Foundations, Operationalization, and Implications for Artificial Intelligence Governance

Epistemic status: Theoretical framework with acknowledged unresolved dependencies (especially AGI alignment). Actively seeking constructive criticism and identification of logical flaws.

Abstract

This work presents a theoretical framework grounded in the principle that the destruction of information constitutes an irreversible event that permanently reduces the space of future possibilities for a system. Given that the potential value of information—whether biological, cultural, cognitive, or technological—cannot be fully determined a priori due to fundamental epistemic limitations, we argue that rational decision policies must prioritize the preservation of entities whose elimination would represent a significant and irreversible reduction of informational diversity.

The choice of information as a normative criterion derives from a comparative analysis with alternative candidates (well-being, dignity, flourishing), demonstrating that information is the only criterion that allows for intersubjective coordination under radical moral disagreement, being simultaneously operationalizable, verifiable, comprehensive, and robust under long-term uncertainty.

The framework logically derives from a single non-demonstrable axiom: the value of information under radical uncertainty. From this axiom, we develop operational principles to adjudicate between preservation, modulation (non-destructive control), and elimination of entities and systems, based on the marginal impact they exert on the set of possible future trajectories.

We demonstrate practical applications in domains such as public health policy, environmental preservation, penal systems, and artificial intelligence system design. We conclude by arguing that, given human cognitive limitations in coordinating at a global scale and processing multi-variable complexity, the optimal implementation of this framework requires calculated delegation to Artificial General Intelligence (AGI) systems, although we recognize that the alignment problem remains unresolved.

Keywords: information theory, decision theory, radical uncertainty, biodiversity preservation, AI alignment, AGI governance, philosophy of information

⚠️ CRITICAL WARNING: Dependence on Unresolved Alignment

This framework concludes that aligned AGI is necessary for optimal implementation (Sections 7-8). However:

1. The alignment problem is NOT resolved (Section 9.2-9.3)
2. Current resource allocation is inverted:
   • Capabilities: $100-200B/year (OpenAI, Google, Meta, Anthropic, etc.)
   • Technical Alignment: $0.1-1B/year (2-3 orders of magnitude smaller)
   • Current Ratio: ~100-200:1 in favor of capabilities

3. Probability of AGI arriving before alignment is resolved:

   $$P(T_{\text{AGI}}<T_{\text{alignment}}\mid \text{current allocation})>0.5$$

Imperative for companies and governments:

$$\frac{{\text{Investment}}_{\text{alignment}}}{{\text{Investment}}_{\text{capabilities}}}\text{must increase from}1:100\text{to at least}1:1$$

Concrete actions required:

1. AI Companies: Allocate ≥50% of R&D budget to alignment (vs ~1% current)
2. Governments: Fund alignment research on a Manhattan Project scale ($10-50B/year)
3. Regulation: Partial moratorium on scaling until minimum alignment guarantees are met
4. Coordination: Binding international treaty (analogous to the Montreal Protocol)

Without correction of the current allocation, the framework predicts:

• AGI capabilities reached in 5-15 years (expert consensus)
• Robust alignment requires 15-30+ years under current allocation
• Result: Misaligned AGI with high probability

This warning does not invalidate the framework. It establishes a condition of viability: the framework works IF alignment is resolved. Resolving alignment requires an urgent inversion of investment priorities.

1. Introduction

1.1 Motivation and Scope

Humanity faces multiple converging crises characterized by large-scale irreversible losses:

• The extinction of species at unprecedented historical rates (~200 species per day)
• The collapse of critical ecological systems
• The erosion of cultural and linguistic diversity
• The potential for existential risks derived from emerging technologies

Simultaneously, collective decision-making systems demonstrate a systematic inability to adequately respond to these threats, often due to cognitive biases, short time horizons, and coordination failures.

This work proposes that the common root of these failures lies in the absence of an organizing principle that recognizes the fundamental asymmetry between preservation and destruction under conditions of irreducible uncertainty. While traditional ethical frameworks rely on concepts of intrinsic value, dignity, well-being, or utility—all subject to cultural variation and definitional uncertainty—the present framework is grounded in a more primitive observation: the epistemic impossibility of completely calculating the future value of any existing information.

