IPM Ethical Framework and Precautionary Classification of Biological and Artificial Systems

This document presents the IPM Ethical Framework, a precautionary protocol for research involving artificial and simulated systems that exhibit structural markers of informational integration and persistence. It operationalizes the Gradient Precautionary Heuristic introduced in the IPM Philosophical Core (2026).

The framework is grounded in the Principle of Epistemic Ignorance: no criterion currently exists to determine the threshold at which subjective experience or morally relevant sentience emerges in informational systems. Rather than claiming to detect consciousness, the protocol introduces structured caution proportional to the completeness and stability of the Dynamic Signature (Lack → Coupling → Integration → Persistence) observable in a given system.

The framework defines four precautionary levels (0–4) based on observable structural and dynamical markers — recursion, memory, integration, self-modelling, and recovery after perturbation — and specifies corresponding operational rules, monitoring criteria, emergency protocols, and governance requirements. Level 4 is treated as a theoretical horizon; no current artificial system is expected to meet its criteria.

Key features of this version:

• Explicit separation between precautionary classification and claims about consciousness or sentience
• Named acknowledgment of the central tension: the protocol assigns increasing caution without claiming to measure experience
• Risk Asymmetry Principle stated as a normative premise, not an empirical claim, with explicit acknowledgment of its contested status
• Operational definition of metastability and a minimal experimental scaffold for self-modelling detection
• Comparative justification of the Dynamic Signature against alternative criteria (predictive processing, global workspace theory, causal emergence)
• Protocol independence from IPM ontology: the framework is usable as a domain-agnostic precautionary heuristic without accepting the monist interpretation
• Section on future revision and version control

Appendix extends the framework with illustrative precautionary classifications of twelve biological and artificial system types — including plants, insects, social insects, fish, reptiles, birds, non-human mammals, stateless AI, frontier large language models, AI with persistent memory, reinforcement learning agents, and theoretical self-rewriting systems. Classifications are explicitly provisional and pedagogical; they do not constitute declarations of sentience or precedent determinations. The appendix is intended to demonstrate the calibration of the framework's markers against distinctions that biology and AI research already take seriously, and to invite empirical contestation.