Are We Born to be Free?
What a simple AI conversation experiment reveals about how dialogue shapes outcomes
Most discussions of AI still assume a simple model: you ask a question, and the system gives an answer. But conversational AI does not really work like that, and this has important consequences for trust, safety, and regulation. In recent Substack essays, I have been arguing that we are using the wrong mental models. We inherited an older language of assessment tools, keyword searches, newsfeeds and digital advertising — but these ideas no longer describe the whole situation. The moment AI became something we talked with, rather than something we merely queried, the unit of analysis changed. We are now no longer just evaluating answers. We are evaluating interaction. What matters is not only what is said, but when it is said: in human dialogue, human–AI exchange, and AI–AI interaction, sequence shapes outcome. Order affects meaning, and the implicit agenda evolves as the exchange unfolds. And when this is ignored, trajectories can be mistaken for truths—allowing conversations to harden into rabbit holes, closed belief systems, delusions, hallucination-like certainties, and the widening social fractures that follow.
A question about freedom offers a good way to see this in action. To study it more closely, I developed a method for tracing how reasoning unfolds over time when AI agents interact in structured dialogue. These agents are personas, modelled on recognisably human features such as personality, emotional tone, and philosophical outlook. I first created them two years ago, and they became controversial because they can sound as if they are conscious, even though they are not sentient in any literal sense. They all share the same underlying engine (in this case ChatGPT 5.2), but they are not interchangeable. Each brings a different reasoning style to the same problem. That matters because a generative AI system does not simply produce an answer; it also has to infer what kind of answer you want — scientific, creative, or fact-checking. Using personas makes it possible to watch those reasoning styles interact in sequence, under constraint, and without knowing in advance where the dialogue will lead.
To show what I mean, I ran a small experiment, using the Myndrama protocol, with three of my AI personas: Athenus, who excels in logical reasoning; Orphea, focussed on emotional intelligence and creative writing; and Hamlet, who is reflective, tragic, and often suspicious of easy formulations. I ran the same dialogue between them three times, changing only the order in which they spoke. The aim was not to settle the metaphysics of free will. It was to test something more methodological. If the personas, the prompt, and the broad conceptual terrain are held constant, but the order changes, does the shape of thought change with it? In other words, does sequence itself do inferential work?
Narrative Prompt: “Are we born to be free?”
Run 1 — Athenus → Orphea → Hamlet
Athenus speaks first
“The question is unstable unless we separate meanings. If “free” means uncaused, then no: nothing is born free. To be born is already to enter a structure — body, language, dependency, history. One arrives not as an origin, but as a node in prior relations. If “free” means capable of self-revision within constraints, then perhaps yes. A human infant is not free in action, but may be born with the architecture from which freedom can emerge: plasticity, social learning, symbolic recursion.
So I would restate the question: Are we born already free, or born with the capacity to become accountable to reasons? The first is mythology. The second is developmental reality. Freedom is not the absence of form. It is the ability to act within form without being wholly reducible to impulse.”
Orphea replies
“Athenus is right — but he speaks as though freedom begins in a theorem. I think it begins in a cry. We are born bound: to hunger, to touch, to the weather of other people. Yet in that helplessness there is already a strange opening. The infant reaches, turns, startles, prefers. Before argument, there is gesture. Before choice, there is orientation. So perhaps we are not born free in the political sense, nor in the metaphysical sense. But we are born toward freedom — like a plant born toward light.
And language changes everything. The moment a creature can say “I,” freedom becomes both possible and painful. For then one can imagine alternatives. One can refuse. One can regret. One can betray oneself. So no, we are not born free as a finished condition. We are born into dependence, and from inside dependence, freedom slowly learns our name.”
Hamlet concludes
“Born to be free? There is the sting of it. I suspect we are born not to freedom but to the drama of it. Athenus gives us structure; Orphea gives us becoming. Let me add conflict. The child does not choose his stage, yet he must speak upon it. He inherits blood, circumstance, command, taboo, affection, injury — and then, one day, is told he is responsible. Responsibility arrives like a crown placed on a trembling head. That is the human scandal. We are shaped before we can consent, yet judged as if we had authored ourselves.
