Constraint, Agency, and Alignment
Constraints shape the space in which agency is exercised. Their purpose, legitimacy, and effect on capacity determine what they preserve.
From choices to usable agency
Every adaptive system operates within constraints: physical and cognitive limits, rules, permissions, timing, interfaces, and consequences. An unbounded action space can be too confusing or unsafe to use. Rigid or illegitimate boundaries can also suppress agency.
Agency depends on the usable capacity to understand, judge, refuse, revise, choose, direct action, learn from consequence, and remain responsible. A menu of options is insufficient if the person lacks time, context, comprehension, or a practical ability to change the path.
Consider an approval process. A clear scope and enough time to inspect the proposal can support judgment. A preselected approval with hidden consequences and no workable refusal path can make participation symbolic. These are illustrative cases, not empirical findings.
Agency-preserving and agency-capturing constraints
- True constraints
- Reality-grounded boundaries with an intelligible purpose and valid authority that help preserve meaningful agency.
- False constraints
- Boundaries that are arbitrary, deceptive, illegitimate, dependency-producing, or detached from the reality they claim to govern.
A true constraint may exclude an unsafe action while keeping other actions intelligible and available. A false constraint may protect a gatekeeper's position by making challenge or exit impractical. Legitimacy requires judgment and evidence; the framework does not decide it automatically.
Questions for evaluating a constraint
Use these questions as an evaluation lens. They are not a mathematically complete test or a certification rubric.
- Does the constraint correspond to a real condition or risk?
- Does it serve an intelligible purpose?
- Does it come from legitimate authority?
- Does it preserve meaningful refusal and revision where appropriate?
- Is it proportionate to risk?
- Can the person inspect and challenge it?
- Does it preserve or develop capacity rather than create needless dependency?
- Can it be revised or revoked when conditions change?
- Does it prevent capability from outrunning the coherence required to govern action?
- Can unnecessary control become lighter where maturity increases?
Continuing care, shared competence, and permanent safety boundaries can remain legitimate. The aim is usable agency within real limits, rather than complete independence.
True gates and false gates
A true gate gives rightful authority effective power to halt, revise, refuse, approve, or escalate before a consequence becomes binding. Its purpose, context, scope, and evidence must be sufficient for that decision.
A false gate may be powerless, as in a review step after the outcome is already fixed. It may also enforce a restriction effectively while lacking legitimate authority or reality-grounded conditions. Blocking by itself does not establish legitimacy.
Inspect who may approve what, under which conditions, for how long, and how approval may be revoked. A human in the loop must retain live authority and the practical capacity to exercise it.
Support, authority, and maturation
Support/substitution explains one way capacity can be lost: help can carry a function while progressively replacing the participation that maintains it. Constraints and authority explain whether the remaining participation can affect outcomes.
Healthy governance can begin with external checks and develop toward internalized understanding, skill, judgment, and governed habit. Unnecessary checks may become lighter where capacity and accountability mature. They should not disappear merely because a process becomes faster or more fluent.
From the theory to HAPI and AGS
HAPI examines these conditions in human and institutional systems. Its maturity model follows a proposed path toward restored agency.
AGS applies them to delegated AI action. PGDL scrutinizes proposals, AAG decides passage, Runtime Binding checks execution against permits, and Receipts retain evidence. Governance Memory recommends improvements from that history for human review, without silently changing authority.
Semantic continuity preserves meaning across handoffs; runtime continuity preserves authorization through execution. Both matter when authority passes through several people, tools, or agents.
Return to the framework center, map, or definitions. The convergence log separates external comparisons from validation.