# Closed-Loop Predictive Intelligence
**Closed-loop predictive intelligence** is an iterative analytic process:
**observe → resolve entities → construct relationships → infer hidden state → quantify uncertainty → forecast transitions → identify the most informative missing observation → lawfully collect or fuse evidence → update → determine whether intervention is justified → audit**
The process is closed because forecasts influence collection priorities, new observations revise the model, and the revised model changes the forecast. Authority, provenance, and auditability remain part of the loop.
## Relationships
- **analytic model:** [[wiki/Counterterrorism Predictive Graph|Counterterrorism Predictive Graph]].
- **collection logic:** [[wiki/Value of Information|Value of Information]] and [[wiki/Collection Management|Collection Management]].
- **operational destination:** [[wiki/Probabilistic Precrime|Probabilistic Precrime]].
- **trust layer:** [[wiki/Machine-Readable Assurance|Machine-Readable Assurance]].
- **collection route:** [[wiki/Counterterrorism Collection - Predictive Policing and Probabilistic Precrime|Predictive Policing and Probabilistic Precrime]].
## Sources / Provenance
- [RAND — Predictive Policing: The Role of Crime Forecasting in Law Enforcement Operations, 2013](https://www.rand.org/content/dam/rand/pubs/research_reports/RR200/RR233/RAND_RR233.pdf)
- [[research/Machine-Readable Assurance and the Convergence of Intelligence|Machine-Readable Assurance and the Convergence of Intelligence]]
**As of:** 2026-09-23