THE GOLD STANDARD OF JUDGEMENT · SPECIALIST STUDY
Represent set-valued evidence while keeping approximation and decision choices visible
A random set assigns uncertainty to sets of possible states rather than forcing every unit of support onto a single state. When focal sets are nested, the resulting consonant belief function corresponds closely to a possibility distribution. This bridge is useful—but converting a general belief function into a possibility envelope can lose conflict and dependence that decisions may need.
This guide belongs to the Judgement subject guide, the Gold Standard of Education, and the How X Works learning library. Choose a reading route, work the examples, then use the transfer checks to test whether the method survives a changed context.
Choose a reading route
Build the foundations · chapters 1–2
Apply the core method · chapters 3–5
Test evidence and alternatives · chapters 6–7
Handle limits and disagreement · chapters 8–9
Transfer and govern · chapters 10
The whole problem at a glance
| Object | Question | Meaning |
|---|---|---|
| Random set | Which subset may contain the state? | A set-valued uncertain variable |
| Mass assignment | How much support is committed to each focal set? | Evidence allocated to sets |
| Belief and plausibility | What is fully supported or not ruled out? | Lower and upper support bounds |
| Consonance | Are focal sets nested? | Possibility-compatible structure |
Every consonant belief function is structured; a general belief function need not have nested focal sets and cannot be reduced without possible information loss.
CHAPTER 1 OF 10 · Build the foundations
1. Treat a random set as a set-valued uncertain object
Core idea. A random set maps outcomes of an underlying uncertainty mechanism to subsets of a frame. In finite evidence theory, this induces mass on focal sets rather than only singleton states. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Define the frame, underlying sampling or epistemic interpretation, focal sets and mass normalization. Keep set uncertainty distinct from an ordinary random variable. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. An inspection can narrow a component fault to {A,B} without saying which one. Mass is assigned to that set, preserving unresolved specificity. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. A location estimate can be a random interval rather than a single noisy coordinate. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Frames can omit possibilities, and epistemic and frequentist interpretations should not be mixed casually. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. A random set is just a probability distribution with braces. Set-valued outcomes retain a different kind of uncertainty. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Explain what real observation or evidence event generates each focal set and why its mass is justified.
- Return to chapter contents
CHAPTER 2 OF 10 · Build the foundations
2. Calculate belief and plausibility from focal sets
Core idea. Belief in A sums mass committed wholly inside A, while plausibility sums mass not disjoint from A. The interval exposes support and unresolved possibility. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Enumerate focal sets, test subset and intersection relations, and verify duality between belief and plausibility of complements. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. For fault A, only mass on {A} contributes to belief, while masses on {A}, {A,B} and {A,B,C} contribute to plausibility. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. For event {A,B}, the nested example yields stronger belief because two focal sets lie wholly inside it. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. The interval is conditional on the frame and mass assignment; it is not automatically a calibrated probability interval. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. Plausibility is a second estimate of probability. It is an upper support measure with different semantics. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Compute belief and plausibility for every singleton and one composite event, then explain every included mass.
- Return to chapter contents
CHAPTER 3 OF 10 · Apply the core method
3. Recognise consonance through nested focal sets
Core idea. A belief function is consonant when its focal sets are nested by inclusion. The structure represents progressively less specific evidence without crossing focal alternatives. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Sort focal sets by inclusion, reject incomparable nonzero sets and record whether equal sets were combined. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. {A} inside {A,B} inside {A,B,C} forms a chain and supports a possibility representation. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. Replacing {A,B} with {B,C} creates crossing focal sets and destroys consonance. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Partial consonance and continuous cases require more careful definitions and algorithms. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. Any smooth plausibility contour proves consonance. Singleton values alone can hide non-nested composite structure. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Inspect every pair of nonzero focal sets and demonstrate inclusion in at least one direction.
