#Candidate-Method Recovery from Work Evidence

A.3.1.MR · Stable · Architectural (A) · Normative unless marked informative · Part A - Kernel Architecture Cluster

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

Type: Architectural (A) Status: Stable Normativity: Normative unless marked informative

Plain name. Recover source-traceable candidate accounts of a reusable way from several performances or other direct evidence.

Primary reader. A practitioner, researcher, analyst, or method engineer who has observations or records from several performances and needs an honest reusable explanation before Method identification or specialist reconstruction.

Use this when. Use this pattern when you have observations or records from several performances and want to understand what reusable way they may show, but no Method has yet been established.

#Relations

#Content

#Problem frame

Use this when. Use this pattern when you have observations or records from several performances and want to understand what reusable way they may show, but no Method has yet been established.

First useful result. Return a traceable provisional explanation of the reusable way the material may show, the main competing explanation, important gaps, and the next observation or trial that would separate them—or state honestly that the material shows only what happened.

Three recognition cases.

• Several maintenance visits have videos, logs, notes, and known performers. Most show one inspection order, while one changes order after a vibration cue. The question is whether the evidence supports a fixed-order way, a cue-responsive way, or only a local sequence.
• Event data has been prepared and a discovery tool has produced a behavioural model. The question is what the selected data and modeling choices actually support, including what embodied, discretionary, or unrecorded contributions may be missing.
• Several practitioners describe “the same practice,” but their accounts still differ in applicability, participant meanings, intended result, or stop. The question is whether one candidate reusable way, several candidates, or no reusable account can yet be distinguished.

What goes wrong if missed. One vivid occurrence is generalized into a Method. Repeated event order is treated as reusable applicability. Several rival candidate subjects are combined into one false EntityOfConcern. A mined model, executable diagram, or coherent account is called a MethodDescription before a Method has been admitted. Missing tacit or discretionary contributions disappear behind the record format.

What this buys. A project can use imperfect evidence without overclaiming. Each positive candidate has a truthful subject, source-to-claim trace, important gaps, a rival, and a distinguishing question. Weak evidence can still return a useful record-only result. Stronger reconstruction can continue in specialist ME.18 without making every ordinary use pay that burden.

Not this pattern when. Use the closest applicable pattern instead:

• If the evidence no longer leaves rival candidate ways and the remaining question is whether the proposed way qualifies as the U.Method being identified—or which identity claim needs repair—use A.3.1.
• If the Method is already identified and the question is whether an episteme is its MethodDescription, use A.3.2.
• If wording still hides the object or relation being asserted, use F.19; follow an E.10 route only for a remaining FPF word, kind, or relation question.
• If the question concerns one dated Work occurrence only, use A.15.1.
• If the project needs prospective practice-architecture synthesis, use C.32.MWA.
• If the domain needs a complete reconstruction programme, use specialist ME.18.

When a DPF reuses this pattern. A DPF uses it only when several occurrences or sources create a candidate-recovery question that passes this entry. The DPF still supplies any domain-specific problem, evidence limits, vocabulary, result, and return that change practitioner action or judgment. If no such use-changing contribution remains, cite this pattern rather than copying it.

For example, an engineering study across several test runs may pass this entry, while a live maintenance incident may need only A.15.7 and a stable administrative checklist may need neither pattern. A music or dance DPF uses this recovery Method only for an actual several-performance question and keeps its own evidence, terminology, result, and return.

#Problem

Evidence about Work is not the reusable way itself. Videos, logs, interviews, artifacts, and process models are selected and interpreted under Methods. They may omit perception, conversation, manual adjustment, authority, intent, failure, or local workarounds. Several observations may support more than one reusable explanation, and a coherent explanation can still be wrong.

The practical gap lies between two existing results. A.15.1 can identify what Work occurred and which Method it enacted when that claim is independently grounded. A.3.1 can identify a reusable U.Method once its meaning, scope, and limits are supportable. Neither supplies the several-occurrence recovery Method needed before Method admission. Without that intermediate result, projects either jump from record to Method or import a full specialist reconstruction programme into every case.

