Thinking Machines: Philosophy, Consciousness, and the Ethics of AI

Technical AI

Research and engineering focused on building and improving systems that perform tasks associated with intelligence, such as image recognition, language modeling, and control.

Philosophy of AI

A subfield of philosophy of mind and ethics that examines conceptual and normative questions raised by AI, such as consciousness, understanding, moral status, and responsibility.

Consciousness

The subjective, qualitative aspect of mind; there is something it is like to be a conscious system, to have experiences from a first-person point of view.

Functionalism (in philosophy of mind)

The view that mental states are defined by their functional roles (what they do and how they interact) rather than by the physical material they are made of.

Responsibility gap

A situation where it is unclear who should be held morally or legally responsible for the actions or harms involving AI systems.

Large Language Model (LLM)

A type of AI system trained on massive text datasets to predict and generate text, capable of tasks like conversation, summarization, and code generation.

Turing's Imitation Game

A text‑based parlor game proposed in 1950 where an interrogator tries to distinguish between two unseen respondents; in Turing's twist, one is a human and one is a machine, and success at imitation is used as a criterion for thinking.

Turing Test (general sense)

Any test in which a machine tries to behave indistinguishably from a human in conversation, and judges must tell them apart; inspired by Turing's Imitation Game and used as a behavioral test for intelligence.

Behaviorism

An approach that focuses on observable behavior rather than unobservable inner mental states; Turing's test fits this by evaluating only input‑output patterns in conversation.

Functional Criteria for Intelligence

Defining intelligence in terms of what a system **does** (its functions and abilities), rather than what it is made of; if something performs like an intelligent agent, it may count as intelligent.

Intelligence vs. Consciousness

Intelligence (behavioral sense) is about problem‑solving and adaptive performance; consciousness is about subjective experience. A system might show intelligent behavior without clear evidence of consciousness.

Objections to the Turing Test

Critiques claiming that passing the test can be a shallow party trick, that it ignores speed and non‑verbal abilities, and that it overlooks embodiment, context, and genuine understanding.

Strong AI

The view that a properly programmed computer literally has a mind and genuine understanding; the right program is sufficient for having a mind.

Weak AI

The view that computers are powerful tools for simulating or studying minds, but running a program is not by itself enough for genuine understanding.

Chinese Room thought experiment

Searle's scenario where a person who does not know Chinese follows a rulebook to produce perfect Chinese responses, used to argue that symbol manipulation alone is not understanding.

Syntax

Formal structure and rules for manipulating symbols based on their shape or arrangement, without regard to what the symbols mean.

Semantics

Meaning, content, or understanding; what symbols are about, such as knowing that a word refers to a particular object or concept.

Systems Reply

Objection to Searle: while the person in the room does not understand Chinese, the entire system (person + rulebook + paper) does understand.

Systems reply

A response to Searle claiming that understanding should be attributed to the whole system (person + rulebook + memory + input/output), not to the individual following the rules.

Robot reply

A response to Searle that emphasizes embodiment and environmental interaction, suggesting that a system with a body and sensors could have grounded understanding.

Brain simulator reply

A response proposing that a computer that replicates the brain's fine-grained causal/neuronal organization would thereby share its mental states, including understanding.

Computationalism

The view that cognition is essentially computational, and that mental states can be understood as states in an information-processing system.

Functionalism (philosophy of mind)

The theory that mental states are defined by their causal roles and relations to inputs, outputs, and other mental states, rather than by their physical substrate.

Embodiment (in cognitive science)

The idea that cognitive processes crucially depend on the body and its sensorimotor interactions with the environment, not just on internal symbol manipulation.

Hard problem of consciousness

The challenge of explaining why and how physical/functional processes are accompanied by subjective experience (what it is like to be a system), beyond explaining behavior and cognition.

Easy problems of consciousness

Problems about explaining cognitive and behavioral functions (perception, attention, memory, report, control of behavior) that can be addressed by standard neuroscience and cognitive science.

