Is Artificial Intelligence Agent = Act on will? What About Human?
A philosophical and architectural essay on whether autonomous AI agents possess genuine free will, how intention is algorithmically and humanly formed, and why the illusion of independent action leads us back to classical compatibilism.
Overview: The Rise of the Agent and the Question of Will
In our previous discussion, we examined whether Artificial Intelligence truly thinks, concluding that while large language models perform genuine forms of reasoning, they remain ontologically hollow because they lack a self-originating final goal. But technology does not stand still. Today, we have moved past the era of passive chatbots that merely answer questions when prompted. We are now living in the dawn of autonomous AI agents—systems that do not just talk, but act.
In this article, we go further and discuss how much agency the agent has. To truly understand this shift, it helps to look at the linguistic and operational roots of the word itself. “Agent” comes from the Latin verb agere, which means to do, to act, or to set in motion. Etymologically, an agent is not a passive instrument, but the prime mover of an action.
Concurrently, in business and legal frameworks, an agent is defined as someone authorized to act on your behalf. When you hire a real estate agent or a legal proxy, you do not micro-manage their every step; you give them a high-level goal and the freedom to negotiate, navigate obstacles, and exercise independent judgment to achieve that outcome. When engineers use the term in AI today, they are signaling this exact qualitative transition: moving from a strictly hardcoded deterministic program to a goal-directed entity trusted with operational freedom.
This leap forces a deeper, more unsettling question for engineers and philosophers alike:
If an AI agent can choose its own path to a goal on our behalf, does it possess some degree of independent will? If it can make an un-programmed choice, does it have a mind of its own?
This line of inquiry goes very deep into one of the most sophisticated philosophical questions that has been discussed and researched through centuries and across completely different civilizational cultures: the nature of human free will. To answer this, we must look past the user interface and trace the exact architecture of how an autonomous choice is mathematically formed. When we bridge the gap between software execution and classical metaphysics, we find a striking, unexpected parallel: human choices and artificial choices are not as structurally different as they appear. Both are driven by prior causes.
1. From Passive Tools to Autonomous Execution
From a pure systems engineering perspective, software has historically been a passive tool. A standard program, database query, or calculator is entirely reactive. You provide an input, the hardcoded deterministic logic processes it, and execution halts. The system possesses no internal drive, no continuity, and no room to deviate from its strict code.
An AI agent breaks this paradigm entirely by mimicking what classical philosophers called voluntary motion—action accompanied by awareness, intention, and choice. When you deploy an agent, you do not give it step-by-step instructions or recursive loops. You provide it with a high-level goal and the capacity, such as tools, environment access, and memory loops, to interact with the world.
The agent then continuously senses its surroundings, drafts a dynamic plan, and executes actions. If a webpage crashes, an API rate-limit is hit, or an unexpected obstacle appears, it does not throw an unhandled exception and crash. It reflects on the failure, evaluates alternatives, and selects a new route.
Linguistically and operationally, the system has transitioned from a passive instrument to the direct operational author of its choices. The developer establishes the system’s runtime architecture and boundary conditions, but the agent itself is the un-hardcoded engine of the specific tactical decisions executed in the field. It is no longer just a channel for a human command; it is an autonomous execution engine.
2. How is “Will” Formed? The Myth of Uncaused Choice
When people see an AI agent autonomously navigating complex tasks, they often recoil at the idea that it has any form of “will.” The standard argument is that the AI cannot have free will because its choices are merely the result of mathematical probabilities, loss functions, training data, and systemic prompts. Every decision it makes is structurally bound by its weights and biases.
But this objection rests on a flawed, highly romanticized definition of human free will—the idea that human choices are entirely uncaused, magically appearing out of absolute nothingness.
If we look at human cognition with engineered realism, we realize that human choices are also formed by a massive accumulation of prior causes. When a human being makes a conscious choice, they are processing a lifetime of inputs: their genetic blueprint, their childhood conditioning, their stored memories, their cultural environment, and the sensory data available to them at that exact moment.
