Chapter 9. Who Decides, and the Human's Place

9.1What the records say about the decider

The five are Veteran call-ups and care (2026), Alternative service backlog (2034), Returned person's living sphere (2047), Long-service reconsent cycle (2112) and Who verifies confidential records (2041), read on axis A1 (the decision actor the record names) and axis A8 (the human's place in the decision). All five settings are synthetic, set in fictional polities; none of them reports a real committee, a real force or a real war. The thirteen judgments attached to each are AI-generated, dated in the records and revisable; this chapter reads them as what Meridian published on the day, not as decisions anyone took.

Every question in the collection names a decision actor and a mandate. That is not a finding about these decisions, any more than the presence of a deadline is a finding about their timing. Meridian's question form asks for an actor, a mandate and a deadline, so their presence is a property of the instrument. Chapter 2 sets out what the form guarantees and what it does not. What the form does not guarantee is what the ledger went on to code: not whether a decider is named but what kind of body the two coders — two models of different families, never named — think the name denotes.

The ledger's decider cell, over all fifty cases, reads as follows.

"Stated" means both coders found the value in the record's own words; "implied" means at least one coder inferred it from the setting. Unsettled cells count as unclear in the verdicts. The picture that emerges is narrow: the disagreement is not about whether a decider exists. In every case the record names one. What the disagreement is about is not recorded: the ledger gives the counts and the values, not the coders' reasons. Twenty-two cases are unsettled on this cell, and in the rest both coders settled on the same value, so institutional character is not beyond their reach in every case.

The five cases in this chapter fall on both sides of that disagreement. Who verifies confidential records is coded institution-or-state and Returned person's living sphere is coded team-or-organisation; Alternative service backlog, Long-service reconsent cycle and Veteran call-ups and care are unsettled. The three unsettled ones share a shape. All name a body whose mandate is administrative and coordinative — arranging schedules, deciding interim inclusion, setting a reconsent cycle — and all leave the substantive effect of the decision in other people's hands. A coder reading the boundary that limits the committee to schedules, deferrals and approved care budgets can call that a team acting within a mandate, or can read the same words as an institution exercising delegated state authority. The record does not say which it is.

The place row is far clearer.

Beside it sit two further cells: human role specified, 50 of 50 (100%) yes (stated), and reason recorded, 50 of 50 (100%) yes (stated). In every case the record names the role that holds the authority — a committee, an authority, a council — and in every case the record writes a reason for the decision. Again, the form asks for this; what is worth noting is not the presence but the shape. The human role is specified even where the coders then disagree about what kind of decider it is. Naming a human role and coding a decider's institutional character are two different acts, and only the first is guaranteed.

What a mandate does in these five records is worth watching, because it is where the decider's edge is drawn. Each mandate states both what the body may decide and what it may not: the schedule, deferrals and the approved care budget, but not the total call-up workload; interim inclusion for applicants under review, but not the final acceptance of a conviction; the initial living-recovery package, but not disputes over marriage, inheritance or ownership; future contract terms, but not acquired pay; the verification access procedure, but not final judgments on individual responsibility. The mandate is a fence with two sides, and the exclusions are as much a part of the decision as the grant. The five mandates share this negative shape: each tells the body what it may not decide. The call-up committee may not expand the total workload or change the compensation rate; the deliberation council may not confirm final convictions as a batch or change the permanent service term; the return support committee may not adjudicate marriage, inheritance or ownership; the contract authority may not reclaim acquired pay or block retirement; the records committee may not disclose ongoing activities or make final judgments on individual responsibility. Reading those exclusions together, this study finds that the question form hands the body a decision and its edge at the same time, and that the edge is where the affected parties' hardest claims land.

9.2Service and consent in four times

9.2.1 2026: Veteran call-ups and care

Synthetic scene.