- Return to chapter contents
CHAPTER 4 OF 10 · Apply the core method
4. Derive a possibility contour from consonant evidence
Core idea. For a consonant belief function, singleton plausibilities form a possibility distribution, and event plausibility is the maximum contour value within the event. Necessity corresponds to belief through duality. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Compute singleton plausibilities, normalize, then verify maxitivity across several unions and equality with event plausibility. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. The nested masses produce highest possibility for A, lower for B and only residual possibility for C. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. A continuous nested family produces level sets analogous to alpha cuts. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Ties and scale interpretation need care; cardinal operations require stronger assumptions than ordinal ranking. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. The contour adds probabilities across states. Possibility aggregation is maxitive, not additive. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Choose two events and prove their plausibility equals the maximum singleton contour inside each.
- Return to chapter contents
CHAPTER 5 OF 10 · Apply the core method
5. Detect what non-consonance represents
Core idea. Crossing focal sets can encode evidence that supports incompatible groupings or multiple dimensions. Forcing them into one nested chain may erase which combinations received support. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Visualise focal-set overlaps, identify source or feature creating each set, and calculate which composite-event bounds depend on crossings. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. Mass on {A,B} and {B,C} highlights B through overlap while preserving two different uncertain alternatives. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. Two sensors with different ambiguity patterns create crossing focal sets even if singleton contours look simple. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Non-consonance can arise from modelling choices or dependent combination, not only genuine conflict. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. Non-consonant evidence is defective and should be smoothed. It may carry decision-relevant structure. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Find two events with identical singleton contours but different composite plausibility under rival mass structures.
- Return to chapter contents
Worked inspection example
| Focal set | Mass | Interpretation |
|---|---|---|
| {A} | 0.2 | Specific support for A |
| {A,B} | 0.5 | Evidence narrows the state to A or B |
| {A,B,C} | 0.3 | Residual ignorance over all states |
| Nested? | Yes | Focal sets form a chain |
| Possibility contour | Derived from singleton plausibilities | Encodes the consonant structure |
Change the middle focal set to {B,C} and nesting breaks; a single possibility contour can no longer preserve the whole mass structure.
CHAPTER 6 OF 10 · Test evidence and alternatives
6. Build a possibility envelope with an explicit guarantee
Core idea. A consonant approximation can be chosen to dominate, be dominated by or minimize a distance from a belief function. Each guarantee answers a different question. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. State whether the envelope must preserve upper bounds, lower bounds, contour values or decision ranking. Define the optimization or construction and verify the promised inequality. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. A safety screen uses an outer possibility envelope so no state previously plausible is understated, accepting wider uncertainty. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. A triage shortlist uses an inner approximation to preserve only strongly supported states, with clear risk of omission. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. An outer envelope may become uninformative; an inner one can exclude real possibilities. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. The closest-looking curve is the best approximation. Visual similarity does not define evidential or decision loss. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Test the approximation on all decision-critical events, not only singleton values.
- Return to chapter contents
CHAPTER 7 OF 10 · Test evidence and alternatives
7. Measure approximation loss on relevant events
Core idea. Loss depends on which beliefs, plausibilities, orderings or decisions the application needs. A global numeric distance can hide a large error on one critical event. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Create an event set tied to decisions, compare lower and upper bounds before and after approximation, and perform worst-case and sensitivity analysis. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. Overall distance is small, but plausibility of a catastrophic event rises sharply; the report highlights that local consequence. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. For ranking candidate diagnoses, preserving singleton order may be sufficient only if actions depend solely on that order. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Enumerating all subsets is infeasible for large frames and requires structured search or domain-selected events. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. One average error number certifies the approximation. Decision-critical tails and constraints can dominate averages. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Choose a protected event and prove the approximation does not narrow its uncertainty without authority.
- Return to chapter contents
CHAPTER 8 OF 10 · Handle limits and disagreement
8. Combine evidence before deciding whether to approximate
Core idea. Combination and consonant approximation generally need not commute. Approximating each source first can discard structure that would affect dependence, conflict and the combined result. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Model source dependence, combine at the appropriate evidence level, then compare with early-approximation alternatives. Report conflict and normalization. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. Two non-consonant inspections combine to strongly isolate B, while separate possibility envelopes remain broad and miss the interaction. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. If independent sources are already consonant under compatible semantics, a possibilistic combination rule may be appropriate. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Independence is difficult to justify, and conflicting sources can make normalized rules unstable. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. Converting every source to possibility first makes fusion safer. Early simplification can erase complementary structure. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Run both pipelines on a small example and identify every event whose bound or action changes.
