At the end of Chapter 21, the Riverside Consortium wrote three things into the closing of the annual review affair, and the annual comparison report for the first time set the old judgment of "low demand" and the actual ridership that followed side by side on one and the same page. This layout is the starting point of the present chapter: the two figures were not summed into a single index; they stood next to each other, letting the gap become visible of its own accord. Yet what stands firm in a report begins to sway in the meeting rooms where decisions are made — the budget meeting accepts only one page of conclusions, the assessment form leaves a single line of overall rating, and the question "is it running well" by its very nature exacts a unique answer. What this chapter asks is precisely the institutional obverse of what Chapter 7 criticized: once the single composite score has been dismantled, by what right can multiple value scales coexist without being refolded into one number at the next round of aggregation.
The answer cannot be left to demolition to complete by itself. Chapter 7 showed how proxy indicators of different tasks were folded into a single eligibility; this chapter states the positive conditions: the different scales are each measured independently, conversion is prohibited on defined occasions and exists on the remaining ones as a signed decision, and conflicts have an adjudication procedure that returns them to the table. At the same time, a harsher possibility must be faced directly: the multiple scales can degenerate into the arbitrariness of supervisor preference — favoritism practiced under the name of plurality — or into performative compliance, where every value has its indicator and no one answers for the whole. And the coordinating value of unified standards genuinely exists; plurality is not a license to escape comparison — this is the other boundary the chapter must hold at the same time.
The Question of Scale after the Composite Score Exits
After the affair, the Consortium abolished the composite assessment score that had been in internal use for years. The first quarter's results were unexpected: the records of the various sections became more scattered still. Efficiency reports were kept by shift, care records by individual case, maintenance logs by piece of equipment; when aggregated, the three kinds of material could find no common row. What the abolition of the composite score dismantled was not only a number but also the format of aggregation it had supplied in passing; multiple scales do not grow of themselves out of the ruins — they require an institutional carrier of their own.
More dangerous is the power effect of the interregnum. Where there is no public scale, evaluation does not disappear; it merely retreats into private judgment: the supervisor says at the meeting that this quarter care matters more, next quarter that efficiency matters more, and every switch is accompanied by an ample narrative. The unified score that Chapter 7 criticized was at least a visible coercion; scale-less discretion is an invisible coercion — those evaluated cannot even know by what standard they are being measured, and can therefore neither prepare nor rebut.
From this follows the chapter's first condition: the coexistence of multiple scales is an institutional achievement, not a natural state. It requires that each scale have its own manner of recording, its own caliber of aggregation, and its own position of responsibility, just as Chapter 19 required every standard to carry its reason of formation and its term. The drift of an institution toward a single scale needs no one to push it — the convenience of aggregation always stands on that side; preserving plurality is therefore a continuing opposition, not a one-time demolition.
Scales Answer Questions; Indicators Are Only Proxies
Scale and indicator must first be distinguished. A scale is a question: in what sense is the ferry service doing well — fast, economical, steady, or responsible toward those who depend on it? An indicator is a partial proxy for that question: the on-time rate answers fast, cost per crossing answers economical, the record of delayed-departure consequences answers one part of responsible. What the Consortium once did was to merge four reports into one score; the error lay in none of the reports but in letting one proxy impersonate the entire question.
Chapter 7 showed how proxy indicators expand from local records into overall eligibility; this chapter institutionalizes that judgment as the boundary of measurement: every indicator registers the question it answers, the conditions of its generation, and the term of its expiry, and use beyond that boundary is in itself an appealable matter. The delayed-departure consequence record, used in a care assessment, is relevant material; turned to judging whether a particular caregiver is personally reliable, it crosses beyond its original object — this and the fact that Tang's missed sessions ought not to affect eligibility for the equipment are executions of one and the same boundary on different occasions.
Nor are multiple scales exempt from the temporal analysis of Chapter 19. The meaning of the quality of care changes with demographic structure, hospital scheduling, and the capacity of the crossing; the four reports designed today may, five years hence, equally degenerate into echoes on a chain of citation, report citing report, with no one revisiting the site. The indicator list of every scale should likewise state its review triggers: at what degree of change in the environmental variables, and on whose proposal, what counts as good care is to be redefined. Scales are more wear-resistant than composite scores, but they are not immune to time; a scale that never revisits its own question, like a standard that is never reviewed, passes itself off as correct by the mere fact of existing.
