Radiant Preprint
Allostatic Control Systems: Predictive Set-Point Governance in Changing Environments
When does a system need to govern the defended reference condition itself, rather than merely regulate against a fixed set point?
Abstract
Many engineered systems are regulated as if the target condition were fixed. That assumption fails when demand, risk, environment, or resource constraints change faster than the original set point remains valid. This paper proposes allostatic control systems as an architectural frame for such cases.
The core distinction is between fast regulation and slow reference governance. A fast loop selects actions against the current defended condition. A slower loop evaluates whether evidence has matured enough to change the defended condition itself. Without that separation, systems can overreact to noise, defend obsolete thresholds, or oscillate between local fixes while missing the larger set-point problem.
The paper treats the framework as a structural template, not a full universal control law. It identifies design obligations around evidence time scales, admissible safety envelopes, scarce-resource validity, anti-gaming checks, proxy bundles, and recovery timing. Example domains include clinical alarms, autoscaling, fraud/risk posture, cybersecurity, and airframe or autonomous-system governance.
Fields and Methods
control systems, systems engineering, predictive control, adaptive thresholds, risk governance.
- control-architecture formalization
- worked-example analysis
- safety-envelope specification
- reference-governance checklist design
- domain-specific falsification criteria
Collaborator Profile
Control theorists, systems engineers, safety-critical systems researchers, aerospace engineers, cyber-physical systems researchers, and applied researchers working on adaptive thresholding under changing environments.
Validation Needed
- Compare allostatic set-point governance against fixed-threshold, adaptive-threshold, and model predictive control baselines.
- Define domain-specific evidence maturity and admissibility predicates.
- Test whether the fast/slow-loop separation reduces false alarms, obsolete-threshold failures, or unsafe reference drift.
- Identify domains where allostatic framing adds no value beyond existing control methods.
Publication Posture
This page describes a prepared public preprint. It should not be treated as a peer-reviewed finding unless and until a later version explicitly says so.
Research Correspondence
For collaboration inquiries or research correspondence about this work, contact contact@theradiantinstitute.org.