Relational Information Framework (RIF) v1.5 — Measuring Relational Information Through a Decoder
Year: 2026
Type: preprint
Abstract: Relational Information Framework (RIF) defines “information” as the rarity of a reading relative to a specific decoder holder and its background: the same sound wave is a warning to a local, noise to a tourist, and an input awaiting fitting to an untrained translation model. Shannon’s measure gives all three the same number; RIF asks how rare this reading is for this holder, and what time and energy the holder spends reading it. It distinguishes structural information—the rarity of an object’s own organization against a given background—from causal information—the rarity of an effect reading against the baseline where the effect would also appear under control conditions. Each class is paired with a persistence time and an energy cost, forming the I–T–E triplet. A decoder holder must maintain an identifiable boundary, extract environmental differences selectively, undergo a state update triggered by that extraction, and have that update matter causally for its own persistence; instruments used only for measurement, and public reading protocols, do not qualify. The measurement layer presupposes no dynamics. Five commitments are already executable: the habitat background Benv is the default and the protocol baseline Bmax must be declared; boundary identification is a four-step procedure delivering a candidate pool plus rejection rules, not a unique solution; causal determination first identifies the effect by control design, then reports rarity with an interval, with equivalence tested by two one-sided tests (TOST); holder-level costs are reported per holder and aggregated only when four explicit declarations are complete (§13.4); and every “typical value” carries its statistic, conditions of validity, and number. Several operationalization questions remain tagged as candidate rules or hypotheses and are registered under “Limitations and Open Problems”; the book discusses “how to read”, not “how the world must be”.
Author: LiangRuifeng
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236695498