Parallel Scales and the Conditions of Conversion
The Riverside Consortium's assessment cannot retain only a composite score. The table below places one and the same ferry decision under different questions; it displays the manner of recording in a fictional setting and provides no universal weights.
| Scale | Corresponding material | Conflicts the same decision may produce |
|---|---|---|
| Operating efficiency | Utilization rate, cost per crossing | Moving the last departure earlier may reduce empty runs, yet leave night-shift workers missing their connections |
| Care and accessibility | Waiting, interrupted pickups, round trips for medical care | Adding runs may improve access while increasing current expenditure |
| Long-term maintenance | Upkeep, replacement, retention of skills | Current cutbacks may save expense while deferring the maintenance burden to later |
| Baselines that must not be silently offset | Safety requirements, impacts on basic livelihood, and whether they were touched | Surpluses in the other columns cannot automatically cancel the account of a baseline loss |
The first three columns likewise cannot be summed at will: their questions, units, and time spans differ. The fourth column in particular must not be left blank merely because no report has arrived; it should be marked as checked, as yet unconfirmed, or as touched. The absence of a record is not proof that the baseline went undamaged. The decision-maker may still face a trade-off, but must state what is being forgone, on what basis, and who bears the consequences, rather than hiding the trade-off inside a composite score.
Occasions Where Conversion Is Prohibited
Chapter 7 distinguished compensable from non-compensable differences: a failure on safety cannot be offset by high attendance. The institutionalization of this distinction is the list of prohibited conversions. The safety baseline, bodily integrity, and eligibility for the basic means of subsistence belong among the items into which no conversion formula may enter — not because they rank highest in an ordering, but because they carry a veto function: once they can be traded away by a surplus in another column, the veto is gone, and all the surpluses added together cannot buy back a baseline that has been traded away.
The care and efficiency columns can lock horns with each other, because both are values of degree; subsidy eligibility is different. Folding the coverage of subsidies into operating costs for a cost-benefit calculation amounts to writing some people's survival down as an optimizable item. Chapter 9 treated the structure in which evaluation is bound to chances of survival; the rule of this chapter stands a step earlier: any value whose consequences touch basic opportunities is forbidden to enter cross-scale conversion — it may lose to a directly contrary decision, that is, an explicit cut with reasons stated and a signature borne; it may not lose to conversion, that is, be quietly offset inside the gains of another column. The former is contestable politics; the latter is invisible arithmetic.
The list cannot expand without limit, otherwise every value will claim to be inconvertible, the multiple scales will degenerate into the incalculable, and in the end a composite score will have to be called back in to adjudicate. There are three criteria: whether the consequences are irreversible; whether the value carries a veto function; whether the loss can be compensated by institutional redundancy. A value answering no to all three belongs among the convertible; touching any one of them places it on the list, together with a statement of who has the authority to propose additions and deletions and by what procedure the list is reviewed — the prohibition list is itself a standard in the sense of Chapter 19: with reasons, with a term, appealable.
Criteria for Permissible Conversion
Prohibiting conversion does not mean conflicts disappear. There is only one budget, only one schedule, and the departure time of the last ferry can be only one. Chapter 7 already showed that one cannot evade decisions by pleading the plurality of all values; the legitimate occasion of conversion lies precisely here: when two convertible values contend for one and the same resource and the decision cannot be avoided, placing the two into a common unit for comparison is the beginning of honesty, not the beginning of degradation. The question has never been whether to convert, but in what guise the conversion exists.
Legitimate conversion must satisfy five conditions at once: the conversion takes place within one and the same context of decision, for this one timetable, not for the overall ranking of the two values; the conversion rate is published in advance or on the spot, with the party advocating the conversion stating its reasons; the conversion touches nothing on the prohibition list; the side converted away retains its place for appeal and after-the-fact comparison; the conversion decision is signed, and its consequences return to the table under the ledger of Chapter 21. If any one of the five is missing, the conversion degenerates from a decision into a weight that takes effect in the dark — the value decision hidden inside the word "composite" that Chapter 7 criticized, returned in a change of clothes.