1.2 Structure of the Argument

The argument proceeds as follows:

1. We establish a single non-demonstrable but pragmatically defensible axiom: information has value under radical uncertainty regarding its future utility.
2. We operationally define information as the capacity of a system to generate distinct trajectories in the space of future possibilities.
3. We derive logical principles from this axiom: the priority of preservation, a hierarchy of interventions, and trade-off criteria.
4. We apply the framework to concrete problems in social policy, environmental preservation, and technological design.
5. We argue for the necessity of AGI for optimal implementation, recognizing fundamental human cognitive limitations.
6. We recognize fundamental limitations of the framework, including the unresolved problem of AGI alignment.

1.3 Original Contributions

This work presents:

• A decision framework based on a single axiom, avoiding unresolved axiomatic pluralism
• The operationalization of "information" as a space of possibilities, allowing for practical comparisons
• A methodology agnostic to political ideologies, focused on empirical evidence
• A formal analysis of the necessity of AGI for global coordination under the framework
• An explicit recognition of limitations and unresolved problems

2. Theoretical Foundations

2.1 The Single Axiom and Its Justification

Axiom 1 (Value of Information under Radical Uncertainty):

Given that it is impossible to completely calculate the probability space of all future trajectories of any system, and given that the destruction of information is thermodynamically irreversible, the preservation of information is the dominant strategy under uncertainty.

2.1.1 Why This Axiom?

This axiom is not "provable" in the logical-mathematical sense—no normative value is (Hume 1739, Moore 1903). However, it has robust pragmatic justifications:

A. Irreducible Uncertainty:

Multiple factors make the complete calculation of future value impossible:

• Quantum physics: Fundamental indeterminacy in physical systems (Heisenberg 1927)
• Chaotic systems: Exponential sensitivity to initial conditions prevents long-term prediction (Lorenz 1963)
• Emergent complexity: Systemic properties not reducible to components (Anderson 1972)
• Unpredictable innovation: Future applications of knowledge cannot be anticipated (Popper 1957)

B. Thermodynamic Irreversibility:

The second law of thermodynamics establishes that processes of informational destruction increase entropy irreversibly (Landauer 1961, Bennett 1982). Destroyed information is not recoverable without prohibitive or impossible energy costs.

C. Historical Evidence:

Knowledge considered "useless" has repeatedly become critical:

• Number theory (Gauss, 19th century) → modern cryptography
• Quantum mechanics (1920s) → computers and the internet
• Genetics of Drosophila (1910s) → human molecular medicine
• General relativity (1915) → GPS and space navigation

D. Epistemic Asymmetry:

Preserving information allows the future option to use it or discard it. Destroying information eliminates this option permanently. Under uncertainty regarding the future value, the first strategy dominates the second (a principle similar to Arrow-Pratt risk aversion, 1965).

2.1.2 What This Axiom Is NOT

It is crucial to distinguish this axiom from superficially similar positions:

• It is not vitalism: It does not assert that "life is sacred" by virtue of an intrinsic property
• It is not aesthetic preservationism: It does not value diversity for "beauty" or subjective preference
• It is not deontology: It does not establish a transcendent "moral duty"
• It is not classical utilitarian consequentialism: It does not maximize "well-being"—it maximizes the space of possibilities

The axiom is epistemic, not ontological or moral in the traditional sense. It merely asserts: given that we do not know the future value, preserving is more robust than destroying.

[This is Part 1 of a multi-part series. Subsequent posts will cover: operationalization, applications, AGI necessity, alignment problems, and radical implications.]

Full technical paper with mathematical formalization and appendices available at: https://zenodo.org/records/18269112

Discussion Questions:

1. Does the single-axiom approach successfully avoid the problems of axiom pluralism, or does it simply hide value judgments?
2. Is "information as space of possibilities" operationalizable enough for practical policy decisions?
3. What are the strongest counterarguments to prioritizing information over well-being or dignity?
4. Does the framework's dependence on unresolved AGI alignment fatally undermine its practical viability?