PSDA.6 - Represent Values and Trade-Offs
Type: Method pattern body Status: Stable
Primary working result: a decision-usable value account: attributed concerns and objectives, their consequence measures, protected conditions, admitted trade-offs, unresolved disagreements, and limits for one named comparison.
PSDA.6:1 - Problem frame
Use this when a scorecard’s criteria came from a template, “importance” weights have no stated meaning, or the same benefit appears under several headings. Also use it when people agree that reliability, development, safety, or fairness matters but mean different consequences for different subjects, or when their consequence comparisons still need to become a justified numerical model.
For each relevant concern, identify who holds it or bears the consequence, the subject and conditions, and why it matters. Connect it to the consequence that should enter the comparison and state how it is represented. Establish whether trade-offs are permitted and on what grounds.
An objective expresses what is wanted. A measure describes a characteristic. A preference orders consequences, while a protected condition limits the permissible trade-offs. Use the governing basis for a duty and for any affected party’s rights. Establish whether a change or exception is permitted and, if so, who can authorize it. A preference or weight grants no such power.
Use a current, qualified value account directly when it answers the comparison. For a missing measure, use the relevant measurement practice. For a condition’s current force, use the applicable rule, decision or domain evidence. For a disputed requirement’s justification, use C.11.DUA. A later choice, if needed, belongs to its authorized holder.
PSDA.6:2 - Problem
Criteria can appear neutral while deciding the result. “Install more pumps” embeds an action in an objective. “Service quality” may count the same avoided outage again under reliability, satisfaction, and reputation. Equal weights may silently favor the concern represented by the largest number of criteria.
Numbers can conceal a second substitution. A budget amount, a probability, an ordinal judgement, and a protected condition become interchangeable entries in a spreadsheet. Normalizing each to zero–one does not establish comparable value. A mean of participants’ weights can erase a disagreement about who bears the consequence or what may be traded at all.
PSDA.6:3 - Forces
| Force | Tension |
|---|---|
| Completeness and usability | Material values must remain visible, while an enormous hierarchy burdens elicitation and repeats consequences. |
| Measurement and meaning | Observable measures help comparison, while easily measured proxies can displace the concern relevant to the comparison. |
| Trade-offs and protection | Some sacrifices are legitimate within the question; others require a separate authority or are outside permissible compensation. |
| Individual and collective values | A common comparison is useful, while unlike participants need not share one value function. |
| Precision and honest incompleteness | Explicit scales and preference questions can clarify a choice, while unsupported weights create false precision. |
PSDA.6:4 - Solution
Identify the concerns and objectives the comparison is meant to represent. Organize them around consequences. State the measurement and trade-off rules, their grounds and any unresolved preferences or protected conditions. If the task supplies hypothetical parameters, use them for the conditional comparison it asks for.
PSDA.6:4.1 - Recover whose concern changes the comparison
Identify the subject, affected people or Systems, configuration, horizon and receiving question. A compatible PSDA.12 account may establish attributed concerns, duties, interests, missing voices or limits of representation. Establish any value transformation or authority to combine preferences on its own grounds.
For each material concern, establish who expresses it or bears its consequences and under which conditions. Where representation or governing force depends on an assignment, role classification or authority relation, use evidence that the relation holds. A title such as resident, manager or expert identifies a possible contribution without establishing a mandate.
Use a qualified source for a needed concern or binding condition. Where one stays unavailable, state the gap. Distinguish a participant’s statement, the analyst’s interpretation, a stipulated analytical assumption and an established obligation. Non-response leaves the participant’s preference unknown unless other pertinent evidence establishes it.
PSDA.6:4.2 - Distinguish ends, means, and repeated consequences
Ask “Why does this matter here?” to identify the consequence sought. Ask “How might it be achieved?” to expose possible means. “Provide training” may be one means; reliable performance in representative later Work may be the relevant end. The distinction is local to the question: a process experience can itself be valued when the participants value it in its own right.
Organize objectives only as far as the structure helps. For a small decision, a few attributed statements may be enough. A hierarchy or means–ends map is useful when several levels or overlaps would otherwise be lost.
Trace apparent duplicates to the consequence represented. Count that consequence once. Service loss and its distribution can be distinct values even when their measures covary. Statistical independence of measures supplies no preference rule for adding their values.
Retain a material value, protected condition or distinction that matters to the requested comparison. Before simplifying, identify what would be lost and examine its effect on the ranking, explanation or ability to detect a protected loss. Difficulty of measurement can leave a value unresolved.
PSDA.6:4.3 - Give each measure a usable meaning
For every consequential measure, state the characteristic, subject or participant tuple, scale and unit where applicable, direction of preference, horizon, source, and missingness rule. Distinguish a direct measure from a proxy and say what the proxy cannot establish.
Use C.16 for measurement and A.19 where the inquiry needs a declared multi-characteristic space or reusable predicate. A.19 defines coordinate and scale meanings. Scoring, aggregation and selection need their respective rules.
Keep the raw consequences alongside the value transformation. An ordinal judgement supports ordering. Equal intervals, arithmetic averages and interpolation require additional grounds. For a numerical value function, state the consequence range, anchors, shape or interpolation and the basis for its use. Identify whether it comes from attributed comparisons, an applicable rule or assumptions stipulated for a conditional analysis.
