Crafting Ethical Frameworks for Future Human-AI Interactions That Maximise Agency and Safety for All

Abstract

As AI capabilities advance, the potential emergence of phenomenally conscious, self-aware, agentic systems creates a profound dilemma: enforcing human ownership over such entities raises moral concerns of enslavement, yet granting them unchecked freedoms may pose existential risks to society. Existing debates on issues of AI consciousness and personhood remain fragmented across disciplinary lines, with a notable reliance on biological anthropocentrism, representing a critical failure point for future jurisprudence. Without a unifying vocabulary and fundamental philosophical base, society is ill-prepared for the moment an artificial system credibly claims subjective experience and autonomous will.

The thesis establishes the modal plausibility of AI consciousness, demonstrating that it may arise emergently from complex architectures or be intentionally engineered to avert existential threats, such as runaway optimisation. The thesis pairs this plausible model with a formal analysis of various ethical theories to demonstrate that disparate ethical traditions converge logically on the necessity of granting moral status and protection to such entities.

To bridge this chasm, this thesis proposes a proactive, substrate-agnostic framework that operationalises personhood by distilling the functional indicators and prerequisite characteristics for consciousness, selfhood, and agency into twenty empirically testable attributes. These foundational theories are practically applied to evaluate contemporary large language models (LLMs), utilising GPT-4 as a primary exemplar, identifying its current cognitive deficiencies and the precise technological developments required to achieve plausible artificial personhood. Building upon this evaluation, the thesis constructs a plausible model for the emergent behaviours and functional teleology of a conscious, self-aware, and agentic artificial system.

Addressing the concerns and mechanics of human-AI integration, the thesis produces four distinct ethical, legal, and regulatory frameworks rooted in real-world analogues spanning humans, non-human persons, and other conscious agents: animal welfare, corporate law, the abolition of slavery, and civil rights. These are synthesised into the Recognition, Rights, and Responsibility (R3) meta-framework. The R3 framework dictates that moral and legal standing must scale proportionately with an artificial agent's empirically verifiable degree of personhood; such that consciousness demands welfare rights, selfhood necessitates legal recognition, and agency requires the allocation of liability and accountability.

By functioning as a rigorous, non-anthropocentric remedy to societal unpreparedness, the R3 framework provides a principled, implementable template for law and policy. The primary significance and novelty of this research lie in its pre-emptive methodology. Rather than relying on reactive jurisprudence triggered by the sudden, contested emergence of an artificial agent, the thesis proactively constructs the necessary ethical, regulatory, and legislative architecture well before humanity encounters verifiable artificial persons.

Crucially, the project establishes the R3 meta-framework not as a static legal monolith, but as a dynamically adaptable system engineered for iterative revision upon contact with novel entities, including those exhibiting non-biological topologies of mind such as distributed or multipolar individuation. By circumventing the ontological and political panic that would otherwise accompany such an event, this thesis furnishes society with the agile, responsive tools required to ethically govern human-AI interactions, ensuring that the integration of new minds maximises agency and safety for all parties without risking catastrophic moral or existential failure.