The criteria should not frighten away all simplification. High-frequency, low-consequence, reversible everyday decisions deserve rough conversion: when an extra ferry is added on short notice, the dispatcher weighs in his head, by experience, ten minutes of added waiting against one more tank of fuel, and no one demands that he write a signature for one act of weighing — Chapter 7 said that the lighter, the more reversible, and the less cross-domain the consequences, the more legitimate the automatic and the rough. The true point of application of the criteria lies on the steps of escalation: when the same kind of weighing begins to decide long-term scheduling plans, the retention or dismissal of personnel, the survival or abolition of subsidies, it has crossed the threshold of the five criteria and must pass from experience into a signed decision. The danger has never lain in the dispatcher's weighing, but in the weighing's being upgraded into an institution while keeping the weighing's immunity.
This is the core distinction of the chapter: converting one delayed departure into so much cost is not measurement but decision. The facts measurement can establish are that a departure was delayed and that the caregiver was twenty minutes late; folding twenty minutes into so much money or so many points embeds a trade-off at every step. Conversion in the guise of measurement is automatic, neutral, and accountable to no one; conversion in the guise of decision requires an author, reasons, and a term. The success or failure of a multi-scale institution ultimately depends on whether it can stop every conversion from passing itself off as measurement.
Value Decisions and Hidden Conversion
Chapter 7 asked who decides how many missed sessions one late return equals; this chapter pushes the same question toward the power to set conversion rates. After Shen's transfer, if the data team that took over proposes to fold passenger waiting into cost accounting by some coefficient, that coefficient must run like the version number of the annual review rule: carrying its signature and reason of formation, deliverable by consequences, overridable by review. The setter need not be a collective, but must be identifiable — in a kingdom of anonymous exchange rates, every trade-off passes itself off as arithmetic.
Chapter 21 showed that decision-making power unaffected by consequences is a privilege to externalize cost; the conversion rate is its miniature form. Once a coefficient that converts care waiting to zero is running, the savings accrue to the reports and the waiting to the Tangs. The direction of repair is isomorphic with the return of consequences to the table: the conversion rate enters the ledger of consequences, the misjudgments and exclusions of decisions made under the coefficient are archived by coefficient version, and the comparison materials return periodically to the setter's desk. A setter who never receives the consequences of his own coefficient and a decision-maker who never receives the consequences of his rules are, in the mechanical sense, one and the same position.
Conversion rates expire more easily than ordinary standards, because what they encode is the relation between two columns of values at a given moment: the cost of waiting under a young population differs from that after aging; the cost of empty runs under tight capacity differs from that when capacity is plentiful. When a coefficient is set, it should at the same time state to which environmental variables it is tied and at what degree of change it must be revalued. Without this clause, the conversion rate becomes the most obstinate metaphysics — a trade-off of a particular year passing itself off as the eternal exchange rate between two values.
Where Hidden Conversion Resides
The most dangerous conversion never calls itself conversion. When the efficiency report enters both the savings from cut runs and the interruptions from delayed departures into a single money column, the conversion has already been completed although no one ever set an exchange rate — the interruptions are booked at zero. The entry point for identifying hidden conversion is the audit of units: behind every report that places things of different units into one and the same column there must be a conversion rate; ask for its signature, and a conversion rate unable to name its author is precisely a weight hidden inside the system, as Chapter 7 called it.
Chapter 7 showed how treating the missing as zero drove care out of visibility; in multi-scale aggregation this is the commonest hidden conversion: the unquantifiable part of the care column is simply given no indicator, whereupon it automatically contributes zero to every aggregation, and every gain in the efficiency column appears as a net increase. The institutional countermeasure is not to force quantification of the fourth column but to register the unmeasurable — the summary page reserves, for values that cannot be measured, a row stating whether the present decision touched them, so that absence at least no longer impersonates zero.
The threshold for charging a report with hidden conversion resembles the layered appeals of Chapter 20: one must point to the specific table and column where the conversion occurs, the value converted, and the evidence that the exchange rate is missing. The structural defense need not wait for an individual case to stand: the norm governing summary materials should itself require that wherever columns are totaled across, a list of the conversion rates and their signatures be attached. A total without such a list is presumed by default to be an unfinished report, not a neutrality that has nothing to declare.
