eduKateSG · Judgement · The Gold Standard of Education
Remove clearly inferior options before negotiating the trade-offs that remain
Choose the explanation you need, or follow all twelve chapters in order.
Full chapter index · Independent practice · Judgement subject guide
Multi-criteria choices become confusing when every option is collapsed into one score too early. Dominance analysis first asks whether one option is at least as good on every relevant criterion and better on at least one. A dominated option can be removed without choosing weights. The remaining non-dominated options form a Pareto set: improving one measured objective requires giving up something else.
Hard constraints and non-compensable requirements come before that trade space. An option that violates a safety limit or essential accessibility requirement does not become acceptable because it is cheap on average. The method therefore has three layers: feasibility, dominance and preference.
The examples below are fictional. A Pareto frontier does not decide values, guarantee fairness or remove uncertainty. It organises the comparison so that a decision-maker can see which trade-off is actually being accepted.
Continue from The Gold Standard of Judgement within The Gold Standard of Education: learning that transfers into life and the How X Works learning library.
Full chapter index
See the mechanism · Chapters 1–2
Build the worked model · Chapters 3–5
Test what changes · Chapters 6–8
Check evidence and limits · Chapters 9–10
Practise and continue · Chapters 11–12
State the decision, options, affected people, time horizon and authority. Then define each criterion with units and direction: higher reliability may be better; lower cost and delay may be better.
Ambiguous directions cause silent errors. A score called “risk” is minimised, while a score called “safety margin” is maximised. Transforming both to larger-is-better can help, but preserve the original units.
Do not add criteria that duplicate the same consequence. Double-counting can make an option appear stronger without adding new evidence.
Evidence check: for Define the decision and measurement direction, preserve the observation before the interpretation. Record population, binary action, outcome and time horizon. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
A non-compensable constraint is a condition that cannot be offset by gains elsewhere. Examples include a legal prohibition, an essential capacity floor or a maximum exposure limit.
Write each threshold before seeing the preferred option. Record the evidence source, uncertainty allowance and who may approve an exception, if any.
Separate true constraints from strong preferences. Calling every preference mandatory can leave no feasible option; treating every constraint as negotiable can hide unacceptable harm.
Evidence check: for Apply non-compensable constraints first, preserve the observation before the interpretation. Record whose tradeoff the threshold represents. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Option A dominates B when A is no worse than B on all included criteria and strictly better on at least one, under the chosen directions. If B is cheaper but slower, neither dominates the other.
Pairwise comparison needs comparable conditions. If one cost includes maintenance and another does not, the apparent dominance is invalid.
Keep a dominance reason: “B removed because A has equal capacity, lower lifecycle cost and shorter delay.” This is more reviewable than a hidden spreadsheet filter.
Evidence check: for Test pairwise dominance, preserve the observation before the interpretation. Record the formula, denominator and threshold weight. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Three feasible learning platforms are compared on annual cost, median task time and accessibility score. A costs $90,000, 8 minutes and scores 92. B costs $100,000, 9 minutes and scores 90. C costs $75,000, 11 minutes and scores 95.
A dominates B: it is cheaper, faster and more accessible. C is cheaper and more accessible than A but slower, so A and C are both non-dominated.
The choice is now clearer. The live trade-off is between $15,000 and three minutes, with an accessibility difference. A weighted total should not bring B back unless evidence or criteria change.
| Option | Annual cost | Task time | Accessibility score | Dominance result |
|---|---|---|---|---|
| A | $90,000 | 8 min | 92 | Non-dominated |
| B | $100,000 | 9 min | 90 | Dominated by A |
| C | $75,000 | 11 min | 95 | Non-dominated |
Evidence check: for Work through a fictional option table, preserve the observation before the interpretation. Record TP and FP counts recalculated at that threshold. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
In two dimensions, plot feasible options and identify those not dominated. In more dimensions, the Pareto set may be a surface rather than a simple line.
A frontier shows efficiency relative to the evaluated options and metrics. It does not prove global optimality if important designs were never generated.
Look for knees where a small further improvement in one criterion requires a large sacrifice in another. A knee can focus discussion, but the decision still needs stakeholder values and uncertainty.
Label every plotted point with its option and preserve the excluded infeasible set. Otherwise a reader cannot tell whether the visible frontier was produced by performance or by hidden filtering. When criteria exceed two dimensions, use a table of pairwise dominance and several projections rather than pretending one chart shows the whole problem.
The frontier can contain many nearly identical options. Cluster or shortlist only with a declared tolerance, because a visual overlap may hide a meaningful threshold crossing.
Evidence check: for Construct and read a Pareto frontier, preserve the observation before the interpretation. Record treat-all, treat-none and current-practice comparators. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Weights express preference among trade-offs; they should not rescue infeasible options. Apply them only after the criteria and feasible set are stable enough to discuss.
Additive weighted scores are compensatory: a large gain on one criterion can offset a loss on another. That is inappropriate for veto conditions and may be misleading when scales are arbitrary.
Test several defensible weight sets. If one option wins only under a narrow weighting, report that fragility instead of presenting the rank as objective.