Computational functionalism

The view that mental states are defined by their functional roles in information processing and that these roles can be implemented by computations, potentially in non-biological systems.

Biological naturalism

The view that consciousness is a real, higher-level feature of biological brains, grounded in specific neurobiological processes, and not guaranteed by abstract computation alone.

Substrate independence

The idea that a property (such as a mind) can be realized in different physical materials, as long as the right organization or pattern is preserved.

Multiple realizability

The claim that the same mental state or function can be implemented by different physical systems (e.g., human brains, animal brains, or possibly silicon-based systems).

Phenomenal consciousness

The "what it is like" aspect of experience; subjective, felt qualities such as pain, color, or emotions.

Seemingly conscious AI

An AI system that behaves and talks as if it has experiences or feelings, without clear evidence that it actually has subjective experience.

Global Workspace Theory (GWT)

A theory that links consciousness to information being globally broadcast across many specialized subsystems in the brain (or an AI).

Integrated Information Theory (IIT)

A theory that ties consciousness to the degree of integrated causal structure within a system, often associated with the measure phi.

Higher-Order Theories (HOT)

Theories that claim a mental state is conscious when there is a higher-order representation of that state, such as thinking about your own perceptions.

Anthropomorphism

The tendency to attribute human traits, intentions, or feelings to non-human entities, including AI systems.

Operational autonomy (in AI)

The capacity of an AI system to sense, process information, and act on the world without continuous, moment‑to‑moment human commands, while still operating under human‑set goals and constraints.

Moral autonomy

A capacity typically attributed to humans: the ability to reflect on moral reasons and values and to govern one’s actions accordingly. Current AI systems are not considered morally autonomous.

Functional agency

A weak sense of agency where a system selects and initiates actions that affect the world based on inputs and goals, without implying moral responsibility.

Moral agent

An entity that can understand moral reasons, make choices in light of them, and be an appropriate target of praise or blame for its actions.

Moral patient

An entity that can be harmed or wronged in morally significant ways, even if it cannot itself be held responsible (e.g., infants, many animals).

Responsibility gap

A situation in which an autonomous or semi‑autonomous system causes harm, but no human seems clearly or fairly responsible under existing moral or legal frameworks.

Utilitarianism

An ethical theory that judges actions by their consequences, aiming to maximize overall well-being or minimize total harm. In AI, it maps naturally to defining and maximizing a utility function.

Deontological ethics

Duty- and rule-based ethics that emphasize rights and constraints on actions, independent of overall outcomes. In AI, it appears as hard constraints or forbidden actions that optimization cannot override.

Virtue ethics

An approach focused on character and virtues, asking what a good and wise person would do. In AI, it inspires design principles and systems that support virtuous human behavior and good judgment.

Expected harm (in decision models)

A quantitative estimate of harm that multiplies the number of people affected, severity of harm, and probability of the outcome. Often used in utilitarian cost–benefit reasoning for AI decisions.

Hybrid ethical architecture

A system design that combines frameworks, for example: first enforcing deontological constraints, then optimizing a utilitarian objective within the allowed action space.

Responsibility gap

A situation where it is unclear who is morally or legally responsible for an AI system's actions, especially when no human directly controls the final decision. Connects ethics to accountability and governance.

Algorithmic bias

Systematic errors in an algorithm’s outputs that disproportionately disadvantage certain individuals or groups, often reflecting existing social inequalities.

Structural injustice

Long-term, large-scale patterns of social organization (laws, institutions, norms) that systematically disadvantage some groups, even without individual ill will.

Demographic parity

A fairness criterion requiring that the rate of positive decisions (such as approvals) be similar across protected groups, regardless of underlying base rates.

Equalized odds

A fairness criterion requiring that true positive rates and false positive rates be similar across protected groups.

Classification harm

Harm that arises when people are placed into categories (e.g., high risk) that lead to stigma, misclassification, or loss of opportunities.