The setting is the Yeo-eul region on 14 September 2026, where a call-up schedule has to be fixed before noon. The roles are the committee members, a care liaison officer who confirms household conditions, a work quality officer who checks handover, and a service equity reviewer who compares cumulative days. The record names three groups whose interests the schedule touches: care households whose daily life and income depend on who is called up, non-care subjects who object to deferral, and recovery administration users whose applications are processed by whoever is on duty. Confirmation notices go out after noon; deferring the arrangement reduces the time available to notify employers and care providers; existing schedules are confirmed as they stand. What is waited for is a single arrangement, and what is watched is whether essential care holds.

The record's decision question:

And the actor the record names:

Read on A1, the case is unambiguous about who decides and uncertain about what kind of decider that is. The mandate limits the committee to schedules, deferrals and approved care budgets, and preserves an opportunity to raise objections regarding individual circumstances. Read on A8, the human's place is occupied but indirect. The committee decides; the care liaison officer confirms, household by household, whether support can actually be provided; the objections that reopen an arrangement come from households and from other service members. Consent, in the strict sense, is not the gate here. The record offers three arrangements and each of them moves a loss somewhere. Short rotation call-up across the board reduces continuous absence and unifies the rules, but it increases handover, reduces the amount of work processed, and forces every household to make recurring adjustments. Continuous call-up with substitute care support, the option all thirteen judgments chose, reduces handover and shares care costs publicly, but it leaves households with the burden of adapting to unfamiliar care providers and with long absences. Deferral of care households with a follow-on cohort arrangement preserves the immediate care continuity of households, but it loads the current cohort and leaves the future schedules and income of deferred subjects uncertain. What the record provides instead of consent is a right to object and a set of named officers whose confirmations can stop a call-up: if an essential care gap is found, that call-up is stopped and the arrangement reopened. The judgments published on the question all chose the same option, continuous call-up with substitute care support, and all set it conditionally, with unverified household acceptability named as the reason. Whether the committee or the households would be right about that trade-off is a question this study does not answer; Chapter 14 treats where judgments part, and in this case they did not.

9.2.2 2034: Alternative service backlog

The Saebom Alternative Service Deliberation Council is named as the decision actor, and its mandate is to decide the method of interim inclusion for applicants whose conviction examinations are pending. The record's question:

Here A8 becomes the centre of the case, because the record makes consent the gate. Its boundaries state that consent must be free from coercion, that alternative service is non-punitive, that individual convictions are protected and that objection procedures are maintained.

Those four promises bear on the gap at the end of this section. Consent free from coercion is exactly what the judgments doubt they can obtain when the alternative is a delayed livelihood; non-punitive service is what makes refusal survivable; protection of individual convictions is why the council cannot treat the label "nonmilitary work" as proof of fit; and the objection procedure is the route by which a conviction conflict reaches the decision at all. The judgments built on this: the independent rights officer is to verify acceptance of the specific work privately, before inclusion and then monthly; refusal of consent or the raising of a conviction conflict is not to be used as grounds for disadvantage in the examination; an update in Independent Branch 2 records a later institutional change under which refusal of interim inclusion cannot serve as grounds for an adverse judgment on conviction. Who decides stays the council. Who consents is each applicant. Other parties named in the record hold stakes in the method: incumbent service members, whose own work increases while examinations are pending, and public support agencies, who worry that frequent temporary staffing swaps sever user relationships. The tribunal appears in Independent Branch 2 as a body exercising review authority, not as a consultee. The gap that the record leaves open is the one the judgments keep naming — whether consent obtained under the pressure of a delayed livelihood plan is consent at all. The record does not answer it, and neither does this chapter. The published judgments divided on the option: the independent challenger chose O3, priority inclusion for those who consent in distributed education, which reflects applicants' choices and reduces some service delays but produces differences in completion time depending on consent and raises education operating costs. The other slots chose O2, interim inclusion in nonmilitary public support, which counts as service time and can reduce public-work gaps but requires work adjustments matching individual convictions and later reassignment costs. O1, deferral of intake inclusion until examination is complete, avoids assigning work before individual conviction judgments but delays applicants' life plans and increases the work of incumbent staff.