- Return to chapter contents
Approximation decision check
| Method | Preserves | May lose |
|---|---|---|
| Outer possibility envelope | Coverage or upper compatibility under its construction | Specific lower support |
| Inner approximation | Committed core evidence | Possible states outside the core |
| Contour matching | Singleton plausibilities | Interactions among composite events |
| Decision interval | Action bounds under a declared utility rule | A unique preference without added assumptions |
No universally best consonant approximation exists. The loss function, ordering and decision consequence must be stated.
CHAPTER 9 OF 10 · Handle limits and disagreement
9. Convert evidence bounds into actions separately
Core idea. Belief and plausibility describe evidence; an action requires utilities, constraints and a decision criterion such as lower expectation, upper expectation, interval dominance or robust admissibility. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Declare utilities and protected constraints, calculate decision bounds under the original and approximated models, and expose incomparability. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. A maintenance action is selected only when its worst supported loss stays below a safety threshold across both models. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. An exploratory action may use optimistic bounds but is limited in scale and paired with monitoring. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Decision criteria embody risk attitudes and values; no unique rule follows from the belief function. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. The most plausible state determines the best action. State evidence and action value are separate. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- For two actions, show the outcome interval and the exact criterion that ranks or leaves them incomparable.
- Return to chapter contents
CHAPTER 10 OF 10 · Transfer and govern
10. Publish the frame, masses, approximation and limits
Core idea. Reproducibility requires the frame, focal sets, masses, source interpretation, combination rule, consonance test, approximation guarantee, event loss and decision rule. The important discipline is to name the object being assessed, the boundary around it and the kind of conclusion the evidence can support. Without those three choices, technical vocabulary creates confidence without creating a checkable claim.
Method. Version the model and code, retain expert rationales, use exact arithmetic tests where feasible and write plain-language bounds without probabilistic relabelling. Write the procedure before looking at the preferred result, retain intermediate evidence and record who can challenge each assumption. A method becomes reusable when another reader can repeat the steps, locate disagreements and understand why a different answer emerged.
Worked example. A report publishes the original non-consonant model alongside the outer envelope used for rapid screening. The example is deliberately small enough to inspect. Its purpose is not to imitate every real case but to expose the decision rule, the evidence it consumes and the point at which judgement enters.
Changed-context transfer. A regulated application adds independent validation and change control around the same transparent model. Keep the underlying principle while changing the actors, incentives or data. If the conclusion survives only because the original labels were familiar, the learner has memorised a case rather than acquired a transferable method.
Limits. Publishing numbers does not validate the frame or sources. Sensitive evidence may require controlled access. A responsible conclusion therefore states what remains unknown, what evidence could reverse it and when the analysis must be repeated. Precision about limits is part of the result, not a disclaimer added after it.
Misconception to avoid. A possibility plot is a complete audit record. It hides how the contour was derived and what structure was lost. Test the tempting shortcut against a counterexample and ask which stakeholder bears the cost when it fails. The correction should alter the workflow, not merely add cautious wording.
Practical evidence check
- Give the record to a reviewer and require reproduction of focal-set nesting, one event bound, the envelope guarantee and the final action.
- Return to chapter contents
Build a random-set evidence worksheet
State the decision, unit of analysis, time horizon and affected people before collecting evidence. Separate observations from interpretations and interpretations from recommendations. Record data provenance, missingness and incentives that could distort reporting.
For every important claim, add a rival explanation and one observation that would favour it. Give uncertainty a visible place instead of hiding it inside a single score. Assign an owner and review date to assumptions that may change.
Run the worksheet with someone who did not design the analysis. Ask them to reconstruct the conclusion and identify where judgement entered. Revise any step that depends on private intuition or undefined terminology.
