Skip to content

Companion to Mind the Gap

BRIDGE language companion

By Robert Wescott · October 2026 · Public · Design proposal 0.3

Purpose

BRIDGE explores whether a shared combination of words, relationship maps, and expandable shorthand could reduce the effort of human–AI communication. It is presently a proposed notation and interface grammar, not a complete constructed language. Its speed and comprehension benefits have not been demonstrated. The gains in the fictional extra are invented.

The design should help a person locate a consequential relationship without reading every statement in sequence. Compression works only for meanings the participants already share. Unfamiliar concepts still require explanation, examples, and learning time.

Message grammar

subject | statement | basis | request | authority

The statement remains ordinary language. The other fields identify evidence, uncertainty, requested action, and who may authorize it. Unspecified values stay unknown unless an explicit shared default applies. Definitions and defaults have versions that users can inspect.

FormMeaning
=Asserted claim; does not certify truth
?Unknown or uncertain
~Estimate; include range and basis when available
->Claimed or predicted consequence; label its uncertainty and source
IF[...]Condition under which a statement applies
REQUIRES[...]Named prerequisite; its status is separately recorded
@Source reference or defined shared context
!Request for an action
ACK[...]Receipt of a named message
UNDERSTAND[...]Person reports understanding a named message
AGREE[...]Agreement with a named claim; grants no action authority
YES[...]Explicit permission for named action, limits, and duration
NO[...]Refusal within named scope
WAIT[...]Request to pause named action; system must report whether it paused
EXPAND[...]Request ordinary-language detail
DONE[...]Completion claim, with result evidence; independent of authorization

Relationship view

Statements can appear together as a small map or table. Each relationship has its own source and status. The visual view does not silently turn association into causation, an estimate into a fact, or a suggestion into an instruction. Labels remain readable without color, and an equivalent linear reading order must be available.

Example: a fictional cooling proposal.

Claim or conditionRelationshipStatus and source
Proposal reduces electricity useDepends on operating at 8°CSupplier estimate; proposal section 3
Product A can be stored at 8°CRequired for proposed operating planContradicted by product specification: maximum 5°C
Products could be relocatedPossible alternativeCost and capacity unknown
Purchase should proceedRequested decisionNo authorization recorded

A compact reply could be:

cooling purchase | WAIT[purchase; storage conflict] | @ product-A/spec-v4 | ! assess relocation cost | authority: Helen, purchasing only

The reply requests a pause and an assessment. A separate status must establish whether the pause took effect. Reading it authorizes neither purchase nor relocation.

A user can expand any node to see its definition, evidence, excluded cases, alternatives, and uncertainty. A system should offer brief, example-based, and full explanations rather than merely replacing a short obscure phrase with a long obscure paragraph. Expansion time belongs in any performance measurement.

Boundaries

Receipt, understanding, agreement, authorization, and completion remain separate. Silence grants nothing. A consequential decision needs adequate opportunity to consider alternatives; faster delivery does not establish readiness to decide. A source link supports inspection, not automatic credibility.

Access must not depend on learning the notation. Offer ordinary language and equivalent ways to communicate, including a person's preferred assistive method and human support where needed. Record assistance distinctly from the person's expressed choice. The proposed grammar grants no legal authority and does not replace an appropriate consent process.

The interface should show both machine response time and the time available to the human. It should never infer free capacity from faster reading alone. Waiting, reflecting, discussing with another person, or remaining undecided are valid states.

Test before claiming a benefit

Compare the same tasks in ordinary English and BRIDGE after a defined learning period. Measure time to a correct decision, errors, delayed recall, and unintended authorization. Count training, expansion, reflection, and correction time. Include unfamiliar topics, different access needs, and adversarial interfaces that encourage quick responses. Test whether institutions actually honor pauses and refusals.

Test single decisions and sustained sessions separately. Finishing one decision faster does not establish that a person can make proportionally more decisions all day. Assess users' control over the time recovered, not only institutional throughput.

The language card

A separate BRIDGE language card provides a short reference and an example.

What is claimed? __________

What supports it, and what remains uncertain? __________

What depends on what? __________

What action is requested, and who may decide? __________

My response: receipt / explanation needed / agreement / permission / refusal / pause

What happened after my response? __________

Understanding does not grant permission. Permission does not prove completion. A request to pause is not confirmation that the action stopped.

Further development

The next useful step is a small interactive prototype with linked evidence, visible conditions, and expandable explanations. An actual speed claim should follow testing. A real deployment would also need research and review appropriate to its setting; this document makes no factual claim about a current clinical or public-sector implementation.

Sources & Record

Source notes checked October 15, 2026. These references inform the design discussion; they do not validate BRIDGE or document the fictional events.

  1. Progressive disclosure

    Jakob Nielsen · Nielsen Norman Group · December 3, 2006

    Describes making secondary interface options available on request. This is a useful design precedent for expandable explanations, not evidence that a new notation increases human communication speed.

  2. Time to understand

    W3C · Understanding WCAG 2.2 · Timing Adjustable

    Explains the need to let users turn off, adjust, or extend many interface time limits, including accessibility considerations. BRIDGE’s proposed protection for reflection goes beyond this specific criterion.