Our approach
Keep the history readable. Keep the person in view.
Trace-informed AI considers the whole interaction: the model, context, tools, records, interface, and the people affected by its use.
Different parts of an AI interaction.
The model
A trained computational component that transforms inputs into outputs.
Why the surrounding system mattersThe system
The model together with instructions, tools, memory, software, operators, and its environment.
See the design aimsThe interface
What a person can see, choose, correct, refuse, and carry forward.
See the design aimsThe record
Sources, observations, decisions, transformations, and later corrections, kept distinguishable.
Read the foundationsWhat this means for Branchline.
Context has a history
A summary should have a readable relationship to the material it came from.
Permission has a source
A fluent recommendation does not supply the user’s consent or authority.
Correction stays possible
A useful record preserves uncertainty and allows revision without rewriting earlier history.
The interface serves the person
What a system shows, hides, remembers, and interrupts changes the practical experience of using it.
Clear claims, with room for open questions.
PAL records earned distinctions and trace within its declared scope. It is not a final theory of physics, consciousness, ethics, or ontology. PECAN, PEA, and SEED retain their separate roles.
Machine consciousness remains unresolved. Generated self-description is an output to examine, not independent testimony that settles the question.