The Person Changes During the Day

Priya opens the system on her phone before an architecture review and says: I have five minutes before the meeting. Just tell me what can break.

Her standing policy asks for implementation detail and equations when they affect the decision. Yesterday, at her desk, that meant a full account of the recovery path. Today the same account would consume the time she has before the meeting. But a short answer that hides the decisive qualification would also fail her.

The channel has to change what it presents without changing what it remembers about her. It can put the failure risks first, keep the qualification beside them, and leave the full derivation available. Tomorrow, when Priya asks for the derivation, today’s urgency must have no authority over that request.

These are constructed scenes, not measured outcomes. They expose the question this chapter addresses:

How should temporary conditions alter communication without becoming durable properties of the person?

Chapter 23 separated a historically true statement from its current applicability and authority. Here the distinction becomes operational. The record of an earlier policy, the circumstances of this task, and the decisions that shape this view are different objects:

durable history
≠
current context
≠
session overlay

Communication policy can persist. Communication state should usually not.

Context describes the present

Current context is the information relevant to this interaction. It describes what Priya is doing and the conditions under which communication must work. The useful inputs are narrow:

current task             device or environment
available time           explicit mode
declared objective       interaction stage
declared urgency         observable constraints

In this scene, the task is an architecture review, the available time is five minutes, the device is a phone, and the objective is to identify failure risks. “Audio only” can specify an output constraint. “I am scanning before the meeting” can identify the interaction stage. Each item has a source in what Priya says or what the situation directly shows.

Those sources still matter. A small screen is a constraint on presentation; it is not an instruction to omit a qualification. An open document identifies a source she is viewing; it does not by itself establish why she opened it.

The channel can sometimes propose a task from visible work. If Priya has two API descriptions open side by side, “Comparing these interfaces?” is a reasonable guess to offer as visible, correctable current context. She may instead be collecting examples. The guess remains marked as inferred, can shape a provisional view, and expires with its scope. It cannot override an explicit objective or supply permission for a consequential change.

inferred task ≠ inferred inner state

The chapter excludes a different kind of input:

inferred mood
inferred stress
inferred fatigue
inferred motivation
personality-state guesses

A pause can have several explanations. It does not establish stress or a need for reassurance. A request for brevity establishes a communication instruction; it does not diagnose impatience. Priya can ask for a slower explanation or a more direct answer without the system deciding what kind of person she is at that moment.

This keeps the earlier boundary intact: observable context is not inferred inner state. The task, declared objective and available constraints define the communication problem without requiring a psychological explanation of Priya.

The overlay decides what changes

Context does not itself say how the answer should be rendered. Five minutes is a time constraint. Putting risks first is a decision about communication. The session overlay contains those temporary decisions, derived from the standing policy, current context and any explicit temporary instruction:

standing policy
+ current context
+ explicit temporary instruction
          ↓
session overlay
          ↓
current communication

For Priya’s review, the distinction can be shown in a small worked example:

Standing policy
  preserve implementation detail
  show equations when they affect the decision

Current context
  task: architecture review
  device: phone
  available time: five minutes
  declared objective: identify failure risks

Explicit temporary instruction
  "Just tell me what can break"

Session overlay
  put failure risks first
  retain the qualification needed to judge each risk
  fold supporting derivations; keep them available
  scope: this pre-meeting review session

The overlay does not claim that Priya prefers short answers. It explains why this answer begins with risks. It also shows what the adaptation leaves in place: the detail still exists, and a qualification that changes the decision stays visible.

If she asks to expand one derivation, the overlay must accommodate that request. If she moves to a larger screen, the context changes and the presentation can change with it. The overlay is a current application of the inputs, not a second standing policy that acquires authority over them.

Temporary instructions have bounded authority

Priya’s immediate request can override her usual presentation default within this task. It cannot remove preservation obligations or accessibility requirements. Nor can the fact that she is hurried authorise the channel to speak or act for her.

Within those boundaries, one candidate ordering for choosing among permitted presentations is:

explicit current instruction
> explicit current mode
> observable current constraint
> visible, correctable task inference
> standing policy
> weak historical evidence
> default

The ordering is a proposed starting mechanism, not an earned result. Preservation and accessibility constrain every choice. A constraint can limit which forms are available; a task guess can shape a provisional view. Neither acquires the authority of an explicit instruction. The overlay records the resulting choices; it is not an independent source of permission.

That matters when the inputs disagree. Suppose the standing policy calls for relevant equations, but Priya asks for a short risk summary. If one equation determines whether recovery can duplicate a write, the summary must retain that equation or preserve its decisive content in an adequate form. The supporting derivation can be folded. The need to fit five minutes cannot turn a conditional claim into an unconditional assurance.

The important boundary is that a current constraint changes how preserved content is made available. It does not decide that the content has stopped mattering.

