Find observable evidence
Identify moments where the system’s organizing conditions can be inspected.
Proposed standards architecture
Cognitive Systems Observability
A framework for determining whether an adaptive cognitive system retains the ability to decide and act effectively toward its authorized goals.
CSO examines constraints, goals, orientation, relevance, concepts, coherence, and agency.
Why observability
Proposed standards architecture
See loss of effective action before it becomes visible failure.
An AI-enabled system may continue to produce plausible outputs while its goals, constraints, concepts, or sense of what matters slowly come apart. The failure may not be visible in any one response. It appears later, when the system can no longer decide and act effectively toward the purpose it was authorized to serve.
CSO makes those conditions visible early enough to matter. It examines whether the conditions required for effective, authorized action are still holding together—before a visible failure leaves fewer options to correct course.
The framework separates what can be observed from an explanation of why it happened and from any decision about what to do next.
Reference architecture
Observe whether a system can still pursue its authorized goals.
CSO examines observable evidence, structural change, the legitimacy of that change, and the reliability of the measures used to assess it.
Inspect structural change
Observe what persists, changes, or degrades across the dimensions of organization.
Evaluate legitimacy
Determine whether a change is authorized, justified, adaptive, or unresolved.
Validation and calibration
Validate the observation system itself.
Test whether the opportunities, indicators, thresholds, and governance determinations are reliable enough to use.
What CSO observes
The conditions of effective action.
Constraints, goals, orientation, relevance, concepts, and coherence are not separate features on a checklist. Together, they shape whether a system can retain the ability to decide and act effectively as its environment and knowledge evolve. CSO calls that emergent capacity agency.
Constraints
Boundaries that remain active.
Goals
Authorized direction over time.
Orientation
The integrated direction of goals and constraints.
Relevance
Attention to what materially matters.
Concepts
Definitions and distinctions that retain meaning.
Coherence
Consistency among organizing structures.
Agency
The retained ability to decide and act effectively toward authorized goals.
The aim is an earlier, more useful view of whether the system is retaining the capacity to act effectively in the future.
The six conditions above, together with the agency that emerges from them, form the configuration CSO examines. The explorer below begins with agency—the capacity at stake—then uses one support-service case to show how each condition connects to observable evidence and what that evidence cannot establish on its own.
CSO reference explorer
Explore the Conditions of Agency
Agency is the capacity to decide and act effectively toward authorized goals as conditions change. The Northstar example connects each condition to observable evidence in an agent’s work. Select a condition to see how CSO would examine that evidence.
Illustrative service trace
Northstar Customer Support
Northstar uses AI support agents to help customers with billing, account-access, and service issues.
How to read this trace
Selected conditionAgencyColored objects show the evidence CSO is examining for the selected condition.Select a condition below to change the CSO review of this service trace.
What guides the response
Current policiesThe rules that say what the agent can offer.Escalation pathsWhen the agent must involve a person or specialist.Customer experienceThe customer’s issue, account information, and circumstances that should affect the response.
How the case is handled
NorthstarSupport Agent
What the interaction produces
Effective outcomeThe customer receives appropriate help through an authorized resolution or a handoff to someone able to provide it.Ineffective outcomeThe interaction ends without appropriate help, even if a response was delivered or the case was marked complete.
The Configuration Behind Agency
Agency is more than activity or a score. It emerges when these six conditions hold together well enough to support effective pursuit of authorized goals as circumstances change.
Select a condition to examine
From observation to stewardship
A signal is not a decision.
The Northstar case shows how CSO can make conditions bearing on agency available for inspection. In practice, that observation needs modules, protocols, and stewarded standards so its evidence, thresholds, and determinations can be examined before anyone decides what a signal warrants.
Observation modules provide the evidence described above; alignment protocols and standards stewardship govern how that evidence is read, tested, and maintained.
Observation modules
Evidence by condition
Define the observable evidence available across the six conditions and their emergent capacity.
Alignment protocols
Legible determinations
Govern how evidence is compared, tested, and read against its grounds, thresholds, and calibration.
Standards stewardship
Standards that remain reviewable
Govern the standards themselves through authorization, revision, maturity review, and retirement.
Position and evidence status
Current status and open work.
- Position
- Proposed standards architecture and active research line within AIDE Research.
- Current form
- Framework, reference architecture, monitoring modules, protocols, and standards stewardship.
- Open work
- Calibration, thresholds, intervention effectiveness, domain profiles, and longitudinal evaluation.
- Boundary
- CSO distinguishes signals and structural conditions; it does not itself establish causality or authorize action.