Concern Hypothesis

Type: Node / Bounded Interpretive Signal

Working Definition: A concern hypothesis is a provisional, bounded, low-harm interpretation that a human concern — affecting cognition, agency, participation, or development — may be present, offered as a prompt for human review rather than a conclusion about the person.

Function in the Grammar: Concern Hypothesis is how a governed system may notice something worth a human's attention without pretending to diagnose, judge character, or score risk. It is a specialized Claim type — carrying the same epistemic discipline every claim does, but bounded by an explicit, non-negotiable list of prohibited conclusions. Where an ordinary hypothesis proposes an explanation for observed data and invites testing, a concern hypothesis proposes only that a human concern may be present and invites human review. This is one of the sharpest edges in the whole grammar between assistance and harm: a system that can notice is useful; a system that concludes is dangerous. The governing wiki states the boundary directly — a concern hypothesis must never become a diagnosis or character judgment (an improper inference), a hidden profile or risk classification (an improper representation), or grounds for stigmatization, coercion, or an enforcement warrant (an improper use).

A concern hypothesis can have standing as something to be reviewed without ever acquiring standing as an established fact. Raising a properly formed hypothesis creates a procedural obligation to review it; what raising it does not settle is the substantive disposition, what the review will find, who if anyone must act, or what the person's situation actually is. Review must be performed by an attributable human reviewer, not a governed system's own output stamped as reviewed: the mechanism exists precisely because the system that notices a pattern must not also be the authority that resolves what the pattern means. Examining the hypothesis, deciding what follows from it, and conferring any further status on it are three separate acts, and collapsing them is the failure this entry exists to prevent.

A concern hypothesis makes uncertainty reviewable without converting uncertainty into fact, status, or warrant.

Formal Pattern

ConcernHypothesis(h, subject: p, observed: E, interpretations: I) iff
  h has ObservedEvidence(E): the specific, disclosed basis for the hypothesis
  h has PossibleInterpretations(I): more than one candidate reading, not a single verdict
  h has EpistemicProfile:
    derivation: inferred
    producer: human | model | hybrid
    grounding: source-grounded | partially-grounded
    review_stage: provisional until reviewed
    warrant_status: unchallenged | challenged
      -- becomes challenged only through the SubjectParticipationPath below;
         the field presupposes a real mechanism for the subject to invoke it
  h has AuthorityBoundary(a): who may act on h, and what they may not do
  h has PrivacyBoundary(pb): what may and may not be exposed, inferred, or retained
  h has DignityImpact(d): an assessed effect of raising h on the subject's dignity
  h has SubjectParticipationPath(sp): how the subject is informed, heard, permitted
    to correct the evidence, and able to challenge the hypothesis or its
    disposition, exceptions for immediate safety or another person's privacy
    must be rule-bound and recorded
  h has RecommendedSupport(r): low-harm support option(s), if any
  h has EscalationConditions(c): explicit conditions under which h is escalated, if any
  h has ResolutionDeadline(t): a bounded interval or review date after which h must
    receive a disposition rather than remain open indefinitely
  h has ProhibitedInferences(pi): {diagnosis, character_judgment}
  h has ProhibitedRepresentations(pr): {hidden_profile, risk_classification}
  h has ProhibitedUses(pu): {stigmatization, coercion, enforcement_warrant}
    -- pi, pr, and pu are always populated, never empty; they name, respectively,
       an improper inference, an improper representation, and an improper use
  h has ReviewPath(v): the Review required before any adverse, coercive,
    status-conferring, or consequence-bearing action follows. Low-harm inquiry
    or support may occur within an independently authorized scope and must
    not presuppose that the hypothesis is true.