Finish with an action table: act now, monitor, investigate or stop. Link each action to evidence and a reversal condition so the method supports learning rather than defending the first answer.
Design a changed-context transfer check
Rewrite the case with a different scale, institution and distribution of power while preserving the underlying structure. Predict which parts of the method should remain stable and which require new evidence.
Remove one convenient data source and add one conflicting stakeholder account. Re-run the reasoning. If the answer changes, explain whether the change reflects better evidence, a different value choice or an unstable method.
Switch the time horizon. A decision that looks efficient this week may weaken resilience over years, while a long-run benefit may impose an unacceptable immediate burden. Report both rather than averaging them away.
Ask a reviewer to produce a counterexample. A transfer check passes when the method reveals why the new case differs, not when it mechanically produces the same conclusion.
Audit the reasoning and the record
Preserve source versions, transformations, thresholds, exclusions and decision notes. A final chart without its lineage cannot be independently checked and may conceal silent changes.
Trace one favourable and one unfavourable case from raw evidence to recommendation. Confirm that identical rules were applied and that exceptions were authorised, explained and reviewable.
Interview affected users about how the process works in practice. Formal policy may differ from operational reality, especially when time pressure, inaccessible interfaces or asymmetric power limit challenge.
Publish a proportionate account of method, limits, ownership and remedy. Protect sensitive data while ensuring that confidentiality is not used to block legitimate scrutiny.
Turn findings into accountable action
Assign every recommendation to a named role with authority, resources and a deadline. Distinguish mitigation from remedy: reducing future risk does not repair harm already caused.
Choose leading indicators that reveal drift early and outcome indicators that show whether people actually benefited. Disaggregate results where averages can hide concentrated burden.
Create an escalation rule for new evidence, missed thresholds and recurring complaints. Independent review should be available when the original decision maker has an incentive to defend the system.
Close the loop by publishing what changed, what did not and why. Re-open the analysis after material changes instead of treating one successful review as permanent certification.
Questions readers ask
What is a focal set?
- A subset of the frame that receives nonzero mass, representing support committed to that set without choosing a member.
What makes a belief function consonant?
- Its nonzero focal sets are nested by inclusion.
How does consonance connect to possibility?
- Singleton plausibilities form a possibility contour, and event plausibility becomes the maximum contour value.
Why not approximate every belief function immediately?
- Crossing focal sets can contain composite-event and interaction information that a single contour loses.
Is there one best possibility envelope?
- No. Outer, inner, distance-based and decision-focused approximations preserve different properties.
Does evidence select an action automatically?
- No. Utilities, constraints and a separate decision criterion are required.
Practise, verify and continue
Create a three-state mass assignment with nested focal sets, compute belief, plausibility and its possibility contour. Then replace one focal set to break consonance, build an outer envelope, measure loss on a critical event and compare two actions.
Move the method to diagnosis, interval measurement or safety screening. Preserve the frame, set-valued evidence, consonance test, approximation guarantee, decision bounds and sensitivity while changing domain evidence.
Related eduKate reading
Dempster–Shafer evidence, belief and plausibility
Possibility theory, necessity and maxitive uncertainty
Imprecise probabilities, credal sets and robust choice
Minimax regret under deep uncertainty
Robust decisions and reversal assumptions
Authoritative sources and scope
Glenn Shafer, A Mathematical Theory of Evidence
Glenn Shafer, Belief Functions and Possibility Measures
Dubois and Prade, Consonant Approximations of Belief Functions
Thierry Denoeux, Belief Functions Induced by Random Fuzzy Sets
Ilya Molchanov, Theory of Random Sets
This article is mathematical education, not clinical, engineering, financial or safety advice. The finite examples are simplified. Production models must justify the frame, focal sets, masses, source dependence, combination rule, consonant approximation and decision criterion. Belief, plausibility and possibility are not frequencies merely because they use numbers between zero and one. Approximation should be evaluated on decision-critical events and protected constraints, independently validated and revised when the frame or evidence changes.