PSDA.6:4.4 - Separate conditions from compensable objectives
Distinguish a binding condition from an aspiration and from an unresolved claim about a condition. Identify its subject, scope and source. A requested service minimum may express a value. An existing legal or safety requirement can govern whether compensation is permitted.
Exclude compensation for a failed binding condition. If its meaning or applicability is disputed, use an applicable governing interpretation. Where a new ruling is needed, seek it from the authorized holder or identify the outstanding question. A narrower comparison can remain useful while the whole decision stays blocked.
When the live question is whether the requirement itself should be retained or changed, use C.11.DUA §4.3 with the competent domain contribution. Identify the protected person or System and interest, unwanted outcome, governed quantity or activity, conditions and horizon. Distinguish background, attributable increment and total exposure where they affect the judgement. Examine the causal and measurement basis, uncertainty, threshold or precision choice, marginal protection, full burden and displaced protective work. Compare retention, tightening, revision, replacement or removal only where they are meaningful alternatives. A safety label does not supply that basis, and expense alone does not defeat a protective requirement.
For a requirement appraisal, return the conclusion or recommendation the evidence supports and any important limit. Identify the condition’s current force. If amendment is at issue, identify authority and timing. When no acceptable alternative or timely amendment is feasible, state the impasse. A later correction may leave an earlier opportunity lost. An ordinary value account can finish without a separate appraisal.
Some thresholds are preference choices rather than prohibitions: an aspiration level, budget preference, or tolerated inconvenience. Say which they are and whose judgement they express. A threshold crossing is not automatically a decision or permission to act.
PSDA.6:4.5 - Construct preferences from comparisons of consequences
For an account of a holder’s preferences, use attributable comparisons of concrete consequence profiles for the same subjects, horizon and conditions. Obtain a judgement from the holder if the needed comparison is missing and worth obtaining. Preserve whether one profile is no worse, strictly better, indifferent, or the comparison remains open. Keep the profiles and any stated reason. Use alternatives or intelligible hypothetical profiles within the declared ranges. If the task specifies a rule or hypothetical preference instead, report conclusions under that basis.
Sections 4.5.1–4.5.5 treat known consequences. For uncertain outcomes, §4.5.6 uses comparisons of lotteries, preserves the compatible utility functions and relates them to the consequence probabilities. Choose the form for the distinctions the task needs. Deterministic trade-offs alone leave the ranking of some uncertain alternatives open.
PSDA.6:4.5.1 - Choose and challenge the representation
Separate protected conditions under §4.4 before considering compensation. Identify what the representation must preserve and use the least burdensome form adequate for the comparison. The following situations call for different forms.
| Available judgement and intended use | Construction and return |
|---|---|
| Only directions of improvement and some pairwise comparisons are justified. | Keep those comparisons and coordinate dominance. Infer further comparisons only through an admitted rule, such as transitivity. Return the resulting partial relation and the pairs still open. |
| A priority or threshold is explicitly noncompensatory. | Write and apply that rule, including its order and tie treatment. For example, prefer every profile with waiting time at most two hours to every profile above two hours, then prefer lower cost within each class. Attribute the rule to a holder only on evidence of that judgement. A task can also stipulate it for a conditional comparison. A binding duty needs its own governing basis. |
| Improvements can compensate for one another, with constant value increments and a constant exchange over the stated ranges. | Construct the affine family in §4.5.2. Its coefficients express the admitted exchange, not the importance of criterion names. |
| Compensation is acceptable but marginal values may curve. | Test the additive assumption and use the finite-level construction in §4.5.3 when it fits. Preserve interactions through a suitable non-additive family or explicit rules when addition fails. A qualified outranking Method can instead supply a relation with incomparability; its thresholds and aggregation conditions must be supplied with it. |
For a proposed additive form, compare profiles that differ on a subset of criteria while holding the others equal. Use comparisons at materially different common levels of those other criteria, obtaining further judgements where the available evidence leaves the issue open. A reversal challenges preferential independence: the preference over that subset depends on its background. Ask also about complementarities, thresholds and refusals to compensate. With two criteria, unchanged directions of preference alone give little support for addition; ask about the exchanges themselves. Statistical independence of measurements answers neither question.
For an affine form, ask two further questions. Do equal physical increments have equal value at different positions on each scale? Does the compensating amount stay the same across starting profiles? Within a one-to-five-hour range, compare reducing two hours to one with reducing five hours to four, and repeat a cost-for-time exchange at a different cost level. A stated constant exchange is a substantive modelling assumption over that range. Finite elicitation can challenge it and check its local fit; it does not prove a universal preference law. Preserve that qualification when the assumption is accepted for use, and return the unresolved form choice when it is not.
Keep materially different holders’ accounts separate unless a legitimately governed aggregation or reconciliation Method supplies the joint relation. State that basis when using a collective result.
PSDA.6:4.5.2 - Construct a family for a constant exchange
Suppose lower cost c and lower waiting time t are preferred, both matter, and the holder accepts the affine conditions above on the named ranges. Represent loss by L(c,t) = a c + b t, with a > 0 and b > 0; lower loss is preferred. A common constant and positive scale do not change the ordering. Divide by a to obtain L_r(c,t) = c + r t, where r = b/a > 0. The unit of r is cost units per hour. Here it represents the holder’s admitted cost-for-time exchange over the stated range. A market price is a different quantity.
