Rights Constraint¶
A categorical constraint — instantiates Tradeoff Guardrail
Marks certain protections as categorically off-limits — not a weight to be balanced but a line no amount of aggregate benefit may cross.
A Rights Constraint designates a protection that may not be entered into the trade-off arithmetic at all. Its defining move — the thing no other guardrail in this family does — is that it refuses to be a quantity. A minimum threshold says "at least this much"; a rights constraint says "not for any amount." It does not lower the priority of a violating option, it removes the option from the choice set, and it does so regardless of how large the countervailing benefit is. Where the other mechanisms here draw a floor you must stay above, this one draws a side condition on permissible action: an option that violates it is not a worse choice, it is not a choice.
Example¶
A public-benefits agency automates eligibility decisions to clear a backlog — a legitimate optimization, and a fast one. Cost and speed pressure push toward auto-denying thin or ambiguous cases, since a denial closes a file as neatly as an approval. A Rights Constraint sits outside that calculation: no benefit may be terminated or denied without the claimant receiving notice, a stated reason, and a real opportunity to appeal to a human — a due-process protection. This is not weighed against throughput. When the vendor proposes a "silent auto-close" for non-responsive cases that would clear thousands of files a week, the constraint does not ask how many files or how much money; it rules the design impermissible because it removes the appeal, and the proposal has to be rebuilt so that even an auto-close still issues notice and preserves the appeal path. The agency logs why the faster design was rejected — which protected right, and the reasoning — so the decision can be audited and so a future team cannot quietly reintroduce the same shortcut.
How it works¶
- Name the protection categorically. The right is stated as a condition on what may be done to a person or interest, not as a level of a metric — so there is no number to negotiate toward.
- Convert it to a hard filter on options. Any option that would violate the right is struck from the set before optimization runs, so the optimizer never gets to weigh it against benefit.
- Refuse the aggregate-benefit override. The constraint is explicitly insensitive to "but the total gain is huge" — that argument is the exact one it exists to block, and admitting it dissolves the constraint.
- Record the rejection reasoning. When an option is removed, the rationale (which right, why it applied) is written down, both to make the constraint auditable and to keep it from being re-litigated case by case.
Tuning parameters¶
- Scope of the right — who and what it covers, and its edge cases. Drawn too broadly it blocks benign action; too narrowly it fails to protect the case that matters.
- Absoluteness — truly exceptionless versus overridable only by a like-kind right (never by mere aggregate benefit). Genuine absoluteness is strongest but can collide with other rights; the harder design question is which right yields to which.
- Grounding — whether the right rests on law, contract, professional ethics, or stated organizational value. Firmer grounding gives it force against pressure; a merely-asserted right erodes first.
- Remedy on breach — reject-and-redesign versus halt-and-escalate-to-a-rights-owner. Automatic rejection is cleaner; escalation handles genuine rights-versus-rights conflicts that no rule can pre-decide.
When it helps, and when it misleads¶
Its strength is holding a line that thresholds cannot: some sacrifices stay wrong even when the spreadsheet says they pay, and a rights constraint is what keeps a sufficiently large benefit from buying its way through a protection. It works precisely because it treats the right as a side constraint on action rather than a value to be maximized — you may not violate it in pursuit of the good, however great.[n1]
It misleads when everything becomes a "right." Inflating ordinary preferences into inviolable constraints paralyses the decision and cheapens the real rights by association — the whole point is that the category is small and load-bearing. It also fails in the reverse direction: a right asserted but not grounded in law, contract, or a defended value is the first thing to fold under pressure, dressed up as principle but traded away in practice. And a rights constraint cannot itself resolve a clash between two rights — that needs a human owner, not a rule. The discipline is to keep the set of true constraints deliberately short, ground each one, and route genuine rights-versus-rights conflicts to escalation rather than pretending the rule already answered them.
How it implements the components¶
protected_invariant— names the protected thing as a categorical right or protection, not a quantity, which is what lets it resist being priced.violation_response— defines the response as impermissible: the violating option is rejected or sent back for redesign, and (unlike a threshold breach) it is not curable by paying more benefit.decision_rationale— records which right applied and why an option was struck, making the constraint auditable and resistant to quiet re-introduction.
It deliberately does not implement a minimum_threshold — a right is categorical, not a dial, which is exactly its distinction from Safety Floor and Compliance Threshold Check. Nor does it run the standing check that the protection held over time (monitoring_and_audit) — that is an Exception Register — or convene the human judgement a novel rights conflict needs, which is an Ethical Guardrail Review.
Related¶
- Instantiates: Tradeoff Guardrail — the rights constraint is the pattern's hardest form: it protects a value by refusing to let it be traded at all.
- Sibling mechanisms: Safety Floor · Minimum Service Guarantee · Nonfunctional Requirement · Stop-Ship Criterion · Ethical Guardrail Review · Compliance Threshold Check · Quality Gate · Budget Floor · Exception Register
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: Rights Constraint operates as a standing rule, threshold, contractual commitment, or policy constraint governing future conduct because it marks certain protections as categorically off-limits — not a weight to be balanced but a line no amount of aggregate benefit may cross.
Independent corroboration: The frozen evidence defines Rights Constraint as 'Marks certain protections as categorically off-limits — not a weight to be balanced but a line no amount of aggregate benefit may cross', so its operative form is Rule, Policy & Commitment.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Convergent development
Present-day reach: Universal
Rationale: Categorical rights that constrain aggregate welfare are canonical deontological and political-philosophy concepts.
Related originating lineages:
- Law & Governance — Constitutional and human-rights law independently institutionalized nontradeable protections.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: marks certain protections as categorically off-limits — not a weight to be balanced but a line no amount of aggregate benefit may cross.
Review resolution: Both blind reviewers agree that philosophy is the primary historical origin. Explicit reconciliation of alternate origin disagreement, origin mode disagreement, domain reach disagreement starts from reviewer_a’s mechanism-specific evidence: Categorical rights that constrain aggregate welfare are canonical deontological and political-philosophy concepts. Reviewer A proposed alternates=law_governance, origin_mode=convergent, domain_reach=universal, and encyclopedia_synthesis=false; reviewer B proposed alternates=law_governance, mathematics, origin_mode=single_lineage, domain_reach=multi_domain, and encyclopedia_synthesis=false. The final record retains every independently supported alternate from either review (law_governance, mathematics) without an arbitrary cap, selects origin_mode=convergent to represent the combined lineage evidence, and keeps domain_reach=universal and encyclopedia_synthesis=false from the more mechanism-specific assessment. Present-day transfer is recorded as reach and is not treated as proof of historical origin.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
The sharp line between a Rights Constraint and a Minimum Service Guarantee or Safety Floor is revisability. A floor is a number you can, in principle, renegotiate, waive under a logged exception, or re-index. A right is defined by not being that: the moment it becomes waivable "just this once for a big enough gain," it has stopped being a rights constraint and become a soft threshold wearing its name. If a protection genuinely needs an exception path, it was a threshold all along — model it as one honestly rather than calling it a right and then breaching it.
[n1] Robert Nozick's notion of rights as side constraints — bounds on what one may do to others in pursuit of a goal, rather than goals to be maximized themselves. It captures why a rights constraint removes options instead of down-weighting them: the right constrains the action set, it doesn't enter the objective function. ↩