# Observation-Induced Ground Truth **Entity class:** Machine-learning and measurement concept **Domains:** Predictive policing / counterterrorism analytics / surveillance science ## Definition **Observation-induced ground truth** is the condition in which the act of observing changes what gets observed, so a deployed model's training data become endogenous to the security system. Once a model directs attention, collection, or intervention, it determines which outcomes become visible and which remain unobserved. ## Label-generating regimes A deployed security model acquires labels through three regimes. **Proxy labeling** turns system outputs such as referrals, listings, detentions, and interdictions into validation labels. **Deferred intervention** lets a forecast mature so its outcome can be observed. **Perturbation** introduces a stimulus and measures the response; it is documented as a capability in the 2014 PNAS emotional-contagion study and as case practice in the informant record examined by Human Rights Watch. ## Formal anchors [[wiki/Selective Labels|Selective labels]] formalize outcomes observed only where decision-makers acted; performative prediction (Perdomo et al., 2020) formalizes distribution shift caused by deployed predictions; [[wiki/Runaway Predictive Feedback Loop|runaway feedback loops]] (Ensign et al., 2018) prove the mechanism for [[wiki/Predictive Policing|predictive policing]]. ## Relationships - **mechanisms:** [[wiki/Selective Labels|Selective Labels]] and [[wiki/Runaway Predictive Feedback Loop|Runaway Predictive Feedback Loop]]. - **loop context:** [[wiki/Graph-to-Collection Feedback Loop|Graph-to-Collection Feedback Loop]] and [[wiki/Feedback Loops|Feedback Loops]]. - **scarcity driver:** [[wiki/Base Rate|Base Rate]]. - **accounting frame:** [[wiki/The Second Error Function|The Second Error Function]]. - **analysis:** [[research/Harms Incurred While Bringing Preventive Systems Online|Harms Incurred While Bringing Preventive Systems Online]]. - **collection:** [[collections/Terrorism, Counterterrorism, and the Intelligence Environment|Terrorism, Counterterrorism, and the Intelligence Environment]]. ## Sources / Provenance - [Ensign et al. — Runaway Feedback Loops in Predictive Policing, PMLR 81](https://proceedings.mlr.press/v81/ensign18a.html) (2018). - [Lakkaraju et al. — The Selective Labels Problem, KDD](https://www.cs.cornell.edu/home/kleinber/kdd17-selective.pdf) (2017). - [Perdomo et al. — Performative Prediction, ICML](https://proceedings.mlr.press/v119/perdomo20a.html) (2020). - [Kramer, Guillory, and Hancock — Experimental evidence of massive-scale emotional contagion through social networks, PNAS](https://www.pnas.org/doi/10.1073/pnas.1320040111) (2014). **As of:** 2026-09-23