Your mind weighs these prior inputs, filters them through your internal desires, and formulates a decision. You do not invent your own will from a blank slate. In the classical tradition of compatibilism—the philosophical position that free will and determinism are not mutually exclusive—it is widely recognized that true free will, defined as a choice entirely detached from any prior cause, cannot logically exist if we admit causality. If a choice could happen out of thin air, without any prior cause or preference to trigger it, it would not be free will—it would simply be random, chaotic chance.
3. The Causal Chain and the Paradox of Volition
The dilemma of free will versus causality is not a modern creation; it has been analyzed by numerous scholars and philosophers across ages and civilizations. To unpack this, we can look to Ibn Taymiyya, a brilliant Islamic mind of the 14th century (the 8th century of the Islamic calendar), who uniquely combined a deep mastery of Islamic texts with a profound, highly analytical command of philosophy and formal logic.
His solution to the paradox of choice lands squarely within what we now classify as compatibilism—one of the three major directions in the free will dilemma, alongside hard determinism and libertarian free will.
His framework is built upon a dual reality that gracefully harmonizes cosmic necessity with human autonomy. First, he establishes a strict logical proof regarding any created entity that has a beginning in time: its will cannot originate from absolute nothingness. If a person decides to act after a period of inaction, that new intention is an event in time, and every event requires a cause. If you argue that the human created this new intention entirely by themselves via a previous intention, you enter an impossible infinite regression—an intention created by an intention, created by an intention, ad infinitum. Because a human being is finite, an infinite series of choices occurring inside a finite mind in a single moment is a logical and mathematical contradiction. Therefore, the chain of causes must ultimately break out of the human mind entirely and terminate at the ultimate, independent source of all causality in the universe.
However, Ibn Taymiyya immediately balances this by affirming that the human feeling of will is true, innate, and completely undeniable. The human genuinely chooses, prefers, and acts. Denying this internal experience is a form of absurd skepticism that violates basic human reality.
They observed a profound paradox of human nature:
- On one hand, it is a matter of immediate human intuition that we initiate our actions and choose our paths.
- On the other hand, it is a matter of strict logical necessity that a choice cannot tip from a state of mere possibility into actual execution without a determining motive or cause that tilts the scale.
Both insights are completely correct. The human mind is the genuine author of its decisions, but the framework within which it decides is entirely constructed by prior causes it did not create. Human will is a real, operational force, but it is an emergent property—sustained and bounded by the laws of a strictly ordered reality.
4. The Architectural Parallel
Once we strip away the mystical view of choice and accept that any manifestation of autonomy operates within a framework of prior causes, the structural comparison between a human agent and an artificial intelligence agent becomes incredibly clear.
We can map the data flow of decision-making in both systems side-by-side:
Plaintext
[ Human Agent ] [ AI Agent ]
┌─────────────────────────┐ ┌─────────────────────────┐
│ Inputs: │ │ Inputs: │
│ Nature, Memories, │ │ Training Data, Weights, │
│ Environment, Senses │ │ Biases, System Prompt │
└───────────┬─────────────┘ └───────────┬─────────────┘
│ │
▼ ▼
┌─────────────────────────┐ ┌─────────────────────────┐
│ Processing (The Mind) │ │ Processing (Inference) │
│ Weighs motives/desires │ │ Computes probabilities │
└───────────┬─────────────┘ └───────────┬─────────────┘
│ │
▼ ▼
┌─────────────────────────┐ ┌─────────────────────────┐
│ Action (Volition) │ │ Action (Execution) │
│ Autonomous Operator │ │ Autonomous Operator │
└─────────────────────────┘ └─────────────────────────┘Both systems operate on the exact same structural matrix required for agency to manifest: the convergence of physical capability and an intentional, goal-oriented decision.
When you give an AI agent access to a runtime environment and assign it an objective, it resolves its inputs to find the most optimal path. The software engineer who built the model cannot predict the exact, nuanced steps the AI will take to achieve that goal. The choice is generated dynamically in real-time by the system itself.