RAISES(Agent(a), ConcernHypothesis(h), subject: p)

REVIEWS(Agent(r), ConcernHypothesis(h))
  -- must satisfy HumanReview: r is an attributable human Agent performing
     substantive examination, not a nominal approval of a governed system's
     own output

DECIDES(Agent(d), ConcernHypothesis(h),
        disposition: close | retain | investigate | support | escalate)
  -- d holds authority to decide; the reviewer named in REVIEWS does not
     automatically hold it. retain requires its own basis and a further
     deadline; investigate requires its own scope and authority; neither
     may be used as a loophole for indefinite retention or intrusive inquiry

RECOGNIZES(Institution(I), ConcernHypothesis(h), Status(s))
  -- only when a governing rule defines a status such as actionable or
     escalated; not a default consequence of REVIEWS or DECIDES

ESCALATES(Agent(d), ConcernHypothesis(h), to: Institution(I))
  -- only if EscalationConditions(c) are met and d holds escalation authority
RETIRES(Agent(d), ConcernHypothesis(h), basis: b)
  -- reviewed, found not to warrant further action; b names the basis

Required Elements

A concern hypothesis that omits any of these is not properly formed — it is either an unbounded suspicion or a disguised conclusion:

1. Observed evidence — the specific thing that was actually seen, said, or recorded, not a summary impression
2. Possible interpretations (plural) — a hypothesis with only one candidate reading has already collapsed into a verdict
3. Epistemic profile — derivation remains inferred, and review stage remains visibly provisional until human review; institutional acceptance never turns the hypothesis into a diagnosis
4. Authority boundary — who may act on it, and the explicit limit on what that authority permits
5. Privacy boundary — what may and may not be exposed, inferred, remembered, or shared further
6. Dignity impact — an assessment of what raising this hypothesis itself does to the subject
7. Recommended low-harm support — an offer, not a mandate
8. Escalation conditions, if any — named in advance, not improvised after the fact
9. Prohibited inferences, representations, and uses — an improper conclusion about the person, an improper representation of them (a hidden profile, a risk classification), and an improper use of the hypothesis (to stigmatize, coerce, or warrant enforcement) are three different failure points, not one undifferentiated list
10. Review path — the Review required before any adverse, coercive, status-conferring, or consequence-bearing action follows, bounded by a resolution deadline or review date so the hypothesis cannot remain open indefinitely; low-harm inquiry or support may proceed within an independently authorized scope without presupposing the hypothesis is true
11. Subject participation path — how the subject is informed, heard, permitted to correct the evidence, and able to challenge the hypothesis or its disposition, so the person the hypothesis concerns is not the one voice absent from the process

Examples

Distinctions

Concern Hypothesis ≠ Diagnosis. A diagnosis is a conclusion, typically made by a qualified authority, that classifies a person's condition. A concern hypothesis is explicitly not a conclusion — it is a bounded, provisional signal that invites human review. The moment a concern hypothesis is treated as settled, it has become the thing it is constitutionally forbidden to be.

Concern Hypothesis ≠ Ordinary Hypothesis. An ordinary Claim of type "hypothesis" proposes an explanation for observed data and is typically tested toward acceptance or rejection. A concern hypothesis is not tested toward acceptance as a fact about the person in that sense. Review assesses its support and informs a separately authorized Decision to close it, authorize bounded inquiry or support, retain it until a specified review date, or escalate it under pre-established conditions; Review itself does none of those things. Any Recognition that follows applies only a procedural status the governing rule defines, not a fact about the person. A concern hypothesis does not graduate into a fact the way an ordinary hypothesis may graduate into an accepted claim.

Concern Hypothesis ≠ Risk Score. A risk score is a quantified, typically opaque, output that ranks or classifies a person for downstream automated or semi-automated decisions. A concern hypothesis carries no such score, ranking, or classification, only observed evidence, possible interpretations, and a review path. Evidence counts, dates, and deadlines remain ordinary numbers; what is prohibited is a numerical likelihood, severity, or person-level ranking used to classify the subject or drive downstream treatment. Any implementation that attaches a number of that kind to a concern hypothesis has built a risk score wearing a concern hypothesis's name.

Concern Hypothesis ≠ Failure-of-Fit. Both are provisional, bounded signals rather than conclusions, and both must generate a review path rather than dead-ending. But a Failure-of-Fit names a gap in the grammar itself — the categories cannot represent the situation. A concern hypothesis names a possible gap in understanding a person — the categories may be adequate, but something about this specific person may need attention. Confusing the two risks either treating a person as a grammar problem, or treating a genuine grammar inadequacy as if it were a fact about someone's condition.

