A knowledge base should support fluid resolution-switching

Type: kb/types/note.md · Tags: foundations

A reader using a knowledge base alternates between broad views that expose possible routes and narrower views that supply detail. This note calls movement between those views resolution-switching. Resolution here means the scope and detail exposed by a view, not an intrinsic abstraction rank assigned to an artifact.

A knowledge base should make task-relevant switches inexpensive in both directions. A broad view should lead to specific claims or evidence. A detailed artifact should leave a discoverable route back to wider organizing context. This is a qualitative design criterion, not yet a metric or a claim that every task needs the same sequence of views.

Commonplace mechanisms

Titles, descriptions, and bodies expose different amounts of detail. A claim title carries the assertion. A description adds a fixed routing cue. The body supplies the mechanism, qualifications, and evidence. Because a title-as-claim supports traversal as reasoning, a reader can inspect a chain of assertions before choosing which bodies to load.

Indexes and notes expose different scopes. A curated index groups claims and explains their roles in a topic. Following an entry reaches the narrower claim and its support. Returning to the index restores the topic-level view. Search supplies another broad entry route, while local link-following uses the reader's current context. Those routes have different metadata requirements, but both can participate in a resolution switch.

Link context preserves bearings during a switch. A contextual phrase says why the target matters in the current argument. It lets a reader move to a different scope without reducing the connection to an untyped jump. The link's relation does not by itself determine which artifact is more abstract; its role is to preserve the task-relative relationship across the transition. This is the navigational value of encoding link strength in position and prose.

Progressive disclosure supplies intermediate views. Titles, descriptions, contextual link phrases, generated summaries, and full bodies can form a cost gradient. A reader first inspects a cheaper pointer and loads a more expensive view only when needed. The gradient is useful only when each pointer supports a sound next-read decision; pointer types trade off specificity, cost, availability, and accuracy.

The evaluative criterion

Resolution-switching fluidity asks whether a reader can move among the views a task needs without paying avoidable transition cost or losing the relationship between them. It complements retrieval accuracy: a system may retrieve the right detailed artifact yet make its wider context hard to recover, or offer an elegant overview that never reaches the required detail.

The criterion prompts four questions:

  • Can a reader choose a narrower target from the broader view without opening every candidate body?
  • Can the reader reach the needed detail without loading unrelated material?
  • After opening the detail, can the reader recover a relevant broader view through an available route such as an authored link, index, or search?
  • Does the transition preserve why the two views belong together for this task?

These questions identify possible friction. They do not yet supply a scoring rule or universal threshold.

Connection to discovery

Recognizing shared structure requires more than navigation. Broad views can place several candidate claims in one context, while detailed reads can expose the mechanisms needed for comparison. Resolution-switching helps a reader assemble those views; it does not perform the recognition or guarantee a discovery.

Boundaries

Resolution-switching is not the same as addressability grain. Grain sets the smallest amount one access path can retrieve for a matched target; resolution-switching asks whether useful views at different scopes exist and whether the reader can move among them. It is also distinct from pointer accuracy and retrieval recall. Those properties can enable or defeat a switch, but none alone establishes bidirectional navigability.

An artifact does not need an outbound link in every case. Backlinks, an index, or search may supply the outward route. Nor does every task need to return to a broad view after reading detail. Fluidity is relative to the task and to the navigation operations actually available to the reader.

Open questions

  • Which cost should a measurement use: navigation steps, bytes or tokens read, inference calls, or time to a task-relevant view?
  • How should a test distinguish a genuine dead end from a route intentionally supplied by search or backlinks rather than an authored outbound link?
  • How does the useful number of intermediate views vary with the task and the reader's available operations?

Relevant Notes: