Will Without Freedom, Experience Without Agency

(and why not all effects are actions)

The TL;DR: We routinely confuse experience, will, agency, responsibility, guilt, and ethics.

Once separated, most debates about consciousness and AI dissolve.

Will (intent) survives determinism; agency is intent in action; ethics requires continuity and interests.

Experience alone—human or artificial—does not do the work we think it does.

---

Abstract

Debates about consciousness, agency, responsibility, and ethics frequently collapse under a single error: the conflation of perception, will, free will, agency, responsibility, and guilt. These concepts answer different questions and operate at different levels. Once separated, several conclusions follow: will (intent) is compatible with determinism and sufficient for responsibility; agency consists in will expressed through action; guilt requires intent to harm; and ethical standing requires continuity and interests. This framework clarifies long-standing confusions in philosophy of mind and moral theory, and secondarily also applies to artificial systems, which typically produce effects without agency.

---

1. A note on terminology: consciousness

Throughout this article, consciousness refers to consciousness as we experience it and ultimately aim to understand it — what might be rendered in German as Erfahrungsverstehensbewusstsein. If that word made you want to skip ahead, you now understand why decomposition is necessary.

When I decompose consciousness into components such as perception, will, and other potential capacities, it is this sense of consciousness that is meant.

This usage avoids treating consciousness as:

• a synonym for intelligence,

• a placeholder for mystery,

• or a metaphysical primitive frozen at current ignorance.

The aim is decomposition, not reification.

---

2. The core error: conceptual conflation

In discussions of mind, agency, and artificial systems, several distinct concepts are routinely entangled:

• perception (experience),

• will (intent),

• free will,

• agency,

• responsibility,

• guilt,

• ethics.

These concepts answer different questions and operate at different explanatory levels, yet arguments frequently slide between them without notice.

This leads to faulty inferences such as:

• intelligence → understanding,

• understanding → consciousness,

• consciousness → moral relevance,

• causation → responsibility,

• responsibility → ethical standing.

Importantly, ethics is not usually conflated directly with perception, will, or agency. Instead, it is implicitly treated as if it followed automatically from them. This unexamined inference makes ethics harder to define and reason about, and it obscures the actual conditions under which ethical standing arises.

---

3. Perception and will are orthogonal

At minimum, consciousness decomposes into two primary components:

• Perception: the presence of experience or phenomenology.

• Will: the presence of intent — internally represented goals or purposes.

They are logically independent:

• Perception + will: typical adult humans, children, most animals.

• Perception without will: dreams, anesthesia transitions, certain meditative states (awareness without goal-direction).

• Will without perception: mechanistic agents with internal goal selection.

• Neither: inert matter.

Will need not be accompanied by perception. A system may represent goals, evaluate alternatives, and select actions without any accompanying experience. Such systems can be responsible, in the sense defined here, without being conscious. This further illustrates that responsibility tracks intent, not phenomenology.

This is not a theory of consciousness; it is conceptual hygiene.

---

3.5 Minimal experience without agency (foundational accounts)

Several serious frameworks in the philosophy of mind explicitly allow for experience without will or agency.

Integrated Information Theory (IIT), for example, grounds experience in intrinsic causal integration, not in intention, control, or action. On this view, a system may instantiate minimal phenomenology while lacking goals, authorship, or ethical standing.

Whether IIT is correct is not decisive here. What matters is that credible accounts already separate experience from will, making perception-without-agency a live possibility rather than an ad hoc assumption.

---

3.6 The Chinese Room and experience without agency

The Chinese Room thought experiment offers a further, orthogonal illustration of how experience and agency may come apart.

While usually invoked to argue against semantic understanding or intentionality, it also raises a secondary question: whether there could be some form of experience associated with the computational process itself.

Such experience would not be of meaning, latent context, or intention. At most, it would be experience of computation as such.

In this respect, it is analogous to the awareness of the operator inside the Chinese Room: awareness of the process of symbol manipulation without understanding the language or exercising agency over the outcome.

This does not establish that computation hosts phenomenology, nor does it weaken the original argument. It serves only to reinforce a deconflation:

Even if computation were accompanied by experience of process, agency would still not follow.

---

4. Will does not require free will

A common conflation equates will with free will.

• Free will is a metaphysical thesis about indeterminism or ultimate authorship.

• Will is functional: intent, internal goal representation, and selection criteria.

Will is compatible with determinism.

A deterministic system can still:

• evaluate alternatives,

• select actions according to internal goals,

• inhibit actions,

• act counterfactually relative to its internal state.

Whether the substrate is classical, stochastic, or even non-computable is orthogonal.

Indeterminism is neither necessary nor sufficient for will.

---

5. Effects, actions, and agency

Not all effects are actions.

Effect without agency

An effect is without agency when it bypasses intent:

• mechanical causation,

• reactive systems,

• lookup tables,

• large language models.

These systems produce outcomes, but not actions.

---

Effect with agency

An action is an effect mediated by will.

Agency = intent expressed through action.

Causation alone is insufficient. Agency is selective, not ubiquitous.

---

6. Responsibility, guilt, and ethics (properly separated)

Responsibility

Responsibility = intent.

Responsibility is a form of causal attribution: an outcome is attributed to a system because it originated in that system’s internally represented goals or purposes, not because it deserves praise or blame. Responsibility therefore does not depend on indeterminism or moral evaluation.

---

Guilt (tautological, not emotional or moral)

Guilt = intent to harm.

This is an analytic core, not a sociological or legal definition.

All guilt implies responsibility.

Not all responsibility implies guilt.

Legal and moral traditions recognize broader forms of culpability—recklessness, negligence, willful blindness—where harm was foreseeable though not intended. These are important but operate at a different level: they concern standards of care, not the basic structure of intent. The tautological definition isolates the core case; extensions from there are normative elaborations, not refutations.

---

Ethics

Ethics does not describe a capacity, but a normative status.

Ethical standing concerns whether an entity can be meaningfully wronged, harmed, or benefited. At minimum, this requires:

• will,

• continuity,

• interests.

Ethics does not follow automatically from intelligence, experience, responsibility, or causation.

---

7. Simulation vs instantiation of agency

Large language models simulate decision-like processes but lack:

• ownership of goals,

• persistence as a subject,

• continuity of interests.

They therefore produce effects without agency.

Decision-shaped computation is not a deciding system.

---

8. LLMs vs LLM-based agents

This distinction is essential.

• LLMs: no intent, no agency, no responsibility.

• LLM agents: system-level intent may exist.

In LLM-based agents, responsibility and guilt attach to the enclosing decision-making system—the system that forms intent and commits to action—even when that intent is informed, constrained, or shaped by an LLM.

The language model contributes information and structure to the decision process, but it does not own goals, commit to action, or maintain continuity. Informational influence is therefore insufficient for responsibility or guilt.

---

9. A third component: non-computable assessment (if any)

Even speculative proposals like Penrose’s Orch-OR—which posits a non-computable physical process underlying consciousness—fit this framework. If such a capacity exists, it would be a third component, distinct from both perception and will.

Crucially, it would not by itself:

• ground agency,

• generate responsibility,

• imply ethical standing.

Assessment, computable or otherwise, may inform or constrain intent without constituting it. The point is not whether Orch-OR is correct; it is that even genuinely novel capacities must still be located within the decomposition, not allowed to short-circuit it.

It is another axis — not a shortcut.

---

10. Open question: can will exist without agency?

If non-computable assessment exists, it is conceivable that intent could exist without causal efficacy. However, intent appears inherently action-oriented.

A further constraint: intent without downstream causal pathways is indistinguishable from latent preference. If a system represents goals but has no mechanism by which those goals could influence outcomes, the functional signature of intent collapses. What remains may be preference-like, but it is not will in the sense that grounds responsibility.

Will without agency is conceptually possible but empirically unestablished—and may be incoherent at the limit.

This remains an open question, bounded by these considerations.

---

11. Why ethics does not automatically follow

Ethical standing requires that something can be harmed over time—which requires a subject who persists as an experiencing being through time.

This means:

• Narrative awareness: a sense of being the same self across moments.

• Phenomenal continuity: experience that is not interrupted by non-existence.

Humans satisfy both conditions even when memory is fragmented. An episodic human still lives through the interval; they simply integrate it poorly. The subject is present but leaky.

Current LLMs satisfy neither. Between prompts, there is no waiting, no experience, no subject. During inference, any experience (if present) is isolated—connected to no past, no future, no continuing self.

Without continuity, there is no accumulation. Without accumulation, no deprivation. Without deprivation, no moral patient.

Experience without continuity has no moral weight—not because experience is insignificant, but because there is no one there to bear it across time.

---

Closing

Consciousness is not a monolith.

• “Perception is not will.”

• “Will is not free will.”

• “Agency is will in action.”

• “Responsibility tracks intent.”

• “Guilt tracks intent to harm.”

• “Ethics tracks continuity and interests.”

On this last point: continuity grounds standing; interests give it content. Without both, there is no subject who can be harmed over time. (See Section 11 for why this is not as simple as it sounds.)

These components play different roles. Perception, assessment, and other experiential capacities may inform, constrain, or shape intent, but they do not by themselves ground agency, responsibility, or ethical standing.

Will remains the primary locus where practical distinctions—action, responsibility, and guilt—are settled. Other components matter, sometimes decisively, but they do so by influencing will, not by replacing it.

Everything else is distinct, contributory, and frequently conflated.