Applied Information Theory for Autonomy in the Multi-Intelligence Collection and Exploitation Workflow
Abstract
The proposed effort develops a foundational theory of autonomy in information gathering systems that enables the quantification of the value of information collection as well as optimization of collection strategies across competing priorities and conjectures. In our probabilistic framework, support for hypotheses is quantified in terms of the likelihood function and its role in formal measures of information including Fisher and Kullback-Leibler. Likelihood functions provide the quantitative capture of the evidential weight arising from data on all considered hypotheses, conditioned on information gathering allocations through time. Diverse data is unified through the product of likelihood functions for the various modes of intelligence bearing on the hypothesis set. Information that is expressed in the likelihood function updates Bayesian posterior distributions across all hypotheses, thus providing the formal mechanisms for direct feedback of the impact and value of new information. This framework is then used to assess and recommend future collections, in the near and far terms, closing the autonomy loop in information gathering and yielding a rigorous and complete Theory of Intelligence Information (TI2). Our approach endeavors to forge a framework that unifies exploitation and planning functions by providing a rigorous means of expressing the value of distinguishing between plausible lines of reasoning through quantitative evidence. The value of that measured resolution between plausible inferences is weighed against the cost of scarce resource allocation to produce an optimized schedule of activities. This can only be done in a quantified structure and this is what we will develop under this effort. These results will have significant and enduring scientific research value on a broad range of Public Sector and National Security problems by providing an enabling and unifying calculus to quantify the value of information collection across competing priorities and conjectures, and thereby improving orchestrated resource management capabilities. In particular, this framework can support the familiar workflow of Key Intelligence Topics (KITs) and Key Intelligence Questions (KIQs), in which intelligence analysts evaluate evidence supporting and refuting hypotheses on KITs through the dialogue of KIQs, often working with intelligence specialists in SIGNINT, ELINT, IMINT, and other domains. KIQs can be augmented to fully capture and express intelligence needs in terms of data gathering requirements. It is expected this research into the Theory of Intelligence Information (TI2) and its applications will advance the state-of-the-art in autonomy of information gathering systems by defining the underlying calculus on which an optimizing control theory can accommodate discrete, asynchronous sensor windows, while simultaneously performing robust planning and decision support for scheduling in the presence of uncertainty. This work will be conducted by Metron, Inc. under the direction of Dr. Jeffrey Silver as Principal Investigator.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 06, 2016
- Source ID
- N002441610020
Entities
People
- Jeffrey Silver
Organizations
- United States Air Force