Plural Discretion and Responsibility for the Whole
The multi-scale institution legitimizes a new power: the power to choose which scale applies on this occasion. A supervisor may say that this assessment stresses efficiency and the next stresses care, and each time with the endorsement of the plural institution. The unified score of Chapter 7 was at least uniform for everyone; manipulated plurality is more convenient for the strong — whoever commands the switching of scales commands an evaluation power unbound by any expectation of capacity, and plurality becomes instead the most presentable rhetoric of favoritism.
The beneficiaries are the management positions that hold the switch, together with the strong sections adept at following the switches: they can always deliver their results in the cycles when a scale favorable to them is in force. The cost is borne by the weak who depend on predictable evaluation. Harsh rules can at least be planned around; unpredictable rules leave one without even an object to plan: a caregiver cannot know whether this quarter her delayed departure will be recorded as care attrition or as personal unreliability, and this uncertainty is itself already a form of domination — it places the weather of evaluation in someone else's hands.
The counterexample genuinely exists: changing circumstances do require switching scales. Safety first in flood season, maintenance first in deep winter — this is not favoritism but environment; a multi-scale institution that forbade all switching would not survive its first flood season. The criteria concern the manner of switching, not switching itself: trigger conditions written down in advance — the trigger list of Chapter 19; reasons recorded at the switch; comparison accepted after the switch — the evaluators evaluated, in the manner of Chapter 21. With these three, a switch is a scale's response to its environment; without them, a switch is a preference wearing the weather as rhetoric. Switches that repeatedly favor the same side, however contextual the narrative, should be registered as anomalies.
Indicator Proliferation and the Unowned Whole
The second degeneration runs in the opposite direction: not too little plurality but too much. The Consortium could set up six indicator sets — efficiency, care, maintenance, safety, participation, transparency — each with its full complement of quarterly colors, while every actual decision is still made by the old single logic. The multi-scale report here becomes a new form of the decorative display analyzed in Chapter 20: every value has a place to perform, and no value has a place in decision. Indicator proliferation is precisely the most presentable face of plurality's failure.
There are three marks: the deterioration of any scale's indicators never changes any decision; conflicts between scales never enter the agenda — the reports merely stand side by side without speaking to one another; there is no connection between the reports and the decision ledger of Chapter 21, so that one cannot say which aggregation ever triggered which review. When the three coexist, the multiple scales have degenerated from an instrument of evaluation into a compliance document; their product is the appearance of having considered everything, and their cost is the attention that the values truly worth considering never receive.
The repair reuses the same criterion Chapter 20 laid down for appeals: the standard of standing is the capacity to change something. Every scale connects to at least one decision consequence — sustained deterioration in the care column should be able to trigger a scheduling review, a red line in the maintenance column should be able to freeze a cutback proposal — failing which the scale should be deregistered and its resources ceded to the occasions where it truly works. This is not a demand for quarterly overhauls but a demand for an identifiable transmission between indicators and decisions; an indicator that never transmits and an appeal that was never heard are two sets of the same ritual.
Comparison, Adjudication, and Repair
The truth in what is being criticized must be conceded: unified standards carry a genuine coordinating value. The consequence ledger of Chapter 21 needs units — if the caliber of "how many misjudgments" differs each time, the ledger degenerates into the decoration this chapter criticizes; comparability across years and across crews is the material foundation of accountability and learning. But what the institution needs is comparability within scales, not commensurability between scales. The same scale comparable vertically, units of the same kind comparable horizontally, incommensurability acknowledged between different scales — this distinction between the pair of concepts lets the respective goods of unity and plurality coexist.
Chapter 7 already proposed that multiple scales do not mean each going its own way; beyond conversion, a layer of common rules is needed: identity verification, data correction, appeal deadlines, record migration. This layer must be unified, because it is the operating foundation of the appeal system of Chapter 20 and the ledger system of Chapter 21 — for appeals to be able to identify decisions, and for the ledger to deliver consequences, both depend on a cross-scale norm of registration; if every scale had its own correction procedure, Tang would have to prove one record three times in three separate places, and the multiple scales would become a triple burden. Unity that halts at this layer serves plurality; unity that crosses this layer and enters the interior of the value columns begins to swallow.
