After the Spill

How far should an embodied system change after unexpected experience? Research OS, Mismatch Lab, and Builder Lab by Seonhye Gu.

View the Project on GitHub higuseonhye/research-os

D001 Claim & Novelty Audit

Purpose

This document asks why the program could be wrong.

Related work is not enough. The audit must identify what existing fields already explain, what this program explains differently, and what this program has not yet earned the right to claim.

Audit fields

Existing field Overlap Difference to test Missing or unresolved
Active Inference Action and perception under model uncertainty Whether post-experience change is framed as operator choice with closure tests Formal comparison pending
Continual Learning Updating from new experience without forgetting Whether the issue is not only retention but deciding how far to change Literature matrix pending
Meta Learning Learning how to adapt Whether adaptation is selected after evidence rather than assumed as the goal Boundary pending
Recovery Planning Returning to success after failure Whether recovery, repair, and expansion are distinct operators Taxonomy pending
Resilience Engineering Sustaining function under disruption Whether embodied model adequacy can be experimentally isolated Evidence bridge pending
Self-healing Systems Automated detection and repair Whether representation expansion is justified rather than merely executed Mechanism audit pending
Adaptive Control Maintaining performance under changing dynamics Whether structural/model-class inadequacy is outside controller adaptation Formal boundary pending
Decision Theory Choosing actions under uncertainty Whether "change the system" is itself the decision object Mathematical reduction pending

Claim audit table

Claim Scientific status What would make it stronger?
Intelligence chooses the smallest sufficient change. Governing principle; not yet a general theorem Define cost, sufficiency, closure, and competing operators
Recoverability is a measurable post-mismatch window. Supported within Paper 001 task family Broader task families and operator sets
Model adequacy can justify expansion over repair. Supported within Paper 002 model-order target drift More mismatch types and real physical validation
Products can test program hypotheses, not merely demonstrate papers. Program design claim Robot Diff/Mismatch Lab user workflows and benchmarks

Rule

No novelty claim is allowed without an overlap/difference/missing row.