Developed by Jonathan Shay in his work with Vietnam veterans and extended by Alford into organizational contexts, moral injury names a conscience-based wound categorically different from fear-based trauma. PTSD is produced by what was done to you; moral injury is produced by what you did or failed to do. The Orange Pill's confession of building addictive products whose downstream harms became visible only later is, in Alford's framework, a textbook case of moral injury. The AI transition generates moral injury at scale: among builders who know what they are amplifying, among users who know the cost of their dependence, among bystanders who saw a colleague destroyed for testifying and chose not to intervene.
Shay's original work showed that Vietnam veterans suffered a form of lasting psychological damage that standard PTSD diagnoses could not capture — damage that tracked not the terror of combat but the moral fracture of acts committed, ordered, or witnessed. Alford extended the framework to organizational contexts: the middle manager who processed the layoffs, the auditor who signed off on what she knew was irregular, the engineer who shipped the system she knew was unsafe.
The distinctive feature of moral injury is its recursive character. The injury is not merely that one participated in the harm; it is that participation is now part of one's story, and the story must either be reckoned with or repressed. Repression works for a while and then fails; reckoning requires a moral vocabulary that the culture often does not supply. The Orange Pill's author is notable for the rare act of public reckoning — the naming of complicity in earlier products — and Alford's framework identifies this as the precondition for any honest engagement with the AI moment.
In the AI transition, the categories of the morally injured are wider than usual. They include the builders of systems whose downstream consequences they can partly foresee and partly not. They include the users whose daily engagement contributes to outcomes — attention colonization, labor displacement, knowledge atrophy — they would not choose if asked. They include the silent middle that saw clearly and said nothing. Each of these positions carries the structure of moral injury even when it lacks the intensity of combat or direct wrongdoing.
The clinical consequence of moral injury is what Shay called the undoing of character: not the loss of specific capacities but the dissolution of the moral self that had organized one's capacities into a coherent life. The cure, Shay argued, is neither therapy nor absolution but communalization — the telling of the story to a community that can hear it without flinching and without dismissing it. Alford's framework predicts that the AI discourse currently lacks the communal structures required for this telling, and that the moral injuries it is generating will therefore compound rather than heal.
Jonathan Shay developed the concept in Achilles in Vietnam (1994) and Odysseus in America (2002), drawing on Homer to name what the VA's diagnostic categories were missing. Alford extended the framework to organizational ethics in Moral Injury and Nonviolent Resistance (2018), arguing that institutional dynamics produce the same character-undoing wound as combat, through different mechanisms.
The contemporary application to AI work is a natural extension, and has been developed by AI safety researchers, labor scholars, and clinicians working with workers displaced by automation.
Conscience, not fear. Moral injury is produced by what you did, not by what was done to you — distinct from PTSD.
Character undoing. The injury dissolves the moral self that organized capacities into a coherent life.
Communalization cures. Healing requires telling the story to a community that can receive it — not therapy, not absolution.
Recursive structure. Participation becomes part of one's story; repression works only temporarily.
AI-scale prevalence. Builders, users, and bystanders all occupy positions that generate the wound.
The framework's extension beyond combat has been criticized as diluting a specific clinical category. Defenders respond that the category's power lies in its recognition of conscience-based wounds as psychologically real, and that the extension does not weaken this recognition but generalizes it to the many contexts where similar structural wounds occur.