Subconscious Computation

The biographical and theological ground (ZORA ZINE, 2021)

The ZORA ZINE interview provides the most direct statement of what "subconscious computation" actually means, grounded biographically and philosophically:

"By creating conditions that have certain outputs that are not fully predictable, I think you can even say that the unconscious is manifested through that."

The context: van den Dorpel's background in Jungian psychoanalysis ("I am fascinated by psychoanalysis") and his upbringing in a religious household ("My father has run a church at some point; theology has always been in my life"). Jung's model — the unconscious as a structured, symbolic, impersonal dimension that exceeds conscious control — maps onto the generative algorithm: both produce what the conscious mind cannot directly intend.

The threefold parallel is stated explicitly:

"Theology is interpreting scripture or deciphering language. As a literary writer, you have power by creating worlds; when you're a coder, you write conditions; and with generative art, you are, on the one hand, a creator, and [because] you cannot fully prepare yourself for what is coming out of this algorithm…you're also a bystander."

The creator/bystander doubled position — which process-legibility calls "both author of the rules and subject to them" and senescenence calls "less composed than initiated" — is here rooted in a theology of creation: the relationship to conditions that exceed the one who sets them. The generative system is not a tool but a site of revelation.

Our Inner Child (2023): from theory to visceral impulse

The exhibition text for Our Inner Child (Upstream Gallery Amsterdam, 2023) offers the most autobiographically direct account of what the subconscious ground of the practice actually feels like — not as philosophical position but as lived crisis and therapeutic recovery.

After a period of recognition (Stedelijk Museum acquisition, institutional arrival of generative art), van den Dorpel found himself overwhelmed. Seeking professional help, he encountered Inner Child therapy — a modality distinct from the Jungian analysis that grounds the ZORA ZINE account. Where Jung maps structural archetypes of the collective unconscious, Inner Child work addresses the personal biography: specific formations of the child self that persist into adult experience.

The therapy sent him back to "the feelings that first inspired me to make images." This is the most precise biographical statement of what impulse and risk calls the "deepest impulse" (→ 793): not an abstract principle but the specific affective origins of the practice — feelings, not theories. For years, van den Dorpel notes, he had been "engaging in highly theoretical and technologically sophisticated approaches to art," and this cerebral mode had come at a cost: "I felt I was obscuring a part of myself." The theoretical apparatus, however generative, had been functioning as a kind of self-concealment.

The Resolution Paintings that resulted are described as an embrace of "immediate sense of joy" — an insecurity resolved: where earlier he "might have sought something more supposedly 'intellectual' to buttress a creation I viscerally enjoyed," here the visceral is admitted as the legitimate starting point.

*Nacre* (2023) — Resolution Paintings; 150×150cm; photographic exposure on metallic paper; the decompressed form at full material resolution

The religious algorithm

Our Inner Child also extends the religious background in a new direction. Where the ZORA ZINE account uses theology positively — as a model for the creator/bystander position, the relationship to conditions that exceed the one who sets them — the Our Inner Child text uses the same biography critically:

"I spent many years trying to figure out the rules — the algorithm, one might say — to being a good person. How could I do this and avoid the terrors of the Hell that I was told about, and which I was constantly at risk of being sent to. Therapy here has helped me understand the limits of narrow legislative approaches to morality, both towards oneself and towards others."

The word "algorithm" applied to the childhood religious moral code is precise. It names exactly what an algorithm is: a rule-set intended to produce correct outputs, applied to the self as substrate. The legislative approach — enumerate the rules, follow them, produce correct behaviour — is the same model that the generative practice refuses at the level of art-making. The system of conditions that allow something unanticipated to emerge is the opposite of the algorithm as moral legislature.

This does not contradict the ZORA ZINE account. Both are true simultaneously: the religious background is both a resource (the creator/bystander theological model, the interpretive relationship to scripture) and a formation to be worked through (the legislative algorithm, the moral rule-set as self-obscuration). Therapy helps distinguish them. The practice enacts that distinction in its form: not rules that must be followed toward a predetermined correct output, but conditions that allow something to emerge that the rule-setter could not have specified.

The phrase and its stakes

The artist bio describes systems that "draw on intuition and subliminal processes of the mind in order to continually output unexpected and curious aesthetic forms that embody a feeling of subconscious computation."