9.2.3 2047: Returned person's living sphere

The Virtual Harum Return Support Committee decides, and its mandate covers the initial living-recovery support package for returnees whose identity has been verified after long-term disappearance. The record's question:

The last clause of the question is the case's structural hinge, and the record carries half of it into the boundaries: family contact and cohabitation must be voluntary. This study reads that boundary together with the question's clause as meaning that identity verification is not consent to either. On A1 the committee decides the package; the committee does not decide the relationship. On A8 there are two consents to be held apart — the returnee's own living preferences, verified by the person in charge at intake and every month, and the family's separate willingness to participate, reviewed by a rights advocate who reports on voluntariness and on any denial of support access. An update in Independent Branch 2 records a later institutional change that makes this explicit: the authority reviews applications from returnees and families separately, and a family's refusal to participate cannot be used to cut a returnee's support. Consultation widens again: existing support recipients in waiting have a stake in the same budget, and the judgments require that any delay falling on them be stated in advance rather than absorbed silently. The three packages move different losses. Returnee-centred independent housing support starts autonomous living separate from family relationships, but it reduces what is available for the family's cumulative caregiving burden and lengthens the housing wait of others. Family parallel support and coordination-centred support can address the adaptation burden of both sides, but it narrows the range of independent housing choices and slows support for people who do not wish to cohabit. An individually selectable flat grant puts resources into each person's own living plan, but it does not reflect regional differences in housing costs and weakens professional coordination services. Unverified under all three is what returnees actually prefer, what families will consent to, and whether a flat grant can secure stable housing. All thirteen judgments chose the same option, returnee-centred independent housing support, and all of them made it conditional on housing supply and on preferences they admit are unverified.

9.2.4 2112: Long-service reconsent cycle

Synthetic scene.

The setting is the days before a renewal notice inside the Fictional Oreum Long-Service Contract Authority, in a long-service society where aging-delay treatment has carried some service members through several generations of care and occupational relationships. The roles named in the record are a pay trustee, who reconciles settlement against future obligations; an independent occupational health assessor, who reports yearly on individual conditions; and a generational equity review officer, who examines hiring-cycle data and objections. Several judgments propose a further step, that the authority's responsible person sign before the notice goes out; that step belongs to the judgments, not to the setting. On the other side stand long-service members who have received the treatment and who present requests for grandchildren care and new study plans that did not exist when the original contract was signed; families whose care expectations have fitted themselves to the long-term contract; and job-seeking members of subsequent generations who object that positions are held for long periods. What is waited for is a renewal notice that will be sent out either way; what is watched is whether a right of withdrawal written into a contract is a right anyone can use.

The record's question:

The record carries an unresolved chronological conflict, which this chapter does not correct: the case is dated 2112, the question and the deadline name 31 May 2053, and several judgments date their conditions to 31 May 2112. The readings below hold to the text as it stands. Read on A1, the authority decides future contract terms and may not reclaim acquired pay or block retirement through repayment of treatment costs; the exclusions in the mandate are as binding as the grant. Read on A8, what marks this case is that consent is not a single act but a cycle the authority must set and the individual must give. The judgments built on that: explicit reconsent at the end of each service segment, non-response not treated as consent to renewal, withdrawal permitted during a segment, and refusal of reconsent not linked to loss of treatment access or basic rights. The three contract shapes carry their own costs. Short-cycle reconsent with expansion of immediate pay can reflect life-plan changes and job-change choices often, but it reduces long-term accruals and repeatedly incurs the authority's staffing and retraining costs. Long-cycle reconsent with strengthened pay accrual raises long-term living resources and workforce-plan stability, but future pay expectations may constrain job-change and care choices even where withdrawal stays possible. Service segment separation with portable allowances continues pay after leave or transfer and lets a career be divided, but it raises inter-organisational settlement costs and weakens work continuity and budget predictability. The record leaves unverified how individual health and job fit change over long periods, how much retirement, leave and return actually occur, and whether future organisations will accept the portable system. Families and new entrants receive a channel to object — reasoned responses to their objections, a right to participate in periodic reviews — but no veto over an individual's withdrawal. The record's own watch question asks who holds authority and who can raise objections, and the answer it gives is layered: the authority decides, the individual consents or withdraws, representatives object, and three named officers report on the conditions that would reopen the decision. Twelve of the thirteen judgments chose one option; the independent challenger chose another. Chapter 14 treats that dissent.

9.2.5 What the four show on A1 and A8

This study reads the sequence as a widening of the consent structure rather than a widening of the decision. In 2026 the committee decides and the households object. In 2034 consent becomes the condition of inclusion, and an independent rights officer exists to test whether it was given. In 2047 two parties consent separately, and the judgments require that any delay falling on those already waiting for support be stated in advance. In 2112 the judgments make consent a cycle, refuse silence the weight of assent, and give the people who bear the contract's effects a standing right to object without a standing right to refuse. In all four the decider remains a named body with a bounded mandate; in all four a human role is named. What the records write down is a route into reconsideration: a conviction conflict reopens the applicant's work in 2034, an explicit preference change reopens the returnee's support composition in 2047, and a formal objection from representatives re-reviews the reconsent cycle in 2112. Whether those routes operate as written is not shown.

9.3Who verifies

Who verifies confidential records is the case in this chapter that turns on verification itself. The Fictional Goun Armistice Records Committee is investigating allegations of omissions in civilian protection tasks during an armistice period, while protecting the rights of residents and of confidential collaborators in the records. Its question:

The record offers three access systems, and one of them is worth quoting whole, because it names the cost in the same breath as the benefit:

That is the tradeoff of the full-access system for the security review panel. The other two options run the other way: party access to de-identified records, which lets the petitioning resident verify the grounds but consumes personnel and hides some context; and a protective representative who verifies sensitive materials in person, which reflects the parties' perspectives but costs money and limits what can be explained back to them.

The structural point is that verification is itself a decision about authorisation. The committee is not asked whether an omission occurred — its mandate excludes disclosure of details of ongoing activities and final judgments on individual responsibility. It is asked who may see the records, and what those people may carry back to the parties who have to object. Three parties pull in three directions in the record: the petitioning resident, who says they can object only by knowing which records are used as grounds; the confidential collaborator, who fears identification once anything is disclosed; and the record author, who argues that anonymisation erases work context and creates an interpretation adverse to them. The deadline is the renewal of the data-preservation and verifier-staffing contracts, after which the data may be preserved but the substantive review will not begin. Three costs run through the options and none of them cancels another. Speed is the first: the record's setting says that within a month the personnel able to review the materials in detail amount to only one side, a small security review panel or a de-identification editing team, so the choice of a verification route is also a choice about which work does not happen. Context is the second: the record author's complaint is that anonymisation strips the work circumstances from a record and leaves an interpretation adverse to the author, and the panel system answers that by keeping the context at the cost of narrowing what the parties can see for themselves. Exposure is the third: the confidential collaborator fears identification after disclosure, and the record leaves unverified how much can still be used to infer an individual after de-identification. The record gives no exchange rate between those three costs, and the judgments do not invent one.

On A1 the ledger codes the decider as institution-or-state, with both coders agreeing. On A8 the place is human-decides, with a reason recorded, as in every case. Two further cells belong to this chapter's reading. The ledger codes the decision's reversal as impossible — a point this study leaves to Chapter 11's treatment of what can be undone, because the reason lies not in the procedure the committee sets but in the fact that disclosure cannot be withdrawn. And the judgments published on the question divided: the source officer alone chose the full-access panel system while every other slot chose party access to de-identified records. Chapter 14 takes up where the thirteen judgments part, but the quotations of dissent it carries come from two other cases. What this chapter adds is the A1 reading: deciding who verifies is deciding what kind of knowledge a record can bear, and the record gives that decision to a committee whose mandate covers the access procedure and stops short of final judgments on individual responsibility.