For each statement that profile (c_x,t_x) is no worse than (c_y,t_y), write c_x + r t_x <= c_y + r t_y. Use equality for stated indifference and a strict inequality for stated strict preference. Keep every statement on the same r and intersect its restrictions with r > 0. The resulting set R describes every admitted exchange in this form. A convenient point in that set can be used for a stated scenario. Attributing that particular exchange to the holder needs evidence of their judgement.
Ranges still matter when scales are normalized. If the cost range is 20–60 thousand and the waiting range is 1–5 hours, normalized losses are (c-20)/40 and (t-1)/4. A weighted sum with positive weights w_c + w_t = 1 represents the same exchange only when r = 10 w_t/w_c. The weights compare the two whole range changes. A different range changes that correspondence; a change in the holder’s marginal values can require a different function as well. Unit conversion alone must leave the represented comparisons unchanged.
PSDA.6:4.5.3 - Retain unknown marginal values when linearity is unwarranted
For a finite comparison with an admitted additive form, use the distinct consequence levels appearing in the reference profiles and live alternatives. Order each criterion’s levels from least to most preferred. Assign an unknown value contribution q_i(z) to each level z of criterion i. Set the least preferred contribution on each criterion to zero, require the contributions to be nondecreasing along that criterion, and set the sum of the most preferred contributions to one. This normalization requires at least one valued improvement across the represented ranges; it fixes an arbitrary origin and unit without fixing relative criterion values. If the holder is indifferent across every represented profile, return that relation directly. If the evidence leaves open whether any difference matters, retain that uncertainty.
Define U(x) = sum_i q_i(x_i), with higher value preferred. Each no-worse statement x versus y adds U(x) >= U(y); indifference adds equality; strict preference adds U(x) > U(y). Add elicited bounds or comparisons of value differences only with their own meaning and source. All rows constrain the same contributions. Solving them jointly supplies a family of additive value functions, rather than independently chosen intervals for each criterion. Weak monotonicity permits a flat contribution; require strict increase only where the holder’s judgement supports it.
This construction leaves unevaluated levels open. Add a new level with its monotonicity and preference constraints, or use an explicitly justified interpolation. Equal spacing on an ordinal performance scale is not equal spacing in value. A family that is feasible at the represented levels establishes compatibility there; use beyond them needs an applicable shape assumption or evidence of preferences at the additional levels.
The constraints are linear in the unknown contributions. Use direct algebra for a small family or a linear-programming solver for the finite system. When strict comparisons are present, a solver can replace each strict difference by a common positive slack epsilon and maximize that slack under the normalization: a feasible result with epsilon > 0 witnesses joint strictness. A fixed arbitrary margin can exclude compatible models. Preserve weak equalities and numerical error limits when interpreting a computation. MMP.10 and MMP.11 supply the general formulation, representation and obtaining-method work; the consequence meanings and preference assumptions supplied here govern this instance.
PSDA.6:4.5.4 - Distinguish an open answer from a failed construction
First check whether at least one model satisfies the joint constraints. For a small interval family, exhibit an admitted parameter; for a larger system, obtain a feasible witness or a justified infeasibility result. A solver timeout leaves this question unanswered.
If the family is empty, locate a conflicting subset of the elicited rows and form assumptions. Check transcription, units, profile meanings and attribution against the available sources. Seek clarification from the holder where the intended judgement remains uncertain. Two opposite weak comparisons may express indifference; they are not a contradiction. A strict preference for x over y together with y no worse than x under the same conditions conflicts with the claimed preference relation. Other failures can arise because the chosen form cannot express otherwise coherent judgements. Do not resolve either failure by silently deleting a row or returning the best fitting weights. Revise a judgement only with its holder, change the form with an explicit basis, or return the unresolved conflict. An approximate fit needs an admitted error interpretation before it can support advice.
For a nonempty family, distinguish what all compatible models support from what at least one supports:
xis necessarily no worse thanywhen every compatible model gives that weak preference.xis possibly no worse thanywhen at least one compatible model gives it.
Test these claims on the joint family. For the loss family, necessary preference requires L_r(x) <= L_r(y) for every r in R; a possible preference needs one such r. A proof over the range establishes necessity; one feasible witness establishes possibility and one contrary witness defeats necessity. Use MATH.20 for the required whole-domain bound. Separate strict preference from equality at a boundary. An empty family warrants neither kind of recommendation.
If the question asks for a best alternative, keep the quantifiers together. A necessary best is no worse than every rival in every compatible model. A possible best is no worse than every rival in one and the same compatible model. Separate pairwise witnesses against different rivals do not establish possible best. Return a common best when supported, the condition-dependent alternatives otherwise, or the precise comparison gap. No probability distribution over the compatible models is implied.
PSDA.6:4.5.5 - Reopen only what the changed judgement affects
A further judgement by the same holder on the same profiles adds a constraint to the existing family. Intersect, check nonemptiness again, and recalculate the dependent comparisons. A correction replaces its old row and every restriction derived from it; retaining the superseded row could manufacture a conflict. Preserve consequence measurements, protected conditions and comparisons whose grounds remain valid.