Subconscious computation is not just a description of medium — it makes a philosophical claim about what the algorithm is for. The system is designed to externalize or model subliminal cognitive processes: to produce what the conscious, intentional mind cannot directly generate or predict. The output is "unexpected" not because the system is random but because it operates below the threshold of deliberate control.

This raises the question: what does it mean to design a system for outputs you cannot predict? The designer is not the author of the specific form — only of the generative conditions.

The specific algorithm: Cartesian Genetic Programming

The generative engine in Mutant Garden is Cartesian Genetic Programming (CGP), developed by Julian F. Miller and Peter Thomson in the late 1990s for circuit board optimisation. Van den Dorpel implemented his own version in TypeScript after encountering it in 2019. CGP represents a programme as a two-dimensional grid of computational nodes; mutation proceeds by modifying genes in the chromosome. Crucially, the graph structure guarantees safe mutation — no infinite loops, no crashes. This solved the specific technical problem that halted earlier attempts to mutate code directly: the halting problem.

The non-coding gene is the most philosophically interesting feature: genes that contribute nothing to the current computation persist, mutate, and can activate later, producing new structures from dormant material. The system carries more potential than it currently expresses. See evolutionary logic for the full account.

The algorithm as a device for following the deepest impulse

Note 793 argues that the most valuable action is discovery rather than decision: "find your deepest impulse, and follow that… the notion that one trusts what is so discovered, although unclear where it will lead." The problem is that the deepest impulse is by definition not available to direct introspection — it is subliminal.

The algorithmic system offers a solution to this problem. If you can encode the conditions that correspond to your intuitions — the aesthetic criteria, the structural tendencies, the feedback parameters — the system can follow the impulse further and faster than conscious deliberation would allow. The output is alien enough to surprise its own maker, yet recognizably consonant with the maker's sensibility.

This is a different understanding of AI in art from both "tool use" (the artist uses AI to execute intentions) and "generative art" (the system produces random variation). It is closer to: the system makes explicit what was previously only implicit in the practice.

Feedback loops as design

The bio specifies that works are "informed by feedback loops and the design of algorithmic systems." In 709 (artist statement), the feedback loop appears as a description of the practice's complexity: "every single parameter depends on all the others, and my interference amplifies as a feedback loop." There it is a condition encountered; here it is something designed.

The shift from metaphor to architecture matters. A feedback loop built into a software system is not just a way of describing complexity — it is a mechanism that makes the system's outputs depend on its own prior outputs. The work is continuously evolving not because the artist keeps revising it but because the system's history shapes its next state. This connects directly to assemblage identity: the identity of the evolving work is constituted by its history, not only its current state.

Net art and the extended lineage

The bio places the practice "within and beyond the lineage of 'net art.'" Net art (1990s–2000s) foregrounded the network as both medium and subject — distribution, participation, link structure, the browser as canvas. Working "beyond" this lineage means the network logic has been internalized into the generative system itself: the connections are now inside the algorithm, not just between websites.

The continuity: the fundamental concern with how structure produces meaning — how the shape of a system (a network, an algorithm, a feedback loop) determines the kinds of things that can emerge from it. This is also the concern of assemblage theory: the emergent properties of a whole are not reducible to its parts but follow from the relations between them.

Craft and skill as enabling conditions

The bio insists on "immense skill and craftsmanship." This is significant against a background where algorithmic and AI art is often presented as post-skill — as if the machine does the work. Here skill is what makes the system capable of subliminal output rather than merely arbitrary output. The difference between a generative system that produces unexpected-but-resonant forms and one that produces noise is the craftsmanship of the designer.

This connects to workmanship of uncertainty (609): it is not randomness but the disciplined engagement with uncertainty — the "reconciliation of accident" — that produces genuine mutation.

Fitness functions as formalised impulse (Spike #70)

The Spike #70 interview names what was previously implicit: a fitness function is the mechanism by which the generative system knows which outputs are better than others. Without a fitness function, randomness produces "no direction or growth." With one, the system can evolve toward something — not a predetermined end, but a direction. The fitness function is the computational encoding of the deepest impulse: what 793 calls "discovery" rather than "decision" is here formalised as a criterion that the system applies in place of conscious deliberation.

The progression across works — manual selection (Death Imitates Language), crowdsourced (Hybrid Vigor), structurally defined (Nested Exchange's hipster algorithm, Mutant Garden's construction/experience complexity) — is a systematic externalisation of taste into code, each iteration delegating more of the selection to the system itself.