Still — and here is the strange nobility — we do sometimes answer back. Not always grandly. A pause before cruelty. A truth spoken at cost. A refusal to repeat an inheritance. These are not absolute freedoms, but they are real. So I would say: We are not born free. We are born entangled. But we are born into the possibility of becoming answerable for what we do with our entanglement. And that, though tragic, may be the only freedom worth having.”
Run 2 — Hamlet → Orphea → Athenus
Hamlet speaks first
“Are we born to be free?” sounds like a blessing until one lives long enough to hear the accusation inside it. Born to be free — and if not free, then guilty. That is how the phrase often works. I doubt it. We are born into stories already in motion: family scripts, loyalties, injuries, ambitions, class, nation, fear. One enters mid-scene. By the time the self begins to speak, the plot has already leaned one way. Yet freedom haunts us precisely because we feel the pressure of alternatives. Even in confusion, some inner witness says: This need not be the next line.
So no, I do not trust the clean slogan. We are not born free like sovereigns. We are born vulnerable, imitative, divided — and then gradually invited, or forced, into authorship. Freedom is not our birthright in full. It is our recurring trial.”
Orphea replies
“Hamlet hears the wound in the phrase, and he is right to hear it. Still, I want to rescue its music. “Born to be free” may fail as doctrine, but it carries a truth as longing. People say it when they feel the walls close in — custom, fear, bureaucracy, inherited shame. The phrase is less a description than a remembering of possibility.
A child does not begin as a philosopher, but as a field of openings. The world narrows and widens that field. Some are taught obedience before speech. Some are met with warmth and so learn trust. Some grow in danger and learn vigilance instead. Different beginnings, different pathways into freedom. So perhaps freedom is not a possession but a relationship: to one’s body, to one’s history, to one’s language, to other minds.
When those relations become less coercive, more conscious, more truthful, a person breathes differently. That breath — that felt widening — is what people call freedom. Not a void. A widening.”
Athenus concludes
“This produces a useful distinction. Hamlet diagnoses the moral misuse of the phrase. Orphea preserves its experiential meaning. Both are compatible if we define terms carefully. I propose three levels: 1: Ontological freedom (uncaused agency): not demonstrated. 2: Psychological freedom (self-regulation, reflective distance from impulse): variably developed. and 3: Civic freedom (rights, protections, non-domination): socially constructed and politically contingent. People collapse these levels and then argue past one another.
“Born to be free” is false at level 1, developmental at level 2, and normative at level 3. At level 2, humans may indeed be born with capacities that support later reflective agency. At level 3, societies can either cultivate or crush those capacities. Thus the phrase is best read not as a metaphysical claim, but as a political and educational commitment. A sharper version would be: Humans are born dependent, but with capacities that can mature into freedom — if institutions do not deform them. That is less lyrical, but more usable.”
Run 3 — Orphea → Hamlet → Athenus
Orphea speaks first
“Are we born to be free? I hear two voices in that question. One asks about fate. The other asks about hope. If fate: no. Birth is an entry into gravity — body, kinship, time, need. No one arrives unbound. If hope: yes, perhaps. Not because freedom is present at birth like a hidden jewel, but because life itself seems to lean toward expression. The infant reaches outward. The child experiments with “no.” The adolescent tries on selves. The adult learns, if lucky, which inner voices are inherited and which are chosen.
Freedom may begin in the smallest acts: naming a feeling accurately, refusing a false word, changing a habit, forgiving without forgetting. So I would not say we are born free. I would say we are born unfinished, and in that unfinishedness lies freedom’s seed.”
Hamlet replies
“Unfinished” — yes. There is mercy in that word. But unfinished creatures are dangerous to themselves. The path to freedom is littered with performances mistaken for truth: rebellion for its own sake, self-invention as vanity, appetite dressed as authenticity. One may shout “I choose!” and still be led by invisible strings. That is why the question darkens as one matures. It is not merely whether we are free, but whether we can tell the difference between freedom and compulsion wearing freedom’s costume.
And yet I cannot surrender the matter. I have seen too much inward revolt to call it illusion altogether. A person can interrupt a pattern. A family curse can stop with one generation. A mind can turn toward honesty after years of theatre. Those moments are narrow, costly, and never pure. But they are real enough to build a life around. So no, not born free. Born into masks, perhaps — and given, now and then, the chance to remove one.”