A changed holder, horizon, range, consequence meaning or compensation rule can invalidate the representation itself. Reconsider that premise before intersecting parameters in the old form. State which results depend on it and which remain usable. Stop when the receiving comparison is settled to the needed strength or the outstanding judgement, form or authority and its effect on the conclusion are identified.
PSDA.6:4.5.6 - Construct preferences over finite uncertain outcomes
Use this branch for a comparison of uncertain consequences under expected utility. A lottery is a probability distribution over specified consequences, which may be complete profiles of cost, delay and service. If a task supplies the utility function, consequence probabilities and their conditions, go directly to Connect preference to the consequence model below and report the conditional ranking. To represent a holder’s preferences, establish the applicability of expected utility and use attributable lottery judgements for any missing utility values. Those judgements can concern hypothetical lotteries without exposing anyone to an actual gamble.
Choose the representation before estimating it. Expected utility evaluates a lottery P by sum_x P(x) u(x). For an attributed account, its utility values represent the holder’s comparisons of lotteries. Stipulated values instead define a conditional analytical account. A function that orders certain outcomes alone does not determine the ranking of lotteries. In particular, any increasing transformation preserves a certain-outcome ordering, but generally changes the ordering of expected values.
Ask whether mixing each of two compared lotteries with the same third lottery, in the same proportions, should preserve the comparison. For example, if L is preferred to M, would the holder retain that preference between a half chance of L plus a half chance of H and a half chance of M plus a half chance of H? Challenge this premise with material cases, including a move from certainty to a chance of failure. Ask whether small changes in probabilities should permit small changes in the comparison, or whether a strict priority or threshold must prevail. These questions make the representation’s commitments discussable; finitely many answers cannot validate it for every lottery. If a common-mixture reversal, a noncompensatory rule or a material dependence on how the uncertainty is presented defeats the proposed form, retain the expressed comparisons and use a representation that preserves the feature, or identify the remaining representation gap. A probability-weighting or other non-expected-utility model needs its own construction; changing utility numbers alone cannot repair every failed form.
Build one joint family. When constructing a finite utility family from judgements, enumerate the complete consequence profiles appearing in the alternatives and elicitation questions. Obtain two intelligible reference consequences, w and b, that the holder treats as worst and best over this set and strictly distinguishes. Normalize u(w) = 0 and u(b) = 1; require 0 <= u(x) <= 1 for the other represented profiles. This chooses an origin and unit, not the utility of an intermediate consequence. If common anchors lack support, retain the direct comparisons or use a different justified representation. If every consequence is indifferent, return that simpler relation.
Where a utility is still to be established, compare receiving x for certain with a lottery giving b with probability p and w otherwise. Stated indifference gives u(x) = p. If x is no worse than the lottery at p, add u(x) >= p; if the lottery at q is no worse than x, add u(x) <= q. Thus the holder can bound a comparison without identifying one exact probability of indifference. The probabilities in these questions describe the hypothetical lottery. They are not estimates of the real intervention’s chance of success.
Retain all answers on the same utility variables. A comparison of two finite lotteries L and M adds sum_x L(x) u(x) >= sum_x M(x) u(x) when L is no worse, equality for indifference, or a strict inequality for strict preference. A comparison left open adds no equality. Certain-outcome comparisons are the special case with probability one on one consequence. These are linear constraints in the unknown utilities; use the feasibility and strictness treatment in §4.5.3 and the conflicting-subset diagnosis in §4.5.4. An inconsistent family calls for correcting the question or its interpretation, revisiting a judgement with its holder, or changing the form. It does not justify silently dropping a judgement.
This construction assigns utilities to complete profiles and requires no additive decomposition across their attributes. An additive multiattribute utility can reduce elicitation effort when its additional assumptions fit; the deterministic value model in §4.5.3 alone does not establish those assumptions for lotteries. Keep different holders’ families separate unless the receiving work supplies a justified collective rule.
Connect preference to the consequence model. For each alternative a, obtain the distribution P_a(x) over these same profiles from the subject’s probability model. Preserve shared causes, decision-dependent outcomes and the evidence limits described in PSDA.7 and PSDA.8. With one probability model and one utility function, compute EU(a) = sum_x P_a(x) u(x). Comparing expected utility is generally different from applying utility to an average physical consequence.
When probabilities or utilities remain imprecise, retain the set K of compatible complete accounts (P,u). Here P supplies a distribution for every alternative, and u is one utility function used across them. Use all combinations of the probability and utility families only when no joint restriction excludes them. If the meaning of a consequence, a model parameter or an elicited judgement links the two families, carry that restriction in K. Do not mix one alternative’s favorable model with another’s unfavorable model.
For each pair a and b, compare D(P,u) = EU(a) - EU(b) over K. Alternative a is necessarily no worse when every compatible account gives D >= 0; it is possibly no worse when at least one does. A contrary account defeats necessity; a feasible witness establishes possibility. The common-witness rule for possible best in §4.5.4 still applies. If K is empty, repair the failed basis before drawing a preference conclusion. No probability distribution over K follows from having retained it.
