3.1Who can hear the question
In a fictional community workshop, two versions of the same question reach the same timetable: 'What time does R-4 leave on Fridays?' and 'What time does the harbour ferry leave on Fridays?' Stipulate that R-4 names this harbour route and that both enquiries concern the same Friday schedule. What differs is not the fact but the reader's ability to place it. A member who has used the route code can read the first question exactly as the second; a member who has not encounters an opaque token rather than a route. The information difference therefore lies in the relation between question and reader, not in the wording alone.
Context should consequently be a declared component of the editorial task. Bender and Friedman propose documenting linguistic datasets to clarify populations, generalization and possible exclusion (Bender and Friedman, 2018). This study's task-minimal rule is likewise editorial: collect only what a reader must know to interpret the question or the answer, and note where that knowledge comes from when its origin matters. The workshop might record that R-4 is the printed code for the harbour route; it need not reconstruct any participant's history with the ferry.
Context can nevertheless be invoked selectively after an unwelcome answer. Declaring the relevant assumptions before evaluating that answer makes such changes easier to inspect, but cannot prevent every selective interpretation. A later discovery may legitimately require revision. The editor should then identify the new evidence and explain what it changes, preserving the earlier account. For this comparison, the useful addition is the route code and its stated meaning. A request for participants' personal histories would add no necessary premise. Knowing that R-4 names the harbour ferry helps a reader identify the question; it does not justify any particular timetable change.
An unfamiliar code need not put a reader outside the question's intended audience. Continue the R-4 example with the two formulations attached to the same Friday timetable. A reader initially unable to identify R-4 can learn that it means the harbour ferry and then use the same departure information as someone who already knows the code. The explanation changes what the reader can do with the wording. It need not change which sailing the question concerns or introduce a second question.
Familiarity at the moment of reading and usability after explanation are different things. A question may name a route code and still serve readers who learn that code from the answer's gloss, while a question that assumes unexplained shorthand may fail even readers who travel the ferry daily but know the route by another name. The explanation should therefore supply the reference points needed for the stated use, unless the surrounding material already supplies them. Unfamiliarity is a signal to check that requirement, not evidence of exclusion.
A gloss can still fail. If the workshop uses R-4 for one route while the attached timetable uses it for another, expanding the abbreviation will not reconcile the conflict. The editor must first identify which designation belongs to which record. Conversely, where a nearby legend clearly defines R-4, repeating it in every question may add little. These cases distinguish a missing explanation from a disagreement between references. The proposed practice is to make the necessary explanation accessible and check it against the intended use. It offers a reason for expecting the wording to be usable, rather than guaranteeing that every reader will understand it.
3.2What interlocutors already share
In the fictional community workshop reviewing ferry-timetable questions, interpretation runs on background. Some background is explicitly acknowledged in the record; other items may be evidenced indirectly, uncertain or simply assumed by an editor. These statuses should be kept apart. Suppose a participant asks whether the early sailing is workable. If the minutes show the group already named the maintenance shift as the current purpose, that is acknowledged background. An editor who assumes everyone lives near the pier is supplying something else, and the difference matters as soon as the question is read.
Clark and Brennan's interactional account (Clark and Brennan, 1991) supplies no record of what these particular workshop participants accepted. For this workshop record, the editor should identify the actual exchange that supports a claim of shared background. A relevant confirmation may support it; mere repetition of the same words may not. Two cautions follow from this study's editorial position rather than from that source. Sufficient understanding for a purpose is not agreement to whatever action the question invites. And a background item treated as shared for one purpose may be open for another, so the declared purpose stays attached to the entry.
The maintenance-shift example also shows why a useful record need not list everything participants might know. It can point to the exchange in which the purpose was named and distinguish that evidence from the editor's unsupported assumption about where people live. If a later message supplies a missing fact, its status depends on the evidence, not simply on arriving later. It may establish a correction, support a tentative interpretation, or leave the issue unresolved. The editor's task is to make that status visible. Explicit confirmation is especially helpful when a material ambiguity remains; it is not the only possible evidence for interpreting an ordinary exchange.
A clarifying exchange can settle the object of a question while leaving the answer disputed. Suppose a workshop participant asks whether the last Saturday crossing has moved. A reviewer initially treats this as a question about frequency. The participant corrects that reading: the issue is the departure time of the final sailing, not the number of sailings. The reviewer restates that distinction, and the participant confirms it. This is evidence that the parties have aligned on a particular inquiry. It does not show whether the departure time changed, whether either person supports a change, or whether the proposed timetable is acceptable.
This precision also bounds what counts as common ground. The task remains limited to context tied to the identified ambiguity. Other context may also support a reading, but the editor should identify that basis instead of reporting confirmation that never occurred. When a question turns on whether "weekend" includes Friday evening sailings, an assumption that everyone knows the ferry's operating convention is not the same as a record in which participants have said so.