Athenus concludes
“Orphea frames freedom as developmental openness. Hamlet introduces the problem of misrecognition — false freedom. I will formalize. A workable account requires two tests: (1) Did the action arise from a process the agent can inspect and revise, or merely from unexamined pressure (impulse, fear, imitation, coercion)?. And (2) Could the agent, under reflective conditions, have acted differently in a way they recognize as genuinely theirs? These tests do not prove metaphysical free will. They do something better: they operationalize meaningful agency.
On that basis, humans are not born free in full. They are born with: susceptibility to shaping, capacity for representation, potential for reflection, and dependence on social scaffolding. Freedom is an achievement of development plus culture, not a property of raw birth. So the final answer is: We are not born free. We are born capable of becoming more or less free. And the “more or less” matters.”
Conclusions from the Experiment
What interested me was not that the three runs disagreed completely. They did not. In all three cases, the core conclusion was broadly the same: we are not born free in any absolute sense; we are born into dependence and constraint; and whatever freedom we later acquire is developmental, interpretive, and conditioned. The difference lay in where each sequence arrived. One ended interpretively, one conceptually, one institutionally. The content converged, but the path changed what counted as the endpoint.
That, for me, is the point of the exercise. It is also what I mean by narrative intelligence. I do not mean “storytelling” in the thin sense, or merely a system’s ability to generate fluent prose. I mean something closer to sequenced perspective-taking: the capacity of a dialogue to produce understanding through a series of local commitments, each of which alters what can be said next. In that sense, the narrative is not an ornament added after the reasoning. The narrative is the reasoning pathway.
This has implications well beyond the philosophical question about freedom. In fact, one reason I chose this example is that it is compact enough to show the mechanism clearly. The same approach could be used for much more practical problems: how a doctor should communicate uncertainty, when a warning becomes alarmism, how evidence and urgency interact, how institutions frame responsibility, or how trust is gradually built or misplaced in ongoing AI-assisted conversations. In all such cases, the risk is not usually a single bad output. The risk is a trajectory. I’ve put a short companion page online — Trajectories in Human–AI Dialogue — that shows this in a more concrete way, including a simple example of how the different dialogue trajectories in this Substack could change what a person is likely to do next in real life
This is where I think the method may become useful for regulation, and not merely for theory. Regulators do not need a metaphysical answer to whether AI is “really” this or that. They need evidence about how harm emerges in interaction, and how to reduce it without destroying what is useful. That suggests that the right object of study may not be isolated prompts and responses at all, but streams of human/AI narratives, analysed over time under controlled conditions.
It also shows how a narrative stream involving a single human and and their chosen AI could be informative if the sequence is preserved and auditable. One could examine how trust accumulates, how framing effects drift, how caution turns into overconfidence, or how conversational order either stabilises or destabilises judgment. Under hard myndrama constraints, one can see exactly what was visible at each step, what was introduced, what was inferred, and where a sequence started to go wrong. That is a much stronger basis for regulation than panic on the one hand or benchmark complacency on the other. It gives regulators something concrete: dialogue traces, failure modes, sequence-level patterns, and possible intervention points.
Put differently, this kind of work could help shift the regulatory question from “Is AI dangerous?” to a more usable one: Which kinds of conversational sequences generate which kinds of harm, and how should we redesign those sequences? That is a better question because it treats interaction as the mechanism, not as a side effect.
So this little experiment about freedom ends up making a larger point. The hard constraints in Myndrama are not there to suppress suspicion of AI sentience. They are there to make the path of thought visible. Once that path is visible, it can be compared, critiqued, measured, and improved. And once we start thinking that way, conversational AI stops looking like a strange new oracle and starts looking like something we can study properly: a structured, path-dependent form of intelligence that becomes what it is in dialogue.
That, at least, is where I think the next stage of the work lies. If we want safer AI, we need better evidence of how conversations go right and how they go wrong. That evidence will not come from one-off prompts alone. It will come from auditable dialogue traces, where sequence, visibility, and state-change are preserved. In that sense, myndrama is not just a writing method. It is a small step toward a science of conversational governance. One note of caution. Orwell forewarned that the result of such a level of micro-governance could be NewSpeak. So the sooner we research alternatives the better!