Abstract
Artificial intelligence increasingly influences high-stakes decisions, from employment and education to healthcare and public services. When such judgments are disputed, affected people may receive an explanation without being able to reach an authority empowered to change the outcome or remedy its consequences. We introduce cultural repairability: the institutional capacity to review an AI-mediated decision, correct the outcome and downstream records, provide meaningful restoration, and learn from failure through procedures people can understand and use. The framework addresses a visibility–accountability asymmetry: people are exposed to judgments about their standing but lack effective means of obtaining an answer or remedy. Culture matters because exposure, legitimate authority, and appropriate restoration may be understood differently across settings. Language accessibility also determines whether review mechanisms are usable in practice. We identify three measurable cultural mechanisms: the meaning of exposure, the credibility of review, and the meaning of restoration. We further distinguish affected people’s appraisals from institutional actors’ responsibility and capacity to respond. Four falsifiable propositions specify how reviewer authority and audience shape responses to repair, and how these effects vary with cultural orientation, institutional history, and the form of restoration. They predict whether people appeal or withdraw, how they recalibrate trust, and whether they remain willing to engage. The resulting Repairability Principle establishes a common floor of substantive review, correction, restoration, and prevention of recurrence, while allowing people to choose how restoration is delivered. Cultural repairability therefore shifts global AI accountability from whether procedures appear acceptable to whether institutions can correct harm in practice.