{"graphSha256":"f29d7b1eedeb88387238d2abda347f810d2179c5fdbead47c3ef48507ab20ca5","node":{"id":"T1391","type":"ficha","title":"The Limits of AI Oversight","url":"https://dakleyer.github.io/perspective-beyond/sources/the-limits-of-ai-oversight-t1391/","original":"https://tegrity.ai/the-limits-of-ai-oversight/","aliases":[],"clusters":["tegrity"],"tags":["autoridad","capacidad-humana","control-minimo","decision","documentacion","empresa","evidencia"],"summary":"","commit":null,"sourcePath":null,"commitUrl":null,"documentId":"doc-4be828e99f81281c6e","tagWeights":{"autoridad":0.0625,"capacidad-humana":0.1563,"control-minimo":0.0938,"decision":0.25,"documentacion":0.2187,"empresa":0.1562,"evidencia":0.0625},"reading":"Enterprise AI is commonly organized as a statistical core wrapped in a deterministic envelope of guardrails, rules, audits, and human review. The implicit promise is that additional oversight will reduce risk while also reducing human burden. This paper argues for a more conditional conclusion. A guardrail operates on formalised representations and records, not directly on the physical or operational system they represent. If input formation maps two decision-relevant world states to the same record, no downstream guardrail reading only that record can distinguish them. The unresolved decision must then be handled through additional information, abstention, risk acceptance, redesign, or external review. In accountable enterprise settings, this residual work commonly falls on people. Guardrails can remove substantial human work, but they also generate exception handling, reconciliation across tools, policy maintenance, and governance overhead. We model the resulting net human oversight load and define the AI Exhaustion Threshold as the point at which this load exceeds available expert capacity. The crossing is identified with Human Intelligence Debt: cognitive capacity diverted from judgment, strategy, design, and knowledge creation into validation and reconciliation. The paper therefore advances a falsifiable, thresholded prediction rather than an unconditional law: oversight may initially reduce Human Intelligence Debt, but beyond a coupling and exception threshold, further embedded guardrails increase it. The next empirical paper in this program will test that relationship and determine whether it is monotone, thresholded, or U-shaped in real deployments.","readingSource":{"url":"https://tegrity.ai/the-limits-of-ai-oversight/","section":"Abstract","observed":"2026-11-04","sourceSha256":"7189f3c1f768e40ebd899cc3cb38793156863796f7da90d502c759fc1ef0550e"},"readingSummary":"Enterprise AI is commonly organized as a statistical core wrapped in a deterministic envelope of guardrails, rules, audits, and human review.","contextTokens":405},"relations":[["apoyo:T1391:C26","T1391","C26","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C28","T1391","C28","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C34","T1391","C34","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C35","T1391","C35","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C37","T1391","C37","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C41","T1391","C41","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C59","T1391","C59","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C71","T1391","C71","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C75","T1391","C75","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C80","T1391","C80","documenta el concepto","",null,"apoyo",""],["apoyo:T1391:C86","T1391","C86","documenta el concepto","",null,"apoyo",""],["documento:D021","T1391","N1279","candidate transfer","Decision authority and timely review capacity should be specified separately in the governance contract.",null,"documento","El paper no demuestra por sí solo un umbral individual de agotamiento ni invalida una autoridad por falta de respuesta puntual."],["documento:D035","N864","T1391","analogy only","The GEPLAN interface illustrates why human authority needs usable decision context, not merely a confirmation button.",null,"documento","Una interfaz simple puede ocultar incertidumbre. El caso no mide formalmente carga de oversight ni demuestra que toda simplificación sea segura."],["documento:M006","E1704","T1391","candidate transfer","Automation