The practical application follows from these limits. Where the reply is specified and the targeted misunderstanding is named, the editor may record understanding as provisionally achieved, conditional on the record's own terms. If the exchange leaves a material ambiguity unresolved, further context should be requested only when it bears on that ambiguity. The record can identify the repaired misunderstanding without turning that limited confirmation into assent to a timetable change.
3.3Time windows and changed stakes
Three temporal references can matter to a question about a changing service, and conflating them can distort the answer. The first is the date on which the question was written; the second is the service period the question refers to; the third is the date by which some decision must be made. These may coincide in a hastily worded enquiry, but nothing requires them to. An editorial task that records all three can say precisely which window it is answering about, instead of treating a single undifferentiated "now" as the only relevant time.
In the hypothetical workshop, a member writes on 10 January asking whether the 07:15 ferry will run on Saturdays in March. The member must decide on travel arrangements by 20 February. The writing date, the intended service period and the decision deadline are now explicit and distinct. A report about January sailings would not answer the March question merely because it was accurate when the question was written. If a March revision is announced on 15 February, that notice may matter to the pending choice. The reason is its bearing on the intended service period, not simply that it is newer than the original enquiry.
These distinctions help an editor locate the gap in an answer. The intended service period identifies which timetable applies; the deadline identifies when the answer must be useful for the stated choice. An accurate March timetable supplied after 20 February may remain informative while arriving too late for that decision. Conversely, a timely answer about January may miss the requested period. If no decision deadline is stated, the record should leave it unspecified rather than inventing one. Later information may revise the interpretation, but the date and reason for that revision should remain visible alongside the original question.
An answer can remain accurate for the period it describes and still be unsuitable for a later decision. In the hypothetical community workshop that reviews questions about a ferry timetable, suppose a question about Saturday sailings is answered with reference to the summer service schedule. That answer correctly reports the schedule in force. If the next decision concerns a changed winter schedule, the summer answer may no longer supply the needed information, while remaining accurate about summer. Suitability is a relation between an answer's dated scope and the period a reader now cares about, not a property the answer carries unconditionally.
The comparison therefore has three terms rather than two: the service period the question refers to, the dated scope of the answer, and the service period relevant to the later decision. If the earlier answer does not cover the relevant winter service, the editor should identify the mismatch and consult evidence for that period. That mismatch does not make the question false; the interrogative requests information rather than asserting the timetable fact. It also gives no reason to rewrite the historical summer answer as though it had reported the winter schedule. The record of what was reported about summer sailings is itself information, and preserving it keeps the timing of the original judgment auditable.
Reuse therefore calls for a scope check. A summer answer that also explicitly establishes an unchanged winter schedule may still be adequate for the later query. An answer confined to summer cannot acquire that wider reach merely because someone copies it into a winter discussion. Nor does preserving it as a historical record endorse its use there. The editor can retain the dated answer, identify the new service period, and state which further evidence supports or defeats reuse. That preserves continuity without confusing an accurate past report, a current answer and a timely basis for action.
3.4Institutional roles and expected answers
In a hypothetical community workshop that reviews questions about a ferry timetable, the request and the documented remit of its recipient together constrain what an authorised answer could do. Asking the information desk for the published hours and asking the scheduling committee to change those hours are not two phrasings of one request. The first seeks a report about what has been announced; the second seeks a decision that alters what will be announced. Treating them as interchangeable invites either an answer beyond the respondent's remit or a refusal that reads as evasiveness when it is only a boundary.
Let the roles be stipulated as follows within this hypothetical setting. The desk clerk is authorised to read the current published schedule, to note any posted revision notices, and to direct further enquiries elsewhere. The clerk is not authorised to promise departures, adjust times, or speak for the committee's future decisions. The scheduling committee is authorised to review proposals, weigh operational constraints, and adopt timetable changes according to its stated procedure. It can report what it has decided and what remains pending, but it cannot treat a fresh question as though it were already a decision. The remit, not the title or the personal goodwill of its holder, sets the available answer types.
The stipulation distinguishes an authorised response from a predicted one. It does not guarantee that a clerk will answer accurately or that the committee will consider a proposal promptly. The desk can supply a report within its remit; the committee can report a decision or explain that a proposal remains pending. Even an authorised committee may refuse a requested change. An editor who records these boundaries can distinguish a request for information from a request for institutional action without treating either as already fulfilled. Routing should follow the relevant remit, including any stated referral procedure, rather than assuming that the person who reports published hours also publishes or controls them.
