New information deserves attention
A credible cancellation can matter more than weeks of preparation. A memory system that ignored recency would miss genuine changes. But the latest recorded item could also be an old anecdote, an uncertain report, or something unrelated to the present concern.
Which time matters?
Imagine a persona preparing for a meeting. Yesterday's record contains the agreed location. This morning's message recounts an amusing incident from ten years ago. The latter is newer in storage, but its subject is older and its relevance to the meeting may be slight.
Now add a current cancellation from the authorized organizer. It should change the plan. Keeping an older promise visible does not mean continuing to act as though the cancellation never arrived.
Future plans have their own sequence. An intended departure at three, the actual departure at three thirty-three, and a later comparison are different records. The comparison explains the delay without rewriting the earlier intention into punctuality.
Pause here: Is the information recent in occurrence, recording, or applicability—and how does it bear on what matters now?
Two different selection questions
PHS's temporal-memory mathematics addresses which claim should operate within one subject's history. It considers permission, effective time, status, and trust before recency breaks a close choice. When evidence is insufficient, the design allows a contested result.
The portfolio-selection mathematics addresses a different problem: which qualified memories deserve limited space for this encounter. A claim being current does not automatically make its subject the most important one to include.
Neither decision makes a selected statement infallible. An urgent warning may justify checking or taking a precaution before it is established as fact. The response should preserve that uncertainty while remaining useful.
My answer: Recency is useful evidence about change, not a universal measure of importance. A good account preserves relevant earlier commitments while recognizing credible updates, future plans, and unresolved uncertainty. Current-state selection and limited attention have related but different jobs.