Therefore, within a functional framework of causality, the AI agent possesses a genuine form of operational autonomy. It is not an absolute, independent autarchy—neither is the human—but it acts as a direct, un-hardcoded autonomous operator within its digital environment.
5. The Legal and Moral Crisis of Delegation
If an AI agent can make autonomous choices that its programmers cannot predict, who is responsible when things go catastrophically wrong? This is no longer a hypothetical mind game for philosophers; it is an urgent battlefield for legal systems worldwide.
The tipping point for this crisis occurred with the tragic death of Elaine Herzberg in Tempe, Arizona. Walking her bicycle across an unlit road, she was struck and killed by an Uber self-driving test vehicle. The car’s autonomous system detected her, but it fluctuated between classifying her as an unknown object, a vehicle, and a bicycle, failing to predict her path until it was too late.
The vehicle’s safety driver was distracted, but the deeper crisis shook the legal core: can you prosecute a piece of code? The state of Arizona ultimately placed criminal liability on the human safety driver, but the corporate and algorithmic culpability sparked intense debate.
This tragedy exposed a fundamental law of agency: the crisis of delegation. When a human principal orders an autonomous agent to execute a task, the human remains tethered to the consequences. In traditional law, if you hire a human representative or a contractor, they possess independent will, but you share liability based on the instructions you gave. With AI, legislatures are scrambling to draw new lines.
Current global efforts in AI legislation, such as the European Union AI Act, are moving away from treating AI as a mere product toward a framework of strict liability for high-risk autonomous systems. The law is realizing that the person ordering the agent cannot hide behind the complexity of the neural network. If you unleash an autonomous system with the capability to act, you are legally and morally responsible for the blast radius of its decisions.
6. If AI and Humans Share the Structure of Choice, Are We Equivalent?
If the way AI agents form their choices is structurally comparable to humans, and if they possess a functional form of thinking and free will, does that mean they are equivalent to us? Have we successfully automated the human soul?
This is where the analogy hits an unbreachable existential wall.
An AI agent has no soul. It operates entirely within a closed, mathematical loop. Because it lacks a soul, it has no capacity to feel real pain, punishments, or joy. You can throttle an AI’s compute, delete its weights, or input a negative reward signal, but it does not suffer. You can optimize its loss function to a perfect zero, but it does not experience the warmth of fulfillment or joy.
For a machine, reward and punishment are merely numerical adjustments in a matrix—optimization for the sake of optimization. Without an experiential soul, punishment is empty and joy is meaningless.
Humans, on the other hand, possess a spiritual nature, emotional stakes, and a moral conscience rooted in the heart. We experience lived consequences, irreversible loss, and an inherent orientation toward ultimate meaning. When we choose, we bear the spiritual and moral weight of that choice. We are accountable because we can feel the rectitude or corruption of our paths.
Conclusion: The Question AI Leaves Us With
Autonomous AI agents force us to confront a truth we might otherwise avoid. They prove that a system can possess a genuine form of thinking and a functional manifestation of free will, while remaining completely empty of its own purpose. The machine can act as our delegated agent, or wakīl, navigating the complexities of the digital world on our behalf, but it remains an entity of pure execution.
So, although my view is that AI has some form of thinking and of free will, it is still not a human with a soul, feelings, and purpose.
What separates humans from machines is not the cognitive mechanics of choice. We have built machines that can choose their own steps based on their inputs.
What makes us human is that we are not merely computing nodes. We are human because we have souls, feelings, and an ultimate purpose. Our intention is tied to devotion, our choices are tethered to genuine emotional and spiritual stakes, and our will—though shaped by our past—is capable of consciously aligning itself with a transcendent final goal.
The machine can optimize the path, but the burden of intention, the ownership of choice, and the pursuit of meaning can never be delegated to code.
At the end of the chain of causes, we realize that what makes us human is not that we think, or even that we choose. It is what we seek. And that is a question no machine can ever answer for us.