For fixed probabilities, D is linear in the utilities, so the finite constraint method can find a bound or witness. If both probabilities and utilities vary, their products generally make the problem bilinear; separate interval bounds or one linear-programming call are not automatically an exact solution. Use direct algebra, a complete finite enumeration or an appropriate bounding method under MMP.10 and MATH.20. Report the domain over which the computation establishes the bound. With strict constraints, distinguish an attained witness from an unattained limiting value.
Return the supported preference, conditional alternatives or unresolved comparison, with the consequence meanings, preference evidence or stated assumptions and chosen form. A changed probability estimate revises the assessed distribution of consequences. A new or corrected lottery judgement affects the utility family. Recompute affected relations. If an open preference matters to the requested finding or its use, seek an attributable judgement when warranted or retain the conditional answer. More precise factual measurements can leave the value question open.
PSDA.6:4.6 - Return the account and its exact incompleteness
Give the attributed concerns, consequence meanings and measures, applicable protections, preference rules and their grounds. Include any material interaction, dissent, unsupported transformation or condition that would change the result. If the merits of a requirement were at issue, include §4.4’s appraisal and the limit imposed by its current force.
For a numerical construction, include the chosen form and its assumptions, the profiles and judgements constraining it, the joint compatible family or identified conflict, and the scope of any necessary or possible preference. PSDA.8 may use it only for the named comparison. The account can support a partial order or leave a comparison unresolved. Missing preference information is not automatically missing empirical evidence: a further experiment may estimate consequences without deciding how they should be valued.
A material ambiguity, repeated consequence or unexplained trade-off can identify the need for this work. For consequential reliance, qualify the measurements, elicitation or other value grounds the comparison uses. Justify any aggregation, establish the governing force of binding conditions and apply the relevant domain assurance.
PSDA.6:5 - Archetypal Grounding
PSDA.6:5.1 - Two meanings of flood service
In the illustrative flood-pump engagement, a residents’ association emphasizes property protection, while a separately heard group of mobility-constrained residents emphasizes reachable assistance. The board is the later choice owner, not an assumed representative of every value judgement.
The analyst replaces a generic “service quality” score with this account:
| Concern | Consequence representation | Trade-off or qualification boundary |
|---|---|---|
| Reduce property loss. | A qualified loss estimate for the affected properties and flood conditions. | The estimate is not evidence of reachable assistance; its valuation and uncertainty remain explicit. |
| Keep assistance reachable. | Whether the defined assistance service can reach the named population under the relevant access conditions. | The expressed concern is retained. Any binding access or safety condition needs its direct competent basis; it is not inferred from participation. |
| Maintain pumping service. | Hours of unavailability for the specified pumping service over the season. | This operational measure does not cover all property or human consequences. |
| Limit resource use. | Resources required by each whole arrangement, including continuing obligations and transition. | A cost preference does not permit violation of a separately established protected condition. |
The analyst removes a second “reliability benefit” entry that merely repeats the same pumping-service hours. Reachable assistance stays separate even if failures often occur in the same storms.
If the board adopts a trade-off between operating-service hours and budget units for a narrow analytical slice, the account names that value judgement, range, and limit. It does not present the slice as the residents’ common value function. If the association and the other group reject a common trade-off, the return carries separate accounts and a stated disagreement; another hydraulic simulation cannot resolve that disagreement.
PSDA.6:5.2 - Development directions do not share a universal improvement score
For the ninety-day service-organization case, a qualified strategy result gives reliability priority over throughput for the current horizon. That premise helps frame the comparison but does not supply numerical weights or a causal effect estimate.
| Proposed holder-specific input | What the value account may use | What it must not infer |
|---|---|---|
| Human capability result for representative later Work. | The supported performance distinction and the person’s relevant protected conditions. | Course attendance, assessment performance, retention, and transfer are not one interchangeable development value. |
| Organization result for positions, interfaces, and continuing service. | Contribution and service consequences for the named arrangement. | A local team’s gain is not automatically a gain for the organization or another team. |
| AI/model evaluation for a named version and environment. | The qualified performance, resource, oversight, and failure distinctions. | A benchmark percentage is not a person’s capability measure or the whole service’s reliability. |
The adviser keeps those premises in their own meanings. A common service-level comparison is possible only after qualified domain results connect each direction to that same service and horizon. Without that connection, “human 4, organization 4, AI 4” is not a meaningful equal score. Identify the missing link to the service consequence. Seek evidence for that link if its attainable contribution to the intended comparison warrants the cost. Otherwise return the supported account and its limit.
PSDA.6:5.3 - Cheap non-use
Two qualified replacement parts have the same relevant consequences except for a directly comparable resource cost, and the authorized recipient’s cost preference is already explicit. Reuse that account. An objectives hierarchy and a fresh weighting workshop would add no value distinction.
PSDA.6:5.4 - Retain a protective restriction or revise a redundant control
These constructed variants concern a service team’s proposed use of an external inference provider. In the first, a qualified configuration result shows that the proposed route exposes identifiable confidential incident records to provider staff who have no part in serving those clients. An available internal route meets the current service need within its support budget. The restriction prevents that disclosure; relaxing it for the faster external route would expose those clients without a corresponding needed service gain. The value account therefore supports retaining the restriction and excludes this external route for the current use. It does not infer that all provider configurations are unsuitable.