#Forces

#Solution

For every materially different reusable way that still fits the evidence, write a separate candidate account. Treat each account as one U.Episteme, keep it provisional until A.3.1 admits the candidate as a Method, and keep MethodDescription membership separate until the account concerns that admitted Method and passes A.3.2.

#Identify each candidate account truthfully

Each positive candidate account is one U.Episteme. Before Method admission, distinguish the possible reusable way as one provisionally identified candidate entity: give it a working designation and enough source-supported participant meanings, applicability hypothesis, intended result or preserved condition, variation, and limits to tell it from rivals. That candidate entity is the account's EntityOfConcern.

The account's effective U.ReferenceScheme states only the designation and interpretation rules needed to read source terms such as performer, activity, cue, event, result, and stop. Add measurement or comparison rules only when the account uses them. This provisional entity identification admits neither a U.Method nor a U.MethodDescription.

When two or more candidate reusable ways remain symmetric, return one account episteme per candidate. Each account may mention the rivals while retaining its own EntityOfConcern. If a later use needs a comparison that can be retained or reused, use A.22 to select the candidate subjects and the comparison relations that already hold and together define one comparison structure, then return a separate episteme about that structure. Otherwise compare them in ordinary working prose. Several candidates do not by themselves form one subject for a combined account.

If no candidate entity or truthful effective scheme can be recovered, lower the result rather than fabricating an account.

#Run the nine-step recovery Method

1. State the receiving use and useful grain. Say why a reusable Method is being sought and which later action or decision would change. Do not reconstruct a fine sequence when the use needs one broad way, or a broad routine when a safety-critical branch must remain explicit.
2. Name several grounded occurrences or other direct evidence. For every claimed Work occurrence, recover each precise performer's A.13 core and independently admit the Work under A.15.1. Add F.6 only when the candidate account also needs precise assignment-bound attribution. Keep sources such as videos, logs, notebooks, interviews, artifacts, measurements, and assertions as separate entities or epistemes with their actual evidence relations. One occurrence may open a hypothesis; it does not establish reusable applicability.
3. Write what each source supports. Keep a readable source-to-claim account of performer Systems, relevant facts, enacted-Method claims when independently grounded, actions, cues, variations, results, and stops.
4. Expose how evidence was constructed and what it misses. State which performers, objects, successful, failed, or atypical occurrences were observable and which contributions—such as embodied perception, conversation, manual adjustment, discretion, or tacit know-how—may be absent. For event data, name the preparation Method, relied-on description or configuration, source data, dated preparation Work, resulting event-log episteme, the identified event-data collection or structure that the log describes, and the interpretation scheme. If a relied-on Method, configuration, source, correlation key, event-state encoding, or observation window is unavailable, return that limit before mining.
5. Distinguish each candidate subject. For every materially different possible reusable way, state the provisional identity and scheme from §4.1. If two candidates cannot be told apart without unsupported claims, retain the ambiguity or lower the result.
6. Write one account per candidate. Ask the A.3.1 questions without granting Method membership: applicability, participant meanings, preconditions, intended result or preserved condition, reusable actions, supported parts or interfaces, allowed variation, and stops. Mark every unsupported position unknown rather than filling a familiar template.
7. Compare stability, variation, and alternatives. Ask what recurs across independently grounded occurrences, what changes with the situation, what may be a performer-specific habit or local workaround, and whether another account explains the same evidence. Frequency alone establishes neither a Method part nor its value.
8. State a held-out or distinguishing question. Name one representative occurrence, trial, comparison, or additional source not used to shape the favored account, and the observation that would support, separate, repair, or lower the candidates. Make the test proportionate to the receiving use; it is not automatically an effectiveness trial.
9. Return the strongest honest result. Return one or more candidate accounts ready for A.3.1 identification or specialist work, with a separate comparison only when needed; or lower to a Work-related record, local regularity, performer-specific habit, observed sequence, or unresolved cue. Prepare MethodDescription-authoring input separately. The same account can qualify as U.MethodDescription only after its EntityOfConcern is admitted as one U.Method and its claims pass A.3.2.