A role label can suggest where to look for a remit, but does not by itself establish the requested answer, a speaker's private intention or an institution's powers. In the hypothetical community workshop that reviews questions about a ferry timetable, the useful entry is therefore the documented boundary of the role: what the role is recorded as authorised to answer, and nothing further. If the record does not state a remit, the entry stays open rather than being filled by inference from the title alone.
Apply that limit to the same visitor's request. The published remit says that the receptionist may relay announced times, while the committee may consider timetable changes. A request to change the last departure therefore exceeds the receptionist's recorded authority. It does not follow that the receptionist personally opposes the change, that the committee will grant it, or that no other office has a relevant role. Those would be additional claims needing additional evidence. If a later document delegates limited change authority to the desk, the routing account must be reconsidered. The title could remain the same while the documented powers changed.
For editorial scope, then, record three items: the question asked, the role addressed, and the documented answer boundary attached to that role. Where the boundary is not documented, mark the inference conditional — for example, if the receptionist may only relay published times, the request needs referral to a role with the relevant authority; the destination remains uncertain until that authority is identified. The reason for referral can be stated without assigning motives to either office. Later evidence may expand or narrow the documented boundary; until then, powers not established by the record remain uncertain.
3.5Context that cannot be collected
When an editorial task requires context, the relevant scope is the context needed to interpret or answer the stated task, not the contributor's entire personal history. Consider a fictional community workshop that reviews questions about a ferry timetable. A question such as "Will the morning ferry run?" is ambiguous between several service periods, because a timetable may differ by weekday, season, or date range. If the possible periods would produce different answers, an editor needs to locate the intended period — the span of dates or pattern of service to which the question refers. An accurate schedule for one period would otherwise leave open whether it addresses the intended one.
Locating the period is a bounded operation when the answer varies with it. In this stipulated timetable task, a contributor's full residence, employment, health, financial or family history is unnecessary. A detail from such a history might sometimes indicate a travel date, but that possibility does not make the whole history a justified input. A direct question about the relevant date or schedule is more closely matched to the uncertainty. If another purpose emerges, such as asking about access to a particular sailing, the editor should identify that changed task before requesting further information. Relevance must be argued for rather than inferred from the fact that information is available.
The proposed limit is a reasoned connection between a missing fact and the comparison it would change. Declaring that connection makes the request easier to challenge, but does not make collection compulsory or prove that the declaration is correct. If the intended service period cannot be obtained, the answer should remain conditional — stating which comparison holds under which period and what stays uncertain — rather than being rescued by unrelated personal details. The missing date can be named precisely without making the contributor's biography the default object of investigation.
A conditional answer can preserve useful information when a missing period cannot be identified. Suppose the workshop receives 'Is a Saturday morning ferry listed?' and cannot reach the contributor. In this invented example, the available timetable lists a Saturday morning crossing for summer but none for winter. The record does not establish which period the question concerns. The editor can state the summer and winter answers with their conditions, while leaving the intended period unresolved. Neither answer should be reported as the contributor's confirmed meaning.
The remaining uncertainty is specific: which of the two service periods is relevant. A broad request for employment, household or health information would not be justified merely by that gap. A narrowly relevant message may later identify the date, but the editor should explain how it bears on this question. The conditional response also has a limit. It cannot tell a traveller which answer to act on until the relevant period is known. It can, however, prevent an undated yes or no from being mistaken for a statement about all Saturdays.
An editor should also check whether the uncertainty matters to the answer at all. If both candidate schedules instead listed the same morning crossing, a bounded answer could report that agreement while still marking the intended period as unknown. The ambiguity would remain in the record without blocking every useful response. Where the schedules differ, the conditional pair keeps the unresolved choice visible. If the service period is later identified, the entry can link that evidence to the appropriate branch. No invented personal explanation is needed to fill the gap, and no claim of certainty is required to make the available information readable.
3.6Disputed interpretations
An ambiguous reference should be handled by setting out plausible readings and naming the words that support each, before anyone asserts what the questioner meant. The point matters because the words on the page constrain interpretation without settling it. Consider the hypothetical community workshop reviewing questions about a ferry timetable, where one request is recorded as: Will the earlier one still take the freight? The record contains no nearby list or revision notice that fixes the comparison. The phrase "the earlier one" carries the dispute, and an editor who silently picks one reading has made a context claim without showing it.
One reading takes 'the earlier one' to mean the earlier of two departures. The comparative 'earlier' needs a comparison, and two listed sailing times could supply it. Another takes the phrase to mean a sailing moved earlier than its customary time. In that case, the former departure time would supply the comparison. The word 'one' leaves the particular sailing unspecified in the isolated sentence; the article 'the' does not by itself establish that a pair appeared on a board. On both readings, 'still take the freight' concerns continued freight carriage. It does not, without further wording, ask whether the timetable revision persists.