Normalisation deserves its own check. Dividing by the observed range makes scores depend on which options happen to be present, while target-based scaling depends on the chosen target. Repeat the calculation after adding a plausible alternative and after changing the scale. A ranking that moves because of a harmless representation choice needs explanation before action.
Evidence check: for Keep weights after feasibility and dominance, preserve the observation before the interpretation. Record the relevant range and uncertainty around differences. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 7 OF 12 · Test what changes
7. Model thresholds, vetoes and lexicographic priorities
Some decisions use an aspiration floor rather than a hard legal constraint. Others use a veto: a sufficiently bad performance on one criterion disqualifies an option regardless of total score.
Lexicographic rules rank one criterion first, using the next only for ties. They fit some emergency or rights-based decisions but can ignore large differences on lower priorities.
Choose the rule because it represents the decision, not because it produces the desired winner. Document why compensation is allowed or forbidden.
Evidence check: for Model thresholds, vetoes and lexicographic priorities, preserve the observation before the interpretation. Record calibration in the intended validation population. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Suppose A’s task time is estimated as 8 ± 2 minutes and B’s as 9 ± 1. The point estimates favour A, but the plausible ranges overlap. Robust dominance may not hold.
Use intervals, scenarios or probabilistic simulation where the evidence supports them. An option robustly dominates another only when the comparison survives the relevant uncertainty set.
Measurement error can alter the frontier. Preserve dominated options in the audit until the evidence and threshold decisions are approved.
| Option | Cost range | Time range | Reliability range | Robust reading |
|---|---|---|---|---|
| A | $88k–$94k | 6–10 min | 94–97% | Not uniformly better |
| B | $96k–$103k | 8–10 min | 95–98% | Costlier, reliability may lead |
| C | $72k–$80k | 10–13 min | 93–97% | Cheaper, slower |
Evidence check: for Add uncertainty to dominance claims, preserve the observation before the interpretation. Record target prevalence and sampling weights. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
An average can hide who bears the loss. A platform may reduce median task time while making access much worse for a small group.
Add distributional criteria or explicit constraints where they matter. Avoid turning affected groups into a small weighted term that can be overwhelmed by majority benefit without an accountable decision.
Consultation may change the criteria, thresholds or acceptable trade-offs. Recompute the feasible and non-dominated sets rather than attaching comments to a fixed result.
Where consequences fall on different groups, show the distribution before aggregation. An option that is non-dominated on population averages may be dominated within the subgroup carrying the greatest risk. The decision record should explain which level of analysis governs and why.
Evidence check: for Check fairness and distribution, preserve the observation before the interpretation. Record resampling boundaries and external validation. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Pareto-efficient does not mean good. Every option on a poor frontier can be unacceptable. Dominated does not mean useless if the model omitted a capability or used incorrect data.
The frontier depends on the available option set. Adding a new option can dominate several existing ones. Removing a criterion can also change dominance.
Do not confuse a visually attractive chart with validated evidence. Units, baselines, missing costs and measurement dates remain part of the judgement.
Dominance is also local to the selected criteria and horizon. A low-cost option can dominate during the first year but lose once renewal, maintenance or exit costs are included. State whether the analysis is annual, lifecycle or scenario-specific, and do not combine horizons without discounting and a common decision date.
Finally, a frontier does not transfer authority from people to a model. The responsible person or group still owns the constraint definitions, the accepted sacrifice and the review trigger.
Evidence check: for Limits and common misconceptions, preserve the observation before the interpretation. Record capacity, downstream harms and missing utilities. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 11 OF 12 · Practise and continue
11. Independent practice and changed-context transfer
Practice case: four plans are compared on cost, time and reliability. Plan D is worse than C on all three. Explain why D can be removed without choosing weights and what evidence must be comparable.
Changed context: D alone meets a newly discovered data-residency constraint. Rebuild the feasible set and explain why the earlier dominance result no longer decides the choice.
Then add uncertainty. If C’s reliability estimate has a wide interval, state the condition under which robust dominance fails and the next measurement worth collecting.
Evidence check: for Independent practice and changed-context transfer, preserve the observation before the interpretation. Record which changed context reverses the preferred rule. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Does the Pareto frontier pick the winner?
No. It removes clearly inferior options and displays the trade-offs among the remaining ones. Values, constraints and authority still determine the decision.
Can a weighted score replace dominance analysis?
It can rank options under one compensatory preference model, but it may hide dominated options, arbitrary scales and non-compensable limits.
What if no option satisfies every constraint?
Do not quietly relax them. Return to design, scope, timing or authority and record which requirement could lawfully change and who decides.
Continue with satisficing and robust feasibility and expected utility and certainty equivalents.
Authoritative applied background includes NASA’s Framework for Multidisciplinary Analysis, Design, and Optimization, which defines a Pareto front as designs where improving one merit requires sacrificing another and demonstrates fidelity limits, and the NASA Marshall Systems Engineering Handbook for documented multi-criteria trade studies and constraints. Sources and links were checked on 22 October 2026.
Evidence check: for FAQ, sources and connected reading, preserve the observation before the interpretation. Record the decision sentence that the curve is meant to support. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.