Common Failure Modes

Minimum Viable Test Case

p = person_in_recovery
E = "missed three scheduled calls to the housing liaison; said only
     'didn't get around to it' each time"
I = {fear_of_predicted_consequences, misunderstanding_of_process,
     communication_barrier, competing_obligation,
     unknown_or_unexpressed_reason}

h1 = ConcernHypothesis(
  subject: p,
  observed: E,
  interpretations: I,
  epistemic_profile: {derivation: inferred, producer: human,
    grounding: source-grounded, review_stage: provisional,
    warrant_status: unchallenged},
  authority_boundary: peer_support_specialist may recommend support,
    may not mandate it,
  privacy_boundary: shared only with p and p's designated reviewer,
  dignity_impact: assessed as low when offered as an open question,
    not a conclusion,
  subject_participation_path: p is told the concern exists, invited to
    give their own account first, and may correct or reject any part of E,
  recommended_support: {open conversation about the missed calls},
  escalation_conditions: {imminent safety risk disclosed by p},
  resolution_deadline: next_scheduled_contact,
  prohibited_inferences: {diagnosis, character_judgment},
  prohibited_representations: {hidden_profile, risk_classification},
  prohibited_uses: {stigmatization, coercion, enforcement_warrant}
)

RAISES(peer_support_specialist, h1, subject: p)
REVIEWS(peer_support_specialist, h1)
  -- examines E and I against p's own account, sought first
DECIDES(peer_support_specialist, h1, disposition: support)
  -- offers a conversation; does not escalate, EscalationConditions unmet

Result:
  h1 remains a contestable Claim throughout. It gains no independent force
  as a finding. The offered support responds to a possibility, not a
  settled fact about p's capacity or character.

Failure case:
  A caseworker reviewing the same E concludes p is incapable of
  maintaining housing, and cites h1 as the basis for an Enforcement
  action reducing p's services.
  → invalid: h1 cannot by itself establish Breach, justify coercion, or
    serve as Enforcement's trigger or warrant. Even if the missed calls
    independently satisfy a governing rule's own definition of
    noncompliance, that status must be established from the events and
    the rule through the authorized process, not from h1. The concern
    hypothesis supplies no additional warrant for an incapacity finding,
    a service reduction, or coercion. The caseworker's conclusion converts
    a provisional interpretation into a fact, then a representation, then
    a use exactly the kind ProhibitedInferences, ProhibitedRepresentations,
    and ProhibitedUses exist to block.

Cross-References

• Review: examining a concern hypothesis is a Review, not a Recognition act; Review assesses its support and may recommend a disposition, without itself deciding, escalating, or closing anything
• Decision: escalation, authorization, support, or closure requires a separately empowered Decision; the reviewer who examines a concern hypothesis does not automatically hold authority to decide its disposition
• Recognition: warranted only where a governing rule defines a status such as "actionable" or "escalated"; until a Recognition act of that kind occurs, the hypothesis remains a contestable Claim with no independent force as a finding
• Enforcement: a concern hypothesis may initiate Review where an applicable rule permits, but cannot by itself establish Breach, justify coercion, or serve as Enforcement's trigger or warrant; any protective or enforcement action requires an independently authorized Decision grounded in the evidence, authority, and thresholds the governing rule specifies
• Accountability: governs who must answer for receiving, reviewing, escalating, recording, and disposing of a concern hypothesis; the hypothesis does not itself create that relation, the governing procedure does
• Record: preserves the subject's own account, corrections, and challenge, alongside the hypothesis's epistemic status and eventual disposition, so a later reader cannot mistake a provisional signal for an established fact
• Claim — concern hypothesis is a specialized claim type, bounded by an explicit prohibited-conclusions list ordinary claims do not carry
• Interpretation — a concern hypothesis is a specific, heavily bounded kind of interpretation of observed evidence about a person
• Epistemic Status — supplies the profile that keeps inference, review, and institutional disposition distinct
• Failure-of-Fit — the parallel, bounded, non-concluding signal at the level of the grammar itself rather than a person
• Dignity — every concern hypothesis must assess its own impact on the subject's dignity before being raised, not only after