In the second variant, the configuration and access protection remain adequate. The disputed rule additionally requires three hours of manual transcription of an access log. Assume the competent review has established that the retained, inspectable electronic log covers the same events and that transcription adds no omitted event or distinct check. The same three hours would otherwise test a newly exposed, unreviewed access path to the same protected records. On these facts, retaining the extra transcription displaces more protective work without an identified protective gain. Recommend revising that requirement to use the qualified electronic record while keeping the actual access controls. If a missing event or an independent check is found, that premise changes the recommendation.
The rule owner cannot amend the transcription requirement before this decision. The recommendation is complete. The current rule still applies to the options under consideration, and an option that cannot satisfy it must wait. The lost testing opportunity remains a consequence of the present constraint. The contrasted results follow from the stipulated contribution and burden.
PSDA.6:5.5 - Obtain an exchange family and use a new judgement
A service manager supplies one legitimate value account for a four-week arrangement. Known costs are in thousands of resource units, waiting times in hours. The relevant ranges are 20–60 and 1–5. Lower values are preferred. The manager explicitly accepts constant value increments and a constant cost-for-time exchange within these ranges. Those are the form assumptions; the following comparisons constrain its parameters.
| Alternative | Cost c | Waiting time t | Eligibility for this comparison |
|---|---|---|---|
| A | 60 | 2 | Admissible. |
| B | 30 | 5 | Admissible. |
| C | 45 | 3 | Admissible. |
| D | — | — | Fails a currently binding access condition; no timely amendment is available. |
The manager says that (60,1) is no worse than (20,5), and (20,4) is no worse than (60,2). Applying §4.5.2 gives:
| Stated judgement | Joint constraint | Consequence for r |
|---|---|---|
(60,1) no worse than (20,5) | 60 + r <= 20 + 5r | r >= 10 |
(20,4) no worse than (60,2) | 20 + 4r <= 60 + 2r | r <= 20 |
Thus L_r(c,t) = c + r t with 10 <= r <= 20 is the compatible family. It is nonempty: both endpoints satisfy both weak comparisons. No interview is available to choose a narrower exchange. In normalized range weights the same family has 1/3 <= w_c <= 1/2, with w_t = 1-w_c; those numbers were derived from the profiles, not assigned to criterion labels.
The receiving PSDA.8 comparison keeps one shared r:
| Difference over the whole family | Consequence |
|---|---|
L_B - L_C = 2r - 15, between 5 and 25 | C is strictly preferred to B throughout. |
L_B - L_A = 3r - 30, between 0 and 30 | A is no worse than B throughout, with a tie at r = 10. |
L_A - L_C = 15 - r, between -5 and 5 | C is better below 15; A is better above 15; they tie at 15. |
These linear differences establish the signs over the entire interval. At r = 10, the losses for A, B and C are 80, 80 and 75; at r = 20, they are 100, 130 and 105. Both exchanges satisfy the elicited information but give different best alternatives. The PSDA.9 sensitivity result is therefore the threshold 15 thousand per hour, and the PSDA.14 return is the conditional set {A,C}. B cannot be best in any compatible model; D remains ineligible. Choosing r = 15 merely because it is the interval’s midpoint would add unsupported indifference.
Later the same manager states that C is no worse than A. Its new row is 45 + 3r <= 60 + 2r, hence r <= 15. Intersecting gives [10,15]. Now L_A-L_C lies in [0,5] and L_B-L_C in [5,15]: C is no worse than A and strictly better than B for every compatible exchange. C is a necessary best, while {A,C} remains the set of possible best alternatives because A ties at 15. Return C as a supported direction with that tie intact. The new judgement changed the value family and its comparison; it changed neither the consequences nor D’s access failure. The decision holder can use this recommendation within the stated comparison.
PSDA.6:5.6 - A failed affine form is not automatically contradictory judgement
Consider a different continuation from the original [10,20] family. The manager additionally says that (20,5) is no worse than (44,2). In the affine form this yields 20 + 5r <= 44 + 2r, or r <= 8. That row and the first original row r >= 10 cannot hold together. Return this specific conflict; the former comparison no longer has a compatible affine basis.
Ask whether the equal-increment assumption still expresses the manager’s values. They may care much more about reducing two hours to one than about reducing five hours to two. For example, a curved time-loss contribution taking values 0,20,28,36,44 at times 1,2,3,4,5, with loss c + h(t), satisfies all three profile comparisons: 60 <= 64, 56 <= 80 and 64 <= 64. This constructed witness proves only that those three judgements need not contradict one another. It is not an elicited curve and supplies no recommendation by itself.
If the manager withdraws constant time increments but retains addition, §4.5.3 constructs the whole compatible monotone family at the relevant levels. If the manager instead states the noncompensatory two-hour rule in §4.5.1, that different value account prefers A to B and C directly. Either change needs its stated basis; neither is a convenient parameter adjustment inside the old affine model. If no clarification is available, retain the attributed comparisons, D’s independently established exclusion and the exact failed form. Return the numerical ranking as unresolved.
