Abstract

This is a methodological proposal; it does not claim new physical laws.
Non-ontic Physics (NOP) is proposed here as a strict, falsification-driven
methodological layer that constrains how we translate between (i) non-numeric, non-ontic
descriptions of structure and (ii) standard (ontic) numerical modeling performed inside a
conventional discrete pipeline.
The core goal is not to replace physical theory, but to reduce interpretive latitude by enforcing
an auditable boundary: NOP specifies what may be exported into a numerical model and how failure
is named, while all arithmetic, fitting, and statistical decisions remain confined to the ontic
(DISCRETE) side.
Here ‘non-ontic’ is used in the sense of a non-numeric boundary discipline (not the ψ-ontic/ψ-epistemic debate).

1. Motivation

Many scientific workflows mix conceptual framing and numerical optimization so tightly that it becomes
difficult to audit where “meaning” ends and where “fitting” begins. This can inflate degrees of freedom,
encourage informal feature engineering, and blur the distinction between explanatory structure and
post-hoc accommodation.

Non-ontic Physics addresses this by introducing a disciplined interface that is intentionally
non-computational: it can name types, constraints, permitted exports, and failure conditions,
but it cannot compute or compare metric values. This design aims to make the bridge between concepts
and numbers falsifiable rather than rhetorical.

2. Working Definitions

2.1 Non-ontic Physics (NOP)

Non-ontic Physics is the layer that defines:
(i) a minimal symbolic alphabet for “boundary states,”
(ii) a configuration syntax over an index frame,
(iii) a rule for what structural contrasts are exportable,
(iv) a contract stating what counts as a permitted export,
and (v) a finite set of named failure conditions (“kill flags”).

NOP explicitly forbids numerical operations, optimization, probabilistic inference, or any manipulation of
metric values. It may reference metric names as labels, but not their values.

2.2 IA: Interdiscrete Algebra

IA (Interdiscrete Algebra) is the formal bridge discipline that regulates how structural
objects from NOP are exported into a discrete numerical modeling environment and how results return back
as auditable statuses and named failures.

In this framing, IA is not “AI.” It is an interface algebra: a rule system for permissible mappings between
a non-ontic description and an ontic computation.

2.3 DISCRETE (ontic side)

DISCRETE denotes the standard environment where all ontic work happens:
equations, fitting, statistical evaluation, comparisons, and baselines.
DISCRETE may produce arbitrary numerical artifacts, but NOP does not directly operate on them.

3. Boundary Discipline (the canonical separation)

The key methodological constraint is a one-way separation of roles:

• NOP/IA: names types, permitted exports, test families, and failure conditions; returns only status and kill flags.
• DISCRETE: computes everything numerical (including any error/fit criteria, residual analysis, and model selection procedures).

This is a deliberate “demarcation” mechanism: a boundary that prevents silent leakage of extra degrees of freedom
(e.g., ad-hoc derived features that were never contractually permitted).

4. Minimal export contract (value-driven, not decorative)

The practical contribution of NOP/IA is measured by whether it can constrain DISCRETE in a way that can
make a model fail. Anything that cannot possibly reduce freedom (and thus cannot possibly “kill” a run) is treated
as non-essential.

4.1 Exhaustive exports

For a fixed protocol version, the set of permitted exports is exhaustive.
Any derived feature that is not a contractual realization of an allowed export triggers a failure condition.

4.2 Named failures (auditability)

DISCRETE produces numerical artifacts; NOP receives only the names of which contractual conditions were violated.
This forces a run log to be interpretable without re-arguing “what counts.”

5. Near-term application target: RAR as an auditable testbed

A concrete near-term use case is the Radial Acceleration Relation (RAR) program, where public datasets
enable repeatable testing. The point of using RAR at MVP stage is not to claim a new physical theory immediately,
but to demonstrate that the NOP/IA boundary can run end-to-end on real data with:
(i) a mandatory baseline,
(ii) an explicit export contract,
(iii) named failure outcomes.

In this view, “success” is primarily the existence of an auditable falsification machine:
runs that can be cleanly ACCEPTed or REJECTed without protocol drift.

6. Relation to established philosophy-of-science norms (minimal note)

The program is aligned with a falsification-oriented attitude: hypotheses and bridges should expose themselves to
decisive tests rather than relying on post-hoc accommodation. NOP/IA is intended to make that principle operational
at the interface level—where many modern modeling pipelines quietly accumulate flexibility.

7. Status and outlook

This text presents definitions and boundary discipline only. The next deliverable is an end-to-end MVP run on a real,
publicly available dataset (RAR), accompanied by a compact stop-frame log that records (a) protocol version,
(b) baseline completion, (c) named failures, and (d) ACCEPT/REJECT status.

Work on the apparatus continues. Some components are not yet polished to the level required for broad public review;
updates will appear as new runs and clarified contracts rather than continuous patching inside a frozen version.
Please revisit the blog for progress notes and released run logs.

References (minimal)

1. Stanford Encyclopedia of Philosophy — Scientific Method
   (entry;
   Spring 2022 archive)
2. Akaike, H. (1974). A new look at the statistical model identification.
   (PDF)
3. (Background pointer) Wikipedia — Akaike information criterion
   (entry)
4. (Context on “reality of the quantum state”) Pusey, Barrett, Rudolph (2011/2012).
   (arXiv PDF)