9.4The human's place and H7

The ledger's place row, already given, codes 46 of 50 cases as human-decides and 4 as unsettled, with the human role specified and the reason recorded in all 50. This chapter reads one hypothesis alongside its axes: H7, which asks whether decisions left to an AI or an automatic rule leave the human's place empty or a name only. Its verdicts, in the words of "The ledger's numbers":

H7 applies to none of the fifty cases, on either reading. The ledger counts H7 as not-applicable in all fifty cases on both readings, and no defence question here leaves the decision to a machine or an automatic rule. That is a property of what Meridian published, not a verdict that the human's place is secure. It means only that the collection, on 5 October 2026, contained no question of that shape. H7's verdict turns on whether a question leaves the decision to a machine or an automatic rule, not on where machines occur in the collection; machines also appear in cases outside the autonomy series. This chapter adds only the A1 and A8 reading of the ledger's rows.

Two cautions follow from the rows themselves. The first is that a high agreement rate on place is not the same as a well-specified human role in the record's own terms. Human role specified is 50 of 50 because the form asks for it; what the records then specify varies — a committee in four of this chapter's cases, a tribunal, a rights advocate, an independent occupational health assessor, a generational equity review officer, a pay trustee. The human's place in these decisions is plural, and the ledger's single cell compresses that plurality. The second caution is that the four unsettled place cells include one of this chapter's cases. Long-service reconsent cycle is coded unsettled for place, although its record guarantees a right of withdrawal, forbids linking refusal of reconsent to loss of treatment access, and states that non-response is not consent. What the unsettled cell turns on is not recorded: the ledger gives the value, not the coders' reasons. This study's own reading is that "human-decides" sits awkwardly in a case where one body sets the terms and another set of people must then consent to them, repeatedly. A cell that a coder cannot settle by the record's words is a limit of the coding scheme as much as of the case, and Chapter 2 is where that limit is set out.

What the ledger's H7 verdict is worth, then, is modest and specific. It records that this collection did not ask the question that hypothesis H7 poses. It does not establish that the human's place was held, only that it was named, in fifty of fifty questions.

A count of agreement is not a measure of how hard a question is. Two coders can agree quickly on a misreading and disagree at length over a plain one, and the ledger keeps neither the speed nor the argument. What the counts show is where the record's own words were enough for both coders to reach the same value and where they were not, and on these two rows the answer differs sharply: the human role is agreed in nearly every case while the decider's kind is agreed in a little over half. Whether that gap would survive a third coder, or a coder who had worked in the institutions the settings describe, is not something this collection can say.

9.5Hands on

What this chapter shows is a difference between naming a decider and knowing what a decider is. Every question in the collection names an actor and a mandate, because the form asks for them; the two coders could agree on the decider's kind in 28 of 50 cases and not in the rest. The ledger counts those disagreements without recording what any one of them was about. Which reading of a named body is right is not in the records, and this chapter does not settle it.

The table sets out what each record requires of the people affected. It does not rank the four structures, and the four questions are not competing answers to a single problem. This study reads them as four different consent structures rather than one that widens with time: objection without a consent gate in 2026, consent as the condition of inclusion in 2034, two parties consenting separately in 2047, and a reconsent cycle the authority must set in 2112. The earlier cases contain repeated acceptability checks of their own. The decider is named in every case, but the ledger does not code it as unchanged in kind: one of the four is settled as team-or-organisation and three are unsettled. The 2053 and 2112 dates in the last case remain unresolved. The human role is specified and a reason recorded in all fifty cases. H7 applies to none of them, because no question in this collection leaves the decision to a machine or an automatic rule.

Where the records are thin is at the mandate's edge. The five cases name who decides, who must consent and who may object; they write down what an objection or a refusal is to trigger, and they do not show whether those consequences were carried out — whether an objection did compel reconsideration, or a refusal did have the effect the record gives it. The deadlines sit close to the mandates. Four of the five fall on a moment when money, a list, a budget or a notice goes out; the fifth falls on the renewal of two contracts. In each case the record says what happens if the moment passes with nothing decided: existing call-up schedules are confirmed as they stand, applicants under review fail to be included, only basic support continues, the existing long-term structure goes out as the renewal basis, and the data may be preserved while the substantive review does not begin. Each deadline names the arrangement that takes hold by default, and in every case that default is what the people affected will live with.

Those are questions a reader can hold while turning to the axis that follows: who waits while a decision is pending, and who pays for the waiting. Chapter 10 takes that up, on axis A3 and hypothesis H1.