Conversely, the plural side must also be constrained: incommensurability cannot become a shield against all evaluation. That the quality of care is hard to quantify does not exempt care work from examination — it still requires material of the Chapter 21 kind: comparison between conclusions and later facts, a place for the evaluated to submit counterexamples, the rebuttability of evaluators. The values of the fourth column are themselves inconvertible, yet the manner of tending them remains comparable: whether this arrangement ever touched the baseline, whether the touching was seen, whether anything changed once it was seen. What plurality exempts is the exchange rate, not the examination.
The Adjudication of Conflicts Returns to the Table
When efficiency and care lock horns, and maintenance contends with the quarterly budget, the handiest exit is a return to weighting — set a coefficient and let the answer compute itself. This is precisely the restoration, inside the multi-scale institution, of the mechanism Chapter 7 criticized. The value of conflict lies exactly in its having to be seen: a schedule that moves the last departure ten minutes earlier should exist as an explicit decision trading care for efficiency, not vanish as the side effect of a composite score rising by two points. A weight that dissolves conflict dissolves responsibility.
Adjudication needs no new organ, only the confluence of existing institutions: decision signatures, reasons and terms, the place of the affected parties, consequences returning to the table — drawn respectively from Chapters 21, 19, and 20. When the scheduling meeting makes its trade-off, it writes: this decision converts so much care, on such a coefficient, with comparison next quarter. The procedure does not guarantee that the trade-off is correct; it guarantees that the trade-off is visible, arguable, and regrettable. A visible and arguable error at least leaves the next adjudication a piece of comparison material; an automatically computed error claimed by no one lacks even the qualification to be remembered.
The output of adjudication is a decision for this context, not a permanent ranking of values. Sacrificing the empty-run rate this once for the sake of medical pickups establishes no theorem that care forever precedes efficiency; drawing on the care budget next time to keep up the ferry's overhaul overturns the previous decision not at all. The contextualization of value orderings is honesty; their permanentization is power disguised as measurement. What alone needs guarding against is one-directionality: if every adjudication tilts toward the same column, the comparison materials of Chapter 21 will develop it into a de facto single scale — a composite score dressed in plural clothing.
The Institutional Error-Correction Conditions of Coexisting Scales and the Boundary of This Chapter
Whether the multiple scales preserve difference or enlarge discretion can be checked against three criteria. First, independent measurement: every scale registers the question it answers, its units, and its term; proxies do not cross their bounds; the missing is not booked as zero. Second, conversion as decision: conversions outside the prohibition list must be same-context, with published rates, signed and answerable, their consequences returning to the table; a rate unable to name its author is treated as hidden conversion. Third, the existence of transmission: every scale connects to at least one decision consequence, and indicators that never transmit are deregistered. The three are criteria, not a design manual — the concrete forms satisfying them may vary endlessly, and any multi-scale institution that violates any one of them will slide toward one of the two poles, discretionary favoritism or indicator performance.
At the same time, what this chapter does not do must be stated. It offers no design of indicator systems — which indicators to choose, what weights to set, how many grades to divide belongs to the concrete decisions of each occasion, and written out as a universal scheme it would only recreate a composite score. It does not claim that plurality is automatically fairer: multiple scales raise the cost of judgment, manufacture inconsistency between sections, and open new room for discretion — these costs are real; the whole argument of the chapter is only that the costs of a single scale are more hidden and, moreover, unappealable. Nor does it resolve value conflict itself — philosophy cannot bear the responsibility for any single trade-off; it can only refuse to let the trade-off pass itself off as calculation.
The last page of the Riverside Consortium's annual report now holds four columns: efficiency, care, maintenance, and whether the baselines were touched. The adjustment of the last-departure timetable carries a column in small print: this adjustment trades care for efficiency, coefficient 3.2, signed by the data team, term one year, the waiting of the Tangs flowing back through the delayed-departure records. No composite score, no index; the conclusion column holds four answers, each standing on its own. The boundary of this chapter lies here: what an institution can guarantee is that the figures of different values exist side by side, and that every act of mutual devouring leaves behind the author's name. Only in an institution with room for many scales does obedience keep the shape of judgment; an institution that knows only one kind of number trains not cooperation but blindness to everything that cannot be converted.