Lifetime follows the declared scope

The overlay needs an end as well as a source. “For this review” and “until this incident closes” identify different boundaries. A timer alone cannot settle either one: a review can finish early, and an incident can continue longer than expected.

Within an active session, the adaptation should end when its named task or condition ends. If Priya finishes the review and begins reading for study, the pre-meeting overlay no longer applies. If the scope is unclear, the channel should make it visible and ask when the distinction matters. “Until I say otherwise” must not quietly turn a temporary mode into an indefinite standing rule.

The session boundary provides a further limit. Closing this session does not grant its overlay authority over a later one. If an incident is still active when Priya returns, any continuing temporary instruction must have an explicit scope that covers the resumed work; the earlier overlay is not inherited as a personal default. The new view must derive its current adaptation from the conditions that actually apply.

The default route for an ordinary session condition is therefore:

current context
→ used in the current interaction
→ expires with its scope

It need not become durable memory before it can be used. Nor does it need to be stored and later forgotten to count as temporary. Chapter 23 governs deliberate retention; this mechanism supplies no additional reason to retain a phone, a time budget or a one-off request.

Shedding must include the effects

The governing rule is:

A temporary communication condition must never silently become a durable property of the person.

When the session ends, discarding a record labelled “overlay” is insufficient if the system learns from behaviour that the overlay helped produce:

temporary overlay
→ changes presentation
→ changes Priya's behaviour
→ behaviour is observed
→ learner treats it as preference evidence
→ standing policy changes later

Suppose Priya explicitly selects a deep-study mode. The channel expands explanations and keeps derivations open; she reads longer in that view. A learner later treats the reading time as evidence that she generally prefers long explanations. The overlay disappeared, yet survived through its effects.

Behaviour observed under that mode carries its provenance. If legitimately retained, it remains evidence about that interaction, including how the presentation was shaped. It can support a proposal for a scoped rule, but changing standing policy requires a separate promotion decision. Complete shedding therefore includes preventing the overlay’s effects from silently changing a standing policy.

overlay removed
≠
overlay left no durable residue

For an interaction that authorises no durable change, the required boundary is:

session ends
          ↓
overlay sheds completely
          ↓
durable history and standing policy remain unchanged

Return to Priya the next morning. She is at her desk and asks for the full recovery derivation. The system should build that view from her standing policy and today’s request. Yesterday’s phone and five-minute constraint should contribute nothing. If the derivation is still collapsed because the system learned that she likes terse views, the temporary condition has become a durable claim.

That is the chapter’s key test of the proposed mechanism. It concerns both direct changes to policy and indirect learning from overlay-shaped behaviour. The scene defines a failure to look for; it establishes no evidence that a particular implementation avoids it.

A standing rule needs a separate decision

Priya may eventually want the temporary instruction to persist. She can say: For incident reviews, make the risk summary my default. That is a new, declared policy with its own scope and provenance. It is not what happens automatically when an overlay expires.

A channel might also propose such a rule after a repeated pattern. But asking about repetition already depends on retained evidence. It may use deliberately retained, scoped records whose use is justified under Chapter 23; it cannot quietly archive every discarded overlay so that it can later claim to have discovered a preference.

If an appropriate record supports the question, the proposal can remain small:

You have asked for risk summaries in these incident reviews.
Make that a standing rule for incident reviews?
[yes]   [no]   [don't ask again]

The question must be rare, dismissible and scoped. A yes establishes a declared rule. A no means no promotion. It must not merely reduce a confidence score while the learner continues moving the policy towards the same declined rule. Repeated temporary behaviour cannot undo that refusal; a later policy change needs a new explicit decision. “Don’t ask again” also stops renewed prompts about that proposal. Neither outcome establishes that Priya generally prefers brevity or has become a different kind of reader.

Asking is a candidate mechanism, not evidence of reduced correction burden. Its purpose here is to keep the durable change identifiable as Priya’s decision. The temporary adaptation and the policy update are separate operations.

What changed this view?

The channel should be able to explain the current adaptation without presenting a theory of Priya: You asked for failure risks before the meeting. This view puts them first, keeps the decisive qualification visible, and folds the supporting derivation for this session.

That explanation identifies the source, effect and scope. Priya can correct it: Keep the derivation open for this risk. The correction applies to the current view unless she explicitly makes it a standing rule. It should not require repairing a profile assembled from temporary conditions.

The same explanation should stop applying when the overlay ends. Continuity comes from the durable record; responsiveness comes from current context; the overlay makes the bounded communication decisions that connect them.

One question remains. An overlay can make an answer fit the present task, but the easiest next action may still conflict with a goal Priya declared for longer than this moment. She may have asked to finish one paper before opening another. Making every related result immediately inviting would then work against an instruction she supplied. The next chapter asks when communication should add a pause or reminder in service of that objective, and whose decision permits it.