readiness should account for the review and exception-handling work that remains after routine execution is reduced.",null,"documento","Es un diseño de evaluación, no una afirmación de que la oferta ya use o valide el umbral de agotamiento."],["documento:M014","N1281","T1391","analogy only","Trust supports collaboration, but timely review still depends on explicit decision context and actual response capacity.",null,"documento","Confianza social y capacidad cognitiva no son medidas equivalentes ni una sustituye a la otra."],["documento:F003","T1391","T1400","documented use","The oversight-threshold study tests whether added review and reconciliation outweigh removed work, rather than treating guardrail count as a measure of debt.",null,"documento","No todos los controles aumentan carga ni la carga necesaria es deuda. La forma en U y el umbral no se presuponen observados."],["documento:F017","T385","T1391","documented use","A perfectly traceable rule chain cannot recover a decision-relevant distinction that input formation has already removed.",null,"documento","La limitación es condicional a recibir la misma representación; no afirma que ninguna tecnología pueda adquirir nueva evidencia o rediseñar la captura."],["documento:F022","T2729","T1391","candidate transfer","Control burden should be compared among configurations that meet the same objective and evidence obligations, including the human response work they create.",null,"documento","Mínimo es relativo al conjunto evaluado, no óptimo global demostrado; un sistema sin evidencia suficiente no es barato sino no cualificado."],["documento:F030","N1285","T1391","analogy only","Reducing decision handoffs helps only if the receiving owner retains enough context, authority and response capacity to resolve the operational case in time.",null,"documento","No toda capa es redundante ni eliminar una revisión independiente es equivalente a eliminar pérdida de traducción."],["documento:A005","U1494","T1391","analogy only","The rejected turnstile is a field example of why control value must include its effect on the operating flow, alongside the work and risk it removes.",null,"documento","El caso no prueba la curva de carga de oversight ni convierte todo control físico o humano en deuda."],["documento:R049","E1570","T1391","editorial comparison","For a complementary reading, see [The limits of oversight and input recovery](https://tegrity.ai/the-limits-of-ai-oversight/).",null,"documento",""],["documento:R135","N1287","T1391","editorial comparison","For a complementary reading, see [What oversight cannot recover from missing inputs](https://tegrity.ai/the-limits-of-ai-oversight/).",null,"documento",""],["documento:R167","T1391","T874","editorial comparison","For a complementary reading, see [The Human Intelligence Debt definitions](https://tegrity.ai/human-intelligence-debt/).",null,"documento",""],["documento:R168","T1396","T1391","editorial comparison","For a complementary reading, see [The structural limits of AI oversight](https://tegrity.ai/the-limits-of-ai-oversight/).",null,"documento",""],["documento:sac-rel-e59c97e8d7eaf3","sac-31c5add4fe1a12","T1391","referencia de la fuente","La fuente enlaza esta lectura en «8.2 Other article sources» con el rótulo «Published source».","https://github.com/dakleyer/structural-awareness-contributions/blob/58662e16cf6641e7f0c73b6ab30047abcf68d659/research/ecosystem-awareness/DDS_POTENTIAL_CONTRIBUTIONS_REGISTER_v0.1.md","documento",""],["documento:sac-rel-958a1f635a4486","GH037","T1391","referencia de la fuente","La fuente enlaza esta lectura en «Annex B — Cross-series bridge map» con el rótulo «AI Operational Integrity ↔ HID».","https://github.com/dakleyer/structural-awareness-contributions/blob/11fbac15d6f2b790d459ab507633b2293e302709/research/ecosystem-awareness/baseline/01K_HUMAN_INTELLIGENCE_CAPACITY_ESCALATION_AND_DEBT_PROFILE_v0.1.md","documento",""],["documento:sac-rel-d132a67c84bded","sac-ebbabcb7ca63fa","T1391","referencia de la fuente","La fuente enlaza esta lectura en «Research lineage» con el rótulo «The Limits of AI Oversight».","https://github.com/dakleyer/structural-awareness-contributions/blob/db6d444867a27510c3e640e3e71904ae11b01fe4/submissions/itu-fg-tida/2026-theme-contributions/operational-human-oversight.md","documento",""]]}