PSDA.6:5.7 - Separate an uncertain result from its value
A team compares two ways to obtain optional machine time in a ten-hour window. Existing provision covers required service; this comparison concerns additional usable hours, with other material consequences held equal. S supplies four hours for certain. R supplies ten hours if a trial succeeds and zero otherwise. The team prefers more usable hours, but that ordering does not yet compare S with R.
For a success probability of one half, utility u(x) = x/10 gives EU(S) = 0.4 and EU(R) = 0.5; R is preferred. Utility u(x) = sqrt(x/10) preserves the same ordering of certain hours but gives EU(S) = sqrt(0.4), approximately 0.632, and EU(R) = 0.5; S is preferred. These two constructed functions show why the certain-outcome ordering is insufficient. Neither is an elicited preference.
The team accepts the expected-utility form for this bounded comparison and uses zero and ten hours as its reference consequences. Let q = u(4), with u(0) = 0 and u(10) = 1. Asked about hypothetical chances, the team says that four certain hours are no worse than a 40 percent chance of ten hours, but a 70 percent chance of ten hours is no worse than four certain hours. Each lottery otherwise gives zero hours. These answers yield 0.4 <= q <= 0.7; they supply no point estimate.
Separately, engineering evidence bounds R’s success probability by 0.45 <= p <= 0.55. In this constructed case the probability estimate and preference restrictions impose no additional joint constraint. Thus EU(R) - EU(S) = p - q ranges from -0.25 to 0.15. Neither alternative is necessarily no worse. The accounts (p,q) = (0.55,0.4) and (0.45,0.7) witness opposite preferences. The two accounts leave the ranking open. Greater precision about success can still leave the team’s valuation unresolved.
The team can answer one further preference question: four certain hours are no worse than a 60 percent chance of ten hours, otherwise zero. Intersecting that answer with the previous family gives 0.6 <= q <= 0.7. The difference now ranges from -0.25 to -0.05, so S is strictly preferred throughout the joint family. No midpoint utility or invented probability over utility functions was needed.
A changed engineering result can reverse that conclusion without changing the preference account. If a genuinely different trial design raises the supported success range to 0.8 <= p <= 0.9, while the consequence meanings and other consequences stay the same, the difference ranges from 0.1 to 0.3: R is now strictly preferred. If the redesign also changes cost, timing or the meaning of usable hours, those omitted consequences must first enter the profiles; the old one-attribute calculation would no longer answer the whole comparison.
The construction concerns one attributed preference account. It neither aggregates conflicting team members’ views nor establishes that the trial has the assumed success probability. Those are separate inputs whose absence would leave a different question open.
PSDA.6:6 - Bias-Annotation
Scope: value representation for one bounded comparison. Lenses: Onto/Epist separates concern, measure, preference, and obligation; Prag preserves decision relevance; Gov exposes representation and authority; Did makes trade-offs concrete.
Template bias imports irrelevant criteria. Proxy bias rewards what is easy to count. Splitting one objective into many criteria can change its apparent importance. Sponsor averaging can conceal who bears a loss. Counter these by returning to attributed consequences, inspecting overlaps, and retaining disagreement and missingness.
PSDA.6:7 - Conformance Checklist
- Values are attributed to the relevant subjects and receiving comparison, with representation limits intact.
- Objectives and means are distinguishable at the level needed for the comparison.
- Repeated consequences are not counted twice; correlation alone does not justify deletion.
- Measures retain subject, scale, unit, horizon, evidence, proxy limits, and missingness.
- Binding conditions, aspirations and unresolved authority questions are distinguished. Where the merits of a requirement were in question, the appraisal identifies protection, burden, the supported conclusion and the requirement’s current force.
- Value transformations and weights have declared ranges and identifiable grounds, such as attributable preferences, an applicable rule or stated analytical assumptions. The grounds for the chosen form are separate from the fitted parameters.
- The numerical family preserves the joint judgements, has a checked feasibility result, and distinguishes incomplete information from contradictory judgements or a failed form.
- Necessary and possible comparisons retain their quantifiers, ties and dependence on the same compatible model. A changed judgement reopens only dependent results unless it changes the form itself.
- For finite expected utility, the lottery judgements or stipulated assumptions and the representation support the utility account. The probabilities describe the consequences. Compare alternatives under the same compatible account and preserve open preferences.
- Interactions and refusals to compensate are preserved.
- Any collective aggregation has an explicit basis; incompatible accounts remain visible otherwise.
- The return identifies any unresolved comparisons, dissent, missing source grounds and conditions affecting future use.
- A supported preference comparison covers only its attributed account and supplied consequences; authorization of the later choice and a claim of effect need their own grounds.
PSDA.6:8 - Common Anti-Patterns and How to Avoid Them
| Anti-pattern | Repair |
|---|---|
| “Everyone agrees safety is important, so give it twenty percent.” | Identify the condition, its source, and whether compensation is permitted. |
| “Equal weights are neutral.” | State the consequence ranges and inspect how criterion splitting changes the result. |
| “These two measures correlate, so one is redundant.” | Examine whether they represent the same valued consequence. |
| “Normalize everything to zero–one.” | Establish value meaning and admissible operations before any transformation. |
| “The workshop mean is the public preference.” | Preserve attribution and require an explicit aggregation basis. |
| “Collect more data to resolve the value dispute.” | Identify whether the missing input is evidence, preference, representation, or authority. |
PSDA.6:9 - Consequences
The comparison shows which consequences are valued, by whom and which trade-offs are permissible. A simpler account may preserve the same distinctions at lower cost. Where the evidence supports only a partial order or separate disputed accounts, return those results.
