DCT Lexicon · Governing Dynamics
Belief-State Control
Management never sees the enterprise's true state, only noisy signals of it. Belief-state control makes the posterior distribution the state, and turns a partially observed problem into a fully observed one.
DCT Governing Dynamics
In depth
Formal statement
With hidden state , kernel and observation kernel , the belief updates by Bayes' rule, and the value function is defined on beliefs.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Sufficiency of the belief
The belief is a sufficient statistic: an optimal policy may be chosen as a function of alone. For linear-Gaussian models the belief is Gaussian and the Kalman filter computes it exactly. In finite POMDPs the value function is convex in the belief, and piecewise linear over a finite horizon.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In DCT
Capability, culture and readiness are hidden states measured by surveys with error. A digital twin that filters these readings is belief-state control in practice: decisions use the posterior, and the decision to survey again is itself a control.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Limits
Belief spaces are infinite-dimensional except in conjugate families; particle filters approximate them, and their degeneracy is a practical failure mode. A misspecified kernel makes the posterior confidently wrong.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Go next
- Value of Information: Information has value because it moves the belief.
- Probability Kernels: The filter propagates through the declared transition kernel (Pathway 1).
- Dynamic Risk Measures: Risk under partial information is risk on beliefs.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
References
- Åström (1965)
- Kaelbling, Littman & Cassandra (1998)
- Bain & Crisan (2009), Fundamentals of Stochastic Filtering
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In the book: Vol. II, Ch. 13: Machine Learning for Enterprise Transformation · Vol. II, Ch. 15: Enterprise Digital Twins and Autonomous Enterprise Transformation
Related: Value of Information · Probability Kernels · Dynamic Risk Measures
GD.14


