasibility and supportability address the pressure side of the process, asking whether people are expected to deliver only what they can within the standard; where feasibility is low, conflicting demands push people toward the justification that the situation left no alternative.
Transparency, discussability, and sanctionability address the points at which disengagement depends most on silence. Transparency makes conduct and its consequences visible to those who could respond. Discussability allows an ethical concern to be raised without penalty, which matters because the shared rationalizations Ashforth and Anand (2003) describe persist only while they go unchallenged. Sanctionability ensures that violations produce consequences consistently, regardless of who commits them. Treviño, Weaver, and Reynolds's (2006) review situates such conditions within a broader finding that ethical behavior is shaped by individual, group, and organizational influences operating together.
The mapping in Figure 2 should be read precisely. Kaptein's model establishes the dimensions and a validated way to measure them; the assignment of each condition to a stage of the disengagement process is a conceptual synthesis offered in this monograph, not the finding of any single study.
The literature has a direct implication for assessment design: measures that ask people to rate their own ethical standards will systematically miss the process described here. Disengagement allows a person to retain a sincere ethical self-image while acting against the standard that image implies, so a self-report of integrity cannot distinguish a person whose conduct matches their standards from one whose standards have been reconstrued in ways they no longer notice. For the same reason, the individual propensity measure developed by Moore et al. (2012), though predictive, is better suited to research and to understanding risk than to evaluating individuals.
An approach consistent with the evidence would measure the conditions that sustain or interrupt the process, and would measure them from behavior rather than stated commitment. Three features follow. First, it would assess conditions at the level of the organization and its units, using dimensions of the kind Kaptein (2008) validated rather than a single impression of ethical culture. Second, it would score each dimension from observed practice under real cost, such as whether a concern raised by a senior, well-regarded person receives the same response as one raised by a junior employee, because discussability and sanctionability are tested precisely when applying them is expensive. Third, it would track the same measures over time, since the defining feature of drift is that no single observation reveals it; only comparison against an earlier baseline, rather than recent practice, can detect a moving reference point.
A practical corollary is the use of fixed anchors. Where a written standard exists, periodic review can compare current practice directly with the standard as originally stated, and can ask evaluators outside the unit to perform the comparison, since insiders share the reference point that drift has moved. The purpose is not to assume that departures are wrongful, as legitimate standards do change through deliberate decisions. It is to make sure that change happens through such decisions, visibly and with reasons recorded, rather than through accumulated exceptions that no one chose to adopt as policy.
Language is a fourth consideration. Euphemistic labeling is one of Bandura's (1999) mechanisms, and ethical fading involves the ethical dimension of a decision dropping out of view (Tenbrunsel and Messick, 2004), so the vocabulary an organization uses for questionable conduct is itself diagnostic. Routine or technical terms are not proof of disengagement, but a systematic review of how recurring problems are described in internal documents can show where the ethical dimension has left the conversation.
Several limitations constrain how far this argument can be pressed. The first concerns the evidence on antecedents. Detert, Treviño, and Sweitzer's (2008) study drew on undergraduate samples and found only partial support for the proposition that disengagement mediates the relationship between individual differences and unethical decisions. Findings from students do not transfer automatically to working adults facing real organizational consequences, and the partial mediation result is a reminder that the causal chain from disposition to disengagement to conduct is incompletely established.
The second concerns causal direction. Welsh et al.'s (2015) multiround design shows the process unfolding across repeated decisions rather than at a single moment, which raises the possibility that disengagement is shaped by prior conduct as well as shaping later conduct. Cross-sectional measures of propensity cannot separate a disposition that precedes misconduct from a rationalization that follows it, and the staircase in Figure 1 should be read as a conceptual device rather than an empirical trajectory.
The third is interpretive. Because disengagement is defined by cognitive mechanisms that are rarely observed directly, it risks becoming an explanation available after the fact for any misconduct, in which case it would explain everything and predict little. The experimental and survey studies reviewed here guard against this by measuring propensity in advance or manipulating conditions, but the organizational case literature, including Vaughan's (1996) account of Challenger, is by nature a retrospective reconstruction of a single episode. Generalizing from one organization requires comparative evidence of the kind Kish-Gephart et al. (2010) assembled for individual decisions, which remains thinner for organizational drift. Finally, Kaptein's (2008) dimensions are validated for measurement; evidence that improving them reduces drift would require longitudinal intervention research. The framework here is best treated as a grounded structure for diagnosis, not a demonstrated model of remediation.
The research reviewed here supports a reframing of ethical failure. Moral disengagement is a normal cognitive process through which people keep a sincere ethical self-image while acting against the standard that image implies, and it is made easier by incremental change, shared rationalizations, the socialization of newcomers, and the absence of conditions that make concerns discussable and violations consistently sanctionable. Individual propensity matters, but the meta-analytic evidence indicates that it operates alongside features of the moral issue and the organizational environment rather than in place of them.
For leaders and for those who assess organizations, the central implication is that ethical risk lies less in the question of who can be trusted than in the question of what an organization's processes allow to become normal. An organization that evaluates each decision against the last one will not detect drift; one that measures behavior under real cost, against a stable standard, and over time can. The conditions that make this possible are identifiable and measurable, and building them before a visible failure is more reliable than discovering their absence afterward.