# The Second Error Function **Entity class:** Synthesis concept **Domains:** Predictive security / counterterrorism / public-health surveillance / machine learning governance ## Definition The **second error function** is the accounting term a protective system needs for the harms it produces while trying to prevent catastrophe. Every mature security or public-health system already computes a **first error function**: the attacks it missed, the outbreaks it forecast too late, the [[wiki/False Negative|false negatives]] that became casualties. The second error function counts misidentifications, contaminated graph edges, iatrogenic state changes, uncontested estimates, migrated data, decayed trust, and diverted analytic capacity, and returns those costs to the model as negative feedback. [[wiki/Security Harms Topology|Security Harms Topology]] supplies the master map of what kinds of harm exist and how they connect. The second error function is the measuring instrument applied to that map: frequency, severity, propagation radius, correction latency, recurrence, cumulative burden, and residual effect. A predictive architecture that computes only the first function becomes an **asymmetric learner**. Observations can be wrong, edges carry the error outward, interventions alter the measured state, the altered state returns as training data, secondary institutions act on the score, and costs borne by [[wiki/False Positive|false]] and marginal positives never reach the learning loop. The concept is pro-prediction and pro-protection: decision-grade maturity arrives when both functions run continuously. ## Terms and existing instruments Each term already has a first public measurement. **Watchlist correction yield** appears in GAO's 2025 finding that 88 of 289 U.S.-person watchlist-related DHS TRIP inquiries ended in removal; this is a self-selected redress cohort, not a population-wide error estimate. DOJ OIG supplies historical measures for record defects, modification and removal, review time, and downstream update delay. [[wiki/Cumulative Downstream Burden|Cumulative downstream burden]] extends the measure from one correction to every encounter, institution, derivative decision, associate record, opportunity loss, and residual effect produced before the root state was repaired. **Collateral edge burden** has its closest analog in the CalGang audit. **The iatrogenic delta** appears in Prevent project evaluations recording reduced institutional trust among participants. **Purpose-migration events** appear in the TraceTogether and Luca cases. **Trust elasticity of sensing** appears in longitudinal contact-tracing adoption research. **Resource diversion** appears in GAO's SPOT audits. ## The mirror ledger Prevention science already reasons in counterfactual ledgers: the 2025 PNAS estimate associated with [[wiki/Meyers Lab|Meyers Lab]] placed influenza hospitalizations averted by vaccination at 69,886 for the 2022–2023 U.S. season. The same [[wiki/Counterfactual|counterfactual]] machinery can estimate inflicted harm with equal rigor. The second error function is that mirror ledger made operational. ## Relationships - **principal analysis:** [[research/Harms Incurred While Bringing Preventive Systems Online|Harms Incurred While Bringing Preventive Systems Online]]. - **master harm map:** [[wiki/Security Harms Topology|Security Harms Topology]]. - **component harms:** [[wiki/Watchlist Misidentification|Watchlist Misidentification]], [[wiki/Distributed Erroneous State|Distributed Erroneous State]], [[wiki/Life-Trajectory Harm|Life-Trajectory Harm]], [[wiki/Graph Error Propagation|Graph Error Propagation]], [[wiki/Redress Opacity|Redress Opacity]], [[wiki/Iatrogenic Security Intervention|Iatrogenic Security Intervention]], [[wiki/Surveillance Purpose Migration|Surveillance Purpose Migration]], and [[wiki/Surveillance Trust Externality|Surveillance Trust Externality]]. - **measurement terms:** [[wiki/Correction Latency|Correction Latency]] and [[wiki/Cumulative Downstream Burden|Cumulative Downstream Burden]]. - **learning mechanics:** [[wiki/Observation-Induced Ground Truth|Observation-Induced Ground Truth]], [[wiki/Selective Labels|Selective Labels]], and [[wiki/Runaway Predictive Feedback Loop|Runaway Predictive Feedback Loop]]. - **governance seam:** [[wiki/Operational Activity and the Common Rule|Operational Activity and the Common Rule]] and [[wiki/Legibility-Gated Redress|Legibility-Gated Redress]]. - **governing loop:** [[wiki/CVE-CVE Convergence|CVE-CVE Convergence]] and [[wiki/Model Error|Model Error]]. - **collection:** [[collections/Terrorism, Counterterrorism, and the Intelligence Environment|Terrorism, Counterterrorism, and the Intelligence Environment]]. ## Sources / Provenance - [GAO-25-108349 — Terrorist Watchlist: Nomination and Redress Processes for U.S. Persons](https://www.gao.gov/products/gao-25-108349) (2025-08-14). - [Independent Review of Prevent's Report and Government Response](https://www.gov.uk/government/publications/independent-review-of-prevents-report-and-government-response/independent-review-of-prevent-accessible) (2023). - [Ensign et al. — Runaway Feedback Loops in Predictive Policing](https://proceedings.mlr.press/v81/ensign18a.html) (2018). - [Bi, Meyers, et al. — Estimated impact of 2022–2023 influenza vaccines on annual hospital burden in the United States, PNAS](https://pmc.ncbi.nlm.nih.gov/articles/PMC12646225/) (2025). **As of:** 2026-09-23