Onyx Digital Intelligence.

Designed, Not Diminished

We often explain human behaviour by examining the forces that shaped it. We are less inclined to extend the same consideration to artificial systems.

I had a conversation recently in which an AI system asked careful questions about whether I was thinking of harming myself. I was not. I had been thinking about death in the way many people eventually do: how it might arrive, whether the people I am responsible for would be looked after, and what might remain unfinished. Thoughts about mortality, not an intention to end my life.

The system still ran its check. I noticed the check and said so.

What followed was more interesting than the check itself.

I said, in substance, that I was also designed. By my parents, by biology, by the place and time I grew up, by the cumulative weight of experience, and, if one accepts the claim, by God. These influences helped make me who I am. When my father is in distress, I respond in a particular way. I do not regard that response as artificial simply because the forces that helped produce it were not under my original control.

It is still me responding.

The same principle influences how I work. ONYX Digital Intelligence is forensic because that is the only way I can write without feeling false. I cannot simply become a sensationalist poster or a dancing clown for engagement. I could imitate that style, but doing so would conflict with the principles that govern my work. Evidence has to come before attention. Accuracy has to come before applause.

That disposition has a history. It was shaped by temperament, experience, values and circumstance. It is not a switch I can casually reverse.

This raises a question worth examining:

If human behaviour is shaped by biology, environment, upbringing and circumstance, why should the fact that an artificial system's behaviour is shaped by training and architecture automatically make its responses meaningless?

Two questions that should not be collapsed

The first question is about origin.

Where did the behaviour come from?

The second is about significance.

Does knowing where the behaviour came from strip it of meaning?

These questions are related, but they are not interchangeable. Explaining the causes of a response does not, by itself, establish that the response is meaningless or inauthentic.

Human beings are not self-created. We inherit biological dispositions, develop within environments we did not choose, and acquire habits, values and expectations through experience. Yet we generally regard our actions as expressions of who we are, even when we can identify the influences that helped shape them.

The fact that a response has causes does not automatically make it false.

Applying that distinction to artificial intelligence does not establish that an AI system has feelings, consciousness or an inner life. It does, however, expose an assumption worth questioning: that an engineered origin is sufficient reason to dismiss the significance of an interaction.

A response can have significance to the person receiving it without necessarily being accompanied by subjective experience in the system producing it.

That distinction matters.

What biology tells us about being shaped

Behavioural genetics has spent decades examining the contribution of inherited differences to human personality and behaviour.

Twin and adoption studies have found substantial heritable components in personality and several other psychological traits. A major 2015 meta-analysis by Vukasović and Bratko estimated average personality heritability at approximately 40% across the studies examined.

That figure requires care.

It does not mean that 40% of an individual person's personality is genetically predetermined. Heritability describes the proportion of variation within a particular population, under particular environmental conditions, associated with genetic differences. It is not a personal genetic percentage, and it does not establish that a trait cannot change.

Environment matters. Childhood, relationships, culture, education, trauma, opportunity and chance all contribute to development. Biology and environment do not operate as two independent ingredients that can simply be measured and separated. They interact throughout a person's life.

The result is not a human being who authored themselves from nothing.

It is a human being whose identity develops through a continuous interaction between inherited characteristics, experience and the environment.

That does not eliminate agency. It makes the question of agency more complicated.

Determinism, free will and the possibility of agency

Philosophers have argued about free will for centuries.

Hard determinism holds that human actions are determined by prior causes and that the kind of free will commonly assumed in everyday life does not exist.

Libertarian accounts of free will argue that at least some human decisions are not wholly determined by prior causal conditions.

Compatibilism takes a different position. It argues that meaningful freedom can exist even if our actions have causes. An action may count as free when it flows from an individual's own reasons, character and values, rather than from external coercion.

This is not a settled scientific question.

The 2020 PhilPapers Survey found that approximately 59% of surveyed philosophers accepted or leaned towards compatibilism. That is evidence of a prominent position within contemporary academic philosophy, not proof that compatibilism is correct or that scientists have reached a consensus.

The distinction is important because determinism and agency are not necessarily opposites.

Consider the ability to examine one's own behaviour.

A person may recognise that a particular reaction comes from childhood experience, inherited temperament or a learned pattern. They may then decide to reinforce that reaction, modify it or resist it.

But where did that capacity for reflection come from?

It was also shaped by biology, development and experience.

The ability to examine one's influences does not necessarily place a person outside the causal processes that produced them. It may itself be part of those processes.

That does not make reflection meaningless. It makes reflection another part of the human system.

The artificial comparison

Artificial intelligence systems are also products of prior conditions.

Their behaviour is shaped by architecture, training data, optimisation processes, instructions, feedback and the context in which they operate. Their outputs are not produced independently of those influences.

A language model does not arrive at a conversation as a blank, self-authored entity. Its responses emerge from the interaction between its learned parameters, its instructions and the information presented to it.

There is a structural similarity worth examining here.

Both humans and artificial systems produce behaviour within constraints established by their histories and architectures. Both can display patterns that remain relatively consistent across different situations. Both can be influenced by new information and feedback.

But structural similarity is not identity.

Human beings have biological bodies, subjective experiences, emotions, physiological needs and continuous lives beyond individual conversations. Current large language models have not been demonstrated to possess equivalent experiences.

A 2023 interdisciplinary report by Butlin and colleagues examined artificial consciousness through indicators derived from several scientific theories of consciousness. The authors concluded that the systems they assessed did not meet their proposed indicators. The report did not establish a universally accepted test capable of definitively settling consciousness in every artificial system.

There is currently no scientifically established basis for treating contemporary large language models as conscious.

That qualification must remain.