Evidence about a displayed pair of departures or an announced revision could help distinguish the readings, depending on what the speaker was referring to. Without such context, these two candidates can be retained with the assumptions that would support each; the pair is not claimed to exhaust every possible reading. This record should also stay distinct from other repairs: correcting a mistranscribed word, offering a new reading of unchanged text, and asking a fresh question are different editorial acts, and only the second is at issue here. A later clarification may support one candidate, suggest another or leave the choice open. The alternatives make the uncertain reference inspectable without claiming to know the speaker's intention.
A revision to a published reading does not by itself show that the earlier editor erred. Three kinds of change must be kept apart: correction of a transcription, a new interpretation of text that stands unchanged, and a question that departs from what the earlier editor addressed. The distinction matters most when revision becomes public, because a new reading can be presented as a refutation when it is in fact a reorientation. Consider the fictional community workshop that reviews questions about a ferry timetable. Suppose a transcript says "first sailing," but inspection of the unchanged original notice shows "last sailing." Correcting the transcript repairs a demonstrated copying error. But if the notice reads "first sailing" and the workshop now proposes to include the boat that departs before the ordinary service, the wording has not changed; what has changed is the interpretation placed on it. Calling it a correction would require evidence that the earlier interpretation was mistaken, not merely the announcement of another reading.
The cases call for different evidence. A transcription correction should identify the original wording and the mismatch in the copy. A revised interpretation should identify what supports the new reading and whether it contradicts, supplements or leaves open the former one. If the unchanged notice explicitly excludes preliminary crossings, including one as the 'first sailing' would be a poor interpretation; unchanged text does not make all readings equally defensible. A new question, such as whether an earlier crossing should be added, changes the inquiry itself. Its relation to the original timetable question can be explained without pretending that it was already asked.
A useful revision note therefore names the act performed and the evidence for it. Correcting a copied word, changing an interpretation and posing a new question may all occur in one revision, but they should not be collapsed into a single claim that the earlier editor was wrong. Nor should a genuine interpretive error be concealed behind a neutral label. The note can state what changed, what remains in dispute and which earlier conclusion is affected. This makes disagreement inspectable while preserving the possibility of warranted correction.
3.7Minimal context with revision paths
Context should be a declared component of an editorial task, not an open-ended commission to reconstruct a life. In the hypothetical community workshop, a question about the ferry timetable can be anchored by a record of four fields: the referent under discussion, the period the question covers, the task the questioner has declared, and the uncertainty already known at the time of asking. These four fields are a starting proposal for this example, not an exhaustive account of every context that could matter. Its purpose is to let a later reader identify the frame an answer was meant to serve, using no more detail than that comparison needs.
The value of such a record appears when interpretation shifts. A revision path does not rewrite the original entry; it appends what changed and which field the change affects. Suppose the workshop learns that the timetable in question was a draft rather than a published schedule. The document's recorded status is corrected, the prior entry remains visible, and the reason for the amendment is identified. Showing the old and new entries makes the change inspectable; it does not prove that the revision is justified. Nor does a small field count guarantee a small privacy burden. A single field could contain unnecessary personal detail. The content placed in each field needs a reason connected to the task, and a later editor should be able to challenge that reason.
This proposal is useful when a later reader would otherwise have to guess which frame an answer addressed. The revision note should identify the evidence for a changed field and whether the change corrects a mistake or introduces a new scope. Preserving that distinction does not require presenting every version to every reader, or claiming that an earlier answer remains appropriate. The record supports a comparison between versions; it does not decide the comparison. Whether even this modest extra record is needed depends on what the exchange already makes clear.
A separate context record can be omitted when the exchange already supplies the information needed for the immediate task. Consider a member pointing to a dated winter timetable and asking, 'Is the 08:00 Saturday harbour crossing listed here?' The reviewer answers by identifying that row on the same displayed sheet. Stipulate that both people are looking at the same document and that the task is only to check its contents. A second entry repeating the referent, period and task would add no necessary distinction to this immediate exchange.
That judgment has a boundary. If the answer is later copied away from the displayed sheet, 'here' may lose its referent. Naming the sheet or retaining a link to it could then become useful. If the task changes from reading the timetable to establishing whether a crossing will actually operate, the same exchange also leaves a new evidential gap: listing a sailing is not a guarantee of its operation. These differences explain why a record can be unnecessary for one use and helpful for another. They do not establish that every ambiguity requires the same four-field form.
A specific later ambiguity can justify adding context while preserving what the earlier exchange actually established. The editor should explain that trigger, rather than silently treating a new use as the original one. Revision may clarify the same question or introduce a different inquiry; it need not resize the question every time. Proportionate recording thus requires a reason for adding information as well as a reason for leaving it out. With that limit in place, the record stays proportionate to the declared task.