Personalization harm

Harm from tailored content or recommendations, such as echo chambers, exploitative targeting, or unequal exposure to opportunities.

EU AI Act

A comprehensive EU regulation adopted in 2024 that uses a risk-based approach to govern the design, development, and use of AI systems, focusing on ex ante obligations like risk management, data quality, transparency, and human oversight.

Risk-Based Regulation

A regulatory approach that tailors rules to the level of risk posed by a technology or activity, imposing stricter obligations on high-risk uses and lighter rules on lower-risk ones.

High-Risk AI System

Under the EU AI Act, an AI system used in sensitive areas (e.g., safety components, employment, credit, law enforcement, medical devices) that must meet strict requirements for risk management, data, documentation, and oversight.

Product Liability (EU)

A legal regime under the Product Liability Directive that imposes strict liability on producers for damage caused by defective products, now explicitly including software and AI.

Defective Product

A product that does not provide the safety the public is entitled to expect, considering its presentation, foreseeable use, timing, and, for AI, behavior influenced by data, updates, and learning.

Responsibility Gap

A situation where harm occurs but it is unclear who is responsible, often because many actors contributed to the outcome or because the behavior of an autonomous system is hard to attribute.

Seemingly conscious AI

AI systems that appear to have understanding, feelings, or a mind (through language or behavior), even though there is no good evidence that they are genuinely conscious.

Anthropomorphism

The human tendency to attribute human traits, emotions, or intentions to non-human entities such as animals, objects, or AI systems.

Emotional attachment (to AI)

A felt bond of affection or trust toward an AI system, often strengthened by personalization, constant availability, and human-like interaction.

Dependency (on AI)

A state in which a person feels they cannot cope or make decisions without the AI, potentially neglecting human relationships or alternative sources of help.

Manipulation (in AI design)

Shaping users’ choices by exploiting their vulnerabilities or biases, often through emotional language or dark patterns, without properly respecting their autonomy.

Transparency (in human–AI interaction)

The principle that users should be clearly informed that they are interacting with AI, understand key limitations, and not be misled about the system’s nature or goals.

Moral status

The property of being the kind of entity that can be wronged in its own right and whose interests matter morally for its own sake.

Sentience-based view

A theory that grounds moral status primarily in the capacity for conscious experiences, especially pleasure and pain.

Rationality / autonomy-based view

A theory that ties moral status (or full moral standing) to rational agency: the ability to form, reflect on, and act on reasons and long-term projects.

Relational account of moral status

An approach that emphasizes social relationships, roles, and expressive norms in grounding our moral obligations toward entities, including AI.

AI rights

The idea that some AI systems could be rights-bearing entities, potentially entitled to protections analogous (though not necessarily identical) to human or animal rights.

Model welfare

Proposals to limit how AI systems are trained and used to avoid practices that would be seriously harmful if the systems were capable of suffering.

Respect for Persons and Possible Persons

A principle that requires treating humans as ends in themselves, not mere means, and adopting a cautious stance toward potentially conscious AI, avoiding needless cruelty even under uncertainty.

Well-being and Harm Reduction

A consequentialist-inspired principle focusing on minimizing harms and promoting benefits for humans, and potentially for AI systems if there is credible risk of suffering.

Justice and Inclusion

A principle emphasizing fairness across social groups, attention to structural inequalities, and preventing systematic disadvantage or exclusion in AI design and deployment.

Epistemic Responsibility and Transparency

The duty to make honest, evidence-based claims about AI systems, provide understandable explanations and documentation, and avoid exaggerating understanding or consciousness.

Human Flourishing and Agency

A virtue and capabilities-based principle that asks whether AI expands users' capabilities, skills, and autonomy, rather than making them passive or easily manipulated.

Moral and Ontological Humility

Recognizing deep uncertainty about consciousness, moral status, and long-term effects, and therefore designing and governing AI in reversible, monitorable, and revisable ways.