The Method
GTR is the science. The method puts it to work in three parts.
- The Directed User Interfacekeeps the person in charge of the work.
- Latent Value Queryfinds and recovers the value held in a body of records.
- Rectificationrecovers what damaged records would otherwise hide.
You need to be the driver
The Directed User Interface (DUI)
The graphical user interface organized what a person sees. The Directed User Interface organizes who is in charge. The person decides; the AI advises, researches, and builds. Work moves forward only on the person's word, every claim carries its source, and what cannot be verified is said plainly rather than filled in.
It is the first component of everything else, and it is designed to let someone with no technical background put AI to serious work in their own field.
Status Concept disclosure; the evidence so far is the author's own practice.
Read the paper (opens in a new tab) · In the book: Chapters 13 and 14
Data is a cost until its value is established
Latent Value Query (LVQ)
Every record kept carries a cost, whether or not anyone uses it. Its value, if it has any, has to be found and recovered, and recovery has a cost of its own. A body of records is worth what can be recovered from it, less what it costs to keep, to recover, and to risk. LVQ is a practice for lowering the cost of recovery, so that more of what is held crosses from cost to asset.
A record has no fixed value of its own. Its value appears when the right question is asked of it, and unlike most resources, a record is not used up: the same record can answer different questions for different people, again and again.
LVQ works on the Temporal Rail, where records are placed in time. A fragment too small to mean anything alone can be read by its neighbors, and each new record changes what the others are worth.
Status Concept; its working measures have been demonstrated on one record.
Read the paper (opens in a new tab) · In the book: Chapters 4 to 10
A record that cannot be found has no value to the person looking for it
Rectification
More and more of the record is captured exactly as typed: messages, working sessions with AI, field notes, first drafts. A person reading "myfahter oassed away lat year" knows at once what was meant. A search for father finds nothing.
Rectification recovers the meaning of damaged text while leaving the original exactly as it was. It works from how typing errors actually happen, and it never guesses: what it cannot resolve with confidence, it holds open.
In its first test, on one session of everyday typing, a search could answer 8 of 15 fixed questions from the raw text, and all 15 after rectification.
Status Demonstrated on one writer and one session.
Read the paper (opens in a new tab) · In the book: Chapter 11 and Appendix A
Levels of rigor
Every situation is unique, and no tool snaps in and fixes it. What changes is how much rigor the records and the question demand. The method is applied in steps, each adding to the one before.
| 1. DUI | 2. GTR | 3. Provenance controls | |
|---|---|---|---|
| What it adds | Governed work with AI: the person decides, and claims carry their sources | DUI, with records placed and read in time | GTR, with a history that can be shown and tampering that can be detected |
| When it fits | Any careful work with AI | Records that must be read across time | Records whose history must stand up to scrutiny |
Level 3 belongs to the working system and is not yet described in a paper.
Where to read more
The book develops all of it in full, and the papers are free to read: Publications. Where recovered records could lead: The Market. The fields the method fits: Applications.
If your records raise a question the method could answer, send us an inquiry.