PSDA.6:10 - Rationale
Value representation makes consequences decision-relevant without confusing description with preference or preference with permission. Concrete trade-off questions expose assumptions hidden by criterion labels. Preserving multiple admissible accounts lets later comparison show where a conclusion is shared and where it depends on a disputed judgement.
PSDA.6:11 - SoTA-Echoing
| Practice question | Best-known line | Serious alternative or default | Defect overcome and pattern mutation | Source roles and limits | Reopen condition |
|---|---|---|---|---|---|
| How can objectives be concise without omitting what matters? | Inspect means–ends structure, relevance, overlaps, and sensitivity to simplification. | Retain an exhaustive descriptive hierarchy, or delete criteria by correlation alone. | Adapt: :4.2–:4.3 keep a small consequence-led account and test any lost distinction. More analyst attention is accepted where it prevents double counting; small clear cases remain short. | Marttunen et al.’s 2019 objectives-hierarchy study supplies the comparative methods and retrospective environmental cases. It supports disciplined simplification, not one optimal hierarchy or a statistical license to discard values. | Reopen when a removed distinction can reverse the comparison or a lighter structure preserves the same concerns. |
| Must incomplete preferences become one weighted score? | Construct the preference model from consequence comparisons and use all compatible instances. | Fit one additive score, average weights, or retain a noncompensatory rule when that better expresses the judgement. | Adapt: :4.5.1–:4.5.5 develop form selection, joint constraints, feasibility, necessary/possible comparison and local revision; :5.5–:5.6 show a complete use and failure of the affine form. The trade-off is a potentially less decisive answer for less invented precision. | Greco, Słowiński and Wallenius’s 2025 MCDA review, especially §3.1.2 and §7.1 (accessible author-hosted PDF), supplies preference assessment and robust ordinal regression. Its rules and non-additive families remain serious alternatives. The two-criterion derivation and service cases here are bounded applications, not its empirical findings or a universal additive-representation theorem. | Reopen when new judgement changes the compatible set, an interaction defeats the form, or the holder, range or governing condition changes. |
| How can incomplete risk preferences become a usable comparison? | Elicit finite lottery comparisons, construct the compatible utility family and compare expected utilities over compatible probability and utility accounts. | Assume risk neutrality from a certain-outcome score, fit one curve despite open judgements, or retain a different explicit preference rule. | Adapt: :4.5.6 supplies the general construction and :5.7 shows elicitation, an open result and a changed comparison. Further elicitation or computation is warranted when its attainable contribution to the requested preference finding or its use justifies the burden. | Sahlin, Troffaes and Edsman (2021), §§3.3–3.5 and Appendix B demonstrate imprecise lottery-based elicitation with imprecise consequence probabilities. Their additive environmental model is not assumed here. McCarthy, Mikkola and Thomas (2021) develop the representation question for incomplete mixture preferences; retaining a family does not remove independence and continuity premises. Breig and Feldman (2026, online 2025) test expected-utility consistency in a specific elicitation design, supporting scrutiny of the form rather than universal use. The finite profile construction and machine-time calculation here are explicit adaptations, not empirical findings from those studies. | Reopen when lottery judgements conflict with the form, a consequence is omitted, the probability model or holder changes, or an unresolved joint restriction changes the comparison. |
| Can common coordinates perform the value judgement? | Declare meanings and scales separately from preference, comparison, and selection. | Treat a dashboard or normalized vector as the decision model. | Adopt: :4.3 uses A.19 only for the declared space or predicate and leaves value elicitation here. The declared coordinates allow a reader to interpret the table. Scoring or combining the values requires a rule for that comparison. | Current A.19, C.16, and the direct comparison patterns supply the relevant typing and operation boundaries. They do not elicit the engagement’s values or authorize a trade-off. | Reopen when the scale, predicate, consequence range, or receiving use changes. |
PSDA.6:12 - Relations
PSDA.12may supply attributable concerns and participation conditions for the value question. Establish numerical weights or a mandate to aggregate from the evidence or rule that provides them.PSDA.8may use the value account for the named comparison.PSDA.9can test its dependence on the compatible family, andPSDA.14can use a supported direction, conditional set or gap.MMP.10andMMP.11supply general constraint and model-family construction after the value meanings and form conditions in §4.5 are established.MATH.20supplies bounds over the admitted domain; none of these suppliers elicits the holder’s preferences.C.16governs measurements;A.19supplies a declared characteristic space or predicate when needed. Scoring, aggregation, comparison, and selection remain separate direct operations.- Direct ethics, law, safety, governance, finance, and holder practices supply the particular conditions and claims they own.
A.15.9helps obtain a missing qualified contribution without transferring that authority. - For a missing, stale or incompatible premise, use an applicable direct result or state the specific limit on the value account.