The Contract Travels with the Claim

A claim rarely stays where it began.

It gets shortened for a page. Pulled into a presentation. Rewritten as a headline. Translated into another medium. Retrieved by an agent. Combined with other claims. Extracted by an answer engine and presented without the argument that once surrounded it.

The language changes. The obligations do not disappear.

I have been thinking about this as contractual continuity across enactments.

The contract does not merely govern the claim at its origin. It accompanies the claim wherever the claim is used.

By contract, I mean the conditions that make an assertion responsible: its evidence, attribution, qualifications, authority, limitations, and the circumstances under which it may be repeated or acted upon.

Some of those conditions are visible in the sentence. Others live in relationships around it.

A number has a reporting period. A recommendation has a scope. A historical description has a source. A promise has someone authorized to make it. A confident sentence may depend on evidence that is current only for a particular place, audience, or version of the work.

When a claim moves, those relationships are easy to lose.

This creates two kinds of inheritance.

Semantic inheritance asks whether the new expression retains the meaning of the original claim.

Contractual inheritance asks whether it retains the obligations governing the use of that claim.

Both matter. Neither guarantees the other.

An answer can appear to summarize a claim accurately while omitting the qualification that prevents it from being misleading. The recognizable meaning survives, but the permission to use it that way does not.

This becomes especially important when systems are built to retrieve and recompose knowledge. The useful unit cannot be the sentence alone.

The agent needs a claim package.

What is being asserted? What supports it? Which version of the evidence applies? What context is required? Who has authority? What must be attributed? Which limitations must appear in the resulting expression?

When several claims become an argument, their contracts compose too. One source may permit public reuse. Another may require attribution. A third may be accurate only within a narrow timeframe. The system cannot silently discard those differences because the combined paragraph reads smoothly.

Carrying the contract is only the first step. The resulting enactment must be checked at the point of use.

Did the summary retain the essential qualification?

Does this audience change what needs to be explained?

Can the evidence support this stronger wording?

Does the chosen medium remove context the claim depended on?

If the obligations cannot be satisfied, the expression needs to change. The system may need more context, an authorized exception, a narrower claim, or a decision not to publish it.

This is where provenance becomes operational. It stops being a record we preserve after the work and becomes part of how the work is allowed to move.

External systems will not always keep the relationships we provide. A search result or generated answer may separate the statement from its source. That makes the published expression itself responsible for carrying any qualification it cannot safely survive without.

The practical test is simple to say and difficult to build:

What survived the journey, what was transformed, what was lost, and may this enactment proceed?

The contract is not documentation beside the intelligence.

It is part of the intelligence.