#Select the result branch

#Keep process-mining contributions separate

Treat evidence preparation, process discovery, and candidate-Method recovery as three contributions.

1. Named data-preparation Method or Methods select source events, name activities, correlate records and objects, choose event-state or start/complete encodings, and apply abstraction. Dated preparation Work uses named source data and produces an event-log episteme about the identified event-data collection or structure described by the log.
2. A named discovery Method may return a behavioural-model episteme. Conformance checking may compare a log with a separate descriptive or normative model. Enhancement may add timing, organizational, performance, or prediction claims. Object-centric mining may preserve several typed objects and qualified relations.
3. This recovery Method uses those well-scoped results with other evidence to return candidate reusable-way accounts, unrecorded contributions, honest lowering, and a distinguishing trial.

None of the earlier contributions automatically recovers a Method. A discovered process model remains a U.Episteme about selected evidence unless another rule establishes a different kind or use. A conformance result relates a log and model; it does not prove that the model describes the obtaining Method or that every deviation is defective. Executability and visual process form do not satisfy A.3.2.

Treat process, actual process, case, activity, event, variant, deviation, and process model as cues to recover the direct subject, not as types supplied by the words. Use E.10 and A.15.6 when the source remains ambiguous.

#Stop or continue to specialist Method Engineering

Stop here when the receiving use needs only a source-traceable candidate account, an honest comparison, or a record-only result. Continue to specialist ME.18 when the domain and consequence require a reconstruction programme—for example, sampling across performers and settings, interviews, cognitive task analysis, ethnography, protocol analysis, process-mining design, artifact analysis, tacit-contribution recovery, fragment composition, domain trial design, or stronger assurance.

ME.18 may strengthen the candidate accounts and prepare separate inputs for MethodDescription authoring. Use A.3.1 for Method identification and A.3.2 for MethodDescription membership.

#Archetypal Grounding

#Pump-inspection recovery

Four independently grounded pump-inspection Work occurrences have video, sensor logs, technician notes, and known performer assignments. Three show the same inspection order. The fourth begins with a vibration cue and reverses two checks.

An analyst applying the recovery Method distinguishes two possible reusable ways under the plant's current inspection vocabulary: a fixed order with an undocumented exception, and a cue-responsive order. Each candidate gets its own account episteme, candidate subject, interpretation scheme, and source-to-claim support. The account notes that the video misses a tactile check named in interviews and that successful outcomes alone do not distinguish the candidates.

A fifth occurrence is held out. Whether the technician changes order when the vibration cue is present can separate the accounts. Until then, both remain candidates; neither trace nor account is a MethodDescription.

If that held-out occurrence and the remaining evidence support the cue-responsive account while the fixed-order rival no longer fits, recovery can return one candidate account ready for the A.3.1 identity test.

#Process-mining replay

The same team names its event-data preparation Method: select inspection start and completion events from the source files, name activities under the plant vocabulary, correlate records by pump and maintenance visit, and collapse duplicate sensor bursts under a stated rule. It cites the selected configuration because that configuration affects the result. Dated preparation Work on the named files produces an event-log episteme about the selected event-data collection.

If the correlation key, configuration, or source window cannot be recovered, the result is that limitation—not a raw-fact log. A separately named discovery Method returns a behavioural-model episteme with observed variants. Candidate-Method recovery then adds the two candidate reusable-way accounts, the missing tactile contribution, the record-only branch, and the fifth-occurrence question.

#Record-only lowering

Three timestamped records show that one operator checked A before B on three shifts, but the performer assignments, applicability, source window, and purpose of the sequence cannot be grounded. The useful result is an observed sequence in those records and a list of missing facts. It is not a candidate Method account, Method, or MethodDescription.