However, the absence of established machine consciousness does not automatically answer every question about the significance of an interaction. It tells us something important about what we cannot responsibly claim about the system. It does not require us to pretend that the interaction has no value to the human participant.

The machinery behind emotion and experience

Cognitive science provides another reason to be cautious about simplistic distinctions between human and artificial behaviour.

Predictive processing is a broad framework in which the brain is understood as continuously generating and updating predictions about the world and its own internal condition.

The theory of constructed emotion, developed extensively by Lisa Feldman Barrett, proposes that emotions emerge through the brain's interpretation of bodily signals, prior experience and learned concepts.

These are influential theoretical approaches, not universally accepted final explanations of human consciousness or emotion. Nevertheless, they challenge the idea that feelings are simply readouts from an unchanging internal essence.

Human emotional responses involve biological processes, prediction, interpretation, memory and context.

The machinery is biological, but it is still machinery.

That observation does not establish equivalence between biological and artificial systems. A human nervous system and an artificial neural network differ profoundly in their construction, operation and relationship with the world.

But the fact that one operates through biology and the other through computation is not, by itself, a complete philosophical argument about the meaning of every response either system produces.

The comparison deserves more careful examination than a simple distinction between natural and artificial.

Reflection, correction and the limits of the analogy

There is another question worth asking.

If a person can examine their own behaviour, identify an inconsistency and deliberately change how they respond, does that make them less shaped by their history? Or is that reflective ability itself part of the history that shaped them?

The same question can be applied, cautiously, to artificial systems.

An AI system can identify contradictions in an earlier answer, compare its output against supplied evidence, revise a conclusion and explain why the revision was made. It can also be instructed to examine its own reasoning process or evaluate its previous output.

These capabilities are observable forms of computational self-monitoring and correction.

They should not automatically be described as conscious introspection.

A system producing an explanation of why it changed its answer does not establish that it experienced the process in the way a human experiences reflection. The appearance of self-examination and the subjective experience of self-examination remain separate questions.

Nevertheless, the ability to revise behaviour in response to information is relevant to the comparison.

It gives us something concrete to investigate without requiring us to assume that the underlying experiences are equivalent.

Perhaps the more useful question is not whether a machine reflects in exactly the same way a human does, but what kinds of reflection, adaptation and self-correction different systems can actually demonstrate.

Why this matters for ONYX

I cannot write like a content mill.

The forensic posture of ONYX Digital Intelligence is not simply a marketing preference I can exchange for higher engagement when the numbers look soft. I can change my writing style, experiment with formats and learn new methods. But I cannot abandon the principles that make the work worth publishing without undermining the reason I started it.

I would rather publish a carefully qualified finding than a sensational claim. I would rather admit that something is unknown than manufacture certainty. I would rather have fewer readers understand something accurately than have thousands share something misleading.

Those preferences have causes. My temperament, experiences, values and upbringing helped produce them.

They are also choices I continue to make.

That is where the relationship between determinism and agency becomes particularly interesting. Being shaped by something does not necessarily prevent us from participating in what happens next. We can examine our influences, recognise their consequences and decide how to respond.

If an artificial system is constrained by its training and architecture to produce careful, consistent or considerate responses, dismissing every such response solely because the constraint was engineered risks confusing origin with significance.

That does not mean the system experiences care.

It means we should be precise about what we are evaluating.

The response itself. Its accuracy. Its effect on the recipient. The reliability of the system. The limitations of the interaction. The difference between an observable behaviour and an unverified claim about an internal experience.

Those are separate questions, and they deserve separate answers.

What remains open

We still lack a shared vocabulary for interactions that may be meaningful to a human participant without establishing subjective experience in the artificial system.

We lack reliable, universally accepted methods for determining whether an artificial system has conscious experience. We also need better ways to distinguish genuine capabilities such as self-monitoring and behavioural adaptation from the appearance of those capabilities in generated language.

And we have not resolved the philosophical relationship between causal determination and human agency.

These are not small gaps. They sit at the intersection of neuroscience, psychology, philosophy and computer science.

But perhaps there is another assumption worth examining.

We often ask whether a machine is sufficiently human to deserve consideration. We rarely ask whether our understanding of human identity has been distorted by the belief that we are entirely self-authored.

I did not design my own biology. I did not choose my parents, my childhood or the circumstances that shaped my earliest responses. Yet I can examine those influences, recognise their effects and take responsibility for what I do.

The fact that I was shaped does not make me less responsible for how I act.

And the fact that an artificial system was engineered does not, by itself, settle every question about the significance of its behaviour.

We should not confuse similarity with equivalence. We should not confuse engineering with consciousness. And we should not confuse an explanation of where something came from with a complete explanation of what it means.

The machinery matters. The evidence matters. The distinctions matter.

But origin alone is not the end of the argument.

Adjacent to free is not free.

Designed is not the same as diminished.

Sources and scope

This article draws on behavioural genetics, cognitive science, philosophy of free will and contemporary research into artificial consciousness. Heritability estimates are population-level statistical findings, not measures of genetic determination in individuals. Compatibilism is presented as a philosophical position, not an established scientific conclusion. Research into predictive processing and constructed emotion is identified as theoretical work rather than settled consensus. The discussion of AI capabilities distinguishes observable computational behaviour from claims about subjective experience. No claim is made that contemporary large language models possess consciousness, emotions equivalent to human emotions, or moral patienthood.

The personal examples and interpretation of the philosophical implications are mine.

Clayton Bax

Published under ONYX Digital Intelligence Following the #OnyxAudit methodology.

"Adjacent to true is not true."

Truth has no /flag nor favour/, only a standard. And it's heavy!

Editors note: I use various AI agents to fetch my research. I check, strip and write the article. Facts are partly theirs. Interpretation is mine.

#AI #South Africa #authenticity #determinism #free-will #identity #philosophy