Hey Siri, You Will Never Know: Human Agency and the Philosophical Limits of Artificial Intelligence

Artificial intelligence now analyses past data, adapts to new environments, and refines its behaviour through feedback. In many practical settings, it appears convincingly goal directed. This raises a fundamental philosophical question. Can artificial intelligence attain the expressivity, perception, attention, creativity, and agency that characterise the human mind? Or does it reproduce only the observable form of these qualities? Machines can generate language, support medical decisions, navigate environments, and perform increasingly complex tasks. Yet impressive behaviour alone does not establish the presence of a human mind or genuine human agency.

This paper examines that distinction by placing artificial intelligence and the human mind within a common computational framework based on the Universal Turing Machine. The aim is to create a shared conceptual ground for computational, mathematical, cognitive, and philosophical inquiry. Within this framework, an artificial intelligence system is physically implemented and therefore resource bounded. Its memory, processing capacity, and physical operation remain finite. This interpretation is consistent with the OECD conception of an artificial intelligence system and with the reinforcement learning model represented by the Markov Decision Process.

The human mind is modelled differently. We represent it as a finite controller coupled to an unbounded tape. This is a deliberate modelling assumption that allows us to investigate whether a structural distinction between bounded and unbounded systems can illuminate differences between artificial and human intelligence. The unbounded structure is associated with expressivity and is considered in relation to Gödel's incompleteness results and Turing's work on undecidability. Within this model, human cognition is not understood only through observable behaviour or information processing. Subjective experience, perception, feeling, volition, attention, meaning, and context also contribute to the way human agency is formed.

We argue that the distinction between bounded and unbounded systems provides a useful basis for examining differences in expressivity, creativity, attention, and accountability. Artificial intelligence may produce behaviour that resembles purposeful action, but behavioural similarity does not necessarily establish equivalent agency. This distinction becomes especially important when responsibility is considered. An artificial system can influence decisions and actions, but ultimate accountability cannot simply be transferred to the computational system that produces them.

The paper also contributes to the wider philosophical question of how different forms of inquiry can be brought into a common framework without reducing one form of knowledge to another. Computational theory provides a shared method of comparison, while philosophy allows questions of experience, meaning, agency, and responsibility to remain visible. We conclude that human centred design and meaningful human oversight are necessary conditions for the responsible development of artificial intelligence. As artificial systems become more capable, responsibility must remain grounded in human agency.