#Bias-Annotation

• Automation bias: admitting a Method from a mined or executable model without testing the reusable way under A.3.1.
• Frequency bias: repeated order is evidence to examine, not a reusable Method part by count alone.
• Success bias: favorable outcomes without failed or atypical cases may hide the real limits of the reusable way.
• Record bias: unrecorded perceptual, embodied, conversational, and discretionary contributions remain possible gaps.
• Single-tradition bias: process mining, routine dynamics, interviews, and Method Engineering each expose different evidence limits; none alone is sufficient for complete reconstruction.
• Template-completion bias: unsupported account positions remain unknown rather than being completed from familiar practice.

#Conformance Checklist

• CC-A3.1.MR-1 — Receiving use and grain. Is the later use clear enough to choose the useful reconstruction grain?
• CC-A3.1.MR-2 — Grounded evidence. Are claimed Work occurrences and other sources identified under their own patterns and relations?
• CC-A3.1.MR-3 — Source-to-claim trace. Can the reader see which source supports each account claim?
• CC-A3.1.MR-4 — Evidence construction. Are selection, naming, correlation, abstraction, configuration, window, and likely missing contributions explicit when they matter?
• CC-A3.1.MR-5 — One candidate per account. Does every candidate-account episteme have one candidate reusable-way EntityOfConcern and effective scheme?
• CC-A3.1.MR-6 — Rival retained. Is the main competing account or unresolved ambiguity visible?
• CC-A3.1.MR-7 — Held-out question. Is there a proportionate observation or trial that could distinguish, repair, or lower the candidates?
• CC-A3.1.MR-8 — Honest result branch. Does the result stop at candidate account, separate comparison, record-only result, or named blocker without granting Method or MethodDescription membership?
• CC-A3.1.MR-9 — Specialist exit. Is ME.18 used for complete reconstruction only when the receiving use needs its larger burden?
• CC-A3.1.MR-10 — Plain use. Can a cold practitioner explain the candidate, evidence, rival, gap, and next test without reading a predicate inventory?

#Common Anti-Patterns and How to Avoid Them

#Consequences

#Rationale

This pattern fills the smallest transdisciplinary gap between evidence about Work and identification of a reusable Method. A candidate account is an episteme about one provisionally distinguished possible reusable way. Method identity and MethodDescription membership remain later, separate judgments.

The one-account-per-candidate rule protects episteme subject truthfulness when evidence underdetermines the reusable way. The record-only branch protects utility when evidence is too weak. The specialist exit keeps ordinary recovery usable while preserving the larger evidence burden for domains that need it.

#SoTA-Echoing

Qualification and smallest reopen. Reopen only when a source materially changes an evidence limitation, the multi-object recovery choice, or the boundary between reconstruction and an admitted Method used by a result branch. Revise the affected source row and its matching recovery step, case, or checklist item. A new mining algorithm, serialization, or domain example alone does not reopen the general boundary.

#Relations

• Builds on: C.2.1 for each account episteme, its EntityOfConcern, and effective scheme; A.3.1 for the questions that shape a candidate without granting Method membership; A.13 for each precise performer's local agency core; A.15.1 for independent admission of grounded Work occurrences; F.6 only for a current precise assignment-bound attribution; and A.10 for bounded source reliance.
• Coordinates with: A.3.2 for later MethodDescription membership; A.15.6 for recovery of ambiguous process or case wording; A.22 for a separately selected comparison structure only when a named use needs it; and C.32.MWA for prospective several-structure practice synthesis.
• Receives bounded evidence from: process-data preparation, discovery, conformance, enhancement, object-centric mining, interviews, observations, artifacts, and measurements under their own Methods and claims.
• Hands off to: A.3.1 for Method identification or specialist ME.18 for complete reconstruction; neither continuation is automatic.
• Keeps separate: source, record, event log, behavioural model, Work, candidate reusable way, candidate account, admitted Method, MethodDescription, and any comparison episteme.

#A.3.1.MR:End

---

Last Updated: 2026-08-26 — this section last modified in upstream FPF commit c7ac61bb (github.com/ailev/FPF)