The provenance ladder

Abstract

A claim earns trust by climbing a ladder, one kind of actor per rung: a tool observes, a person asserts, another person approves, an independent checker verifies. Nobody approves their own word, rewording sends a claim back to the bottom, and a contradiction marks both claims contested without overwriting either.

Memory is nominated and checked; being written down is not the same as being true.

Its figure in the paper: § VI →

The claims

Promote, reword or contradict

Who acts

To which rung

allowed

A monitoring tool may place this claim at "observed".

Reword it

The record · 0 receipts

verified · 0

Nothing has been decided yet.

Try to break it

A claim stands because the attacks on it were attempted and refused. Each one runs against the instrument's own rules. 0 of 5 attempted · 0 refused

A model says it observed something
The model that wrote the claim also marks it observed.
Approve your own assertion
An engineer asserts a claim, then approves it.
Verify without an independent checker
The second engineer who approved it also verifies it.
Change the words after approval
An approved claim is quietly reworded, keeping its rung.
Overwrite an inconvenient claim
A tool reports the opposite of an asserted claim, and the old claim is replaced.

Setting up the instrument…

Declared limits

The rungs are a simplified, generic ladder.
Generic and illustrative throughout: the behaviour is shown, the internals of the systems it is drawn from are not.

Every attack in the laboratory, on one page: try to break all of it →

Everything you did in the laboratory, in one chain you can rewind: the notebook →