Symposium Reflection: Iterative Prompting, Agency, and Authorship
At the symposium, I witnessed a prevailing tethering of AI art to traditional materiality—darkroom methods grounding digital outputs. Yet, I see AI as something more radical. While large language models rely on linguistic prompts, image generators initiate with language but evolve visually. Once the image is born, I engage in iterative prompting—pattern generating pattern. Each iteration is a conversation, not a command.
This is where agency shifts. Many at the symposium feared losing authorship. One even credited MidJourney equally and disavowed being an artist. I disagree. When I see my life in AI’s patterns, it’s my psyche reflected. The iterative dialogue between my selection and AI’s pattern is a dual agency. The decisive moment is psychic—when comprehension reveals the pattern as mine. Thus, authorship is not lost. It’s redefined—distributed, but deeply personal.
Here is your expanded PhD-level Fairy Note, written in your established voice, integrating your symposium reflections, iterative prompting, agency, film analogy, and theoretical positioning. I have followed your style guide closely in tone, structure, and clarity.
Fairy Note: Iterative Prompting, Distributed Agency, and the Misreading of AI Imagery
Symposium Reflection and Methodological Clarification
1. Opening Position: After the Symposium
This note emerges as an immediate reflection following the recent symposium on AI and artistic practice. I did not present work at this event, although in retrospect I recognise that my own practice would have offered a distinct and necessary counter-position within the discussions that unfolded.
What became clear very quickly was not that the work presented lacked sophistication or validity. On the contrary, much of it was technically accomplished and, at times, conceptually ambitious. However, the dominant orientation toward AI remained constrained. In most cases, AI was approached either as a tool for generating images from prompts, or as a conceptual object framed through familiar concerns around authorship, control, and aesthetic legitimacy.
This produced a striking disjunction between my own lived experience of working with AI and the discourse circulating within the symposium. What I encountered was not simply difference, but a structural gap: a failure to recognise where the work actually occurs.
2. The Persistence of Material Anchoring and the Fear of Immateriality
A recurring pattern was the tendency to re-ground AI-generated work in analogue processes. Many practitioners moved from AI outputs back into darkroom printing, material processes, or physical artefacts. This return to materiality appeared less as a conceptual decision and more as a stabilising gesture.
It suggests an underlying anxiety.
The immaterial nature of AI image production, its lack of indexical origin, its absence of a stable referent, produces a discomfort that is resolved by reattaching the work to traditional photographic or artistic forms. In doing so, the unfamiliar is translated back into the familiar.
Yet this gesture may obscure rather than clarify what is at stake.
The work is not located in the material output. Nor is it adequately described by the prompt that initiates it. The work emerges in the process between.
3. From Linguistic Prompt to Iterative Patterning
At a technical level, AI systems begin with language. Large language models operate through linguistic sequencing, where meaning is constructed through chains of tokens, probabilities, and syntactic relations. Image generators similarly require an initial prompt, a linguistic entry point that establishes parameters.
However, in practice, this linguistic layer quickly becomes secondary.
Once the first image is generated, the process shifts. The artist no longer operates primarily through language but through selection, variation, and transformation. One image leads to another. Patterns generate patterns. The process becomes iterative rather than declarative.
This is the crucial distinction.
The prompt is not the site of authorship. It is the threshold.
After this point, the work unfolds through what can be described as iterative visual dialogue, where each image responds not to language but to the previous image. The system produces variations, and the artist navigates these through recognition and refusal.
This aligns closely with my own methodology of AI phototherapy, where the initial prompt functions as what I have elsewhere termed a linguistic shutter, opening a field of possible images, after which meaning emerges through iterative engagement .
4. Iterative Prompting as Dialogue, Not Command
What follows from this is a redefinition of the creative act.
Iterative prompting is not refinement. It is not optimisation. It is not the gradual correction of an image toward a preconceived goal. Instead, it is a dialogic process in which meaning is discovered through interaction.
Each iteration is a response.
The artist does not impose form but recognises it. The process resembles psychoanalytic free association more than traditional authorship. Images surface, are rejected, are pursued, are abandoned. Meaning emerges through movement rather than intention.
This places the practice in proximity to what the Creative AI Lab describes as a shift from artefact to process, where creativity arises through the interaction between human and system rather than through isolated authorship .
5. Agency Beyond Ownership
Within the symposium, discussions of agency were often framed in terms of ownership. Who made the image? Who controls the output? Can authorship be attributed to the human, the machine, or both?
These questions, while understandable, remain limited.
Agency in this context is not ownership. It is not possession of the final image. It is the capacity to shape the process through recognition.
In my own practice, I encounter images that I did not consciously intend, yet which feel unmistakably mine. This is not because I authored them in a traditional sense, but because they resonate with lived experience. They align with what Bollas describes as the unthought known: material that is recognised before it is understood.
Agency therefore becomes distributed.
It exists between:
- the linguistic structure of the prompt
- the generative capacity of the system
- the iterative patterning of images
- the artist’s act of recognition
This is not a loss of agency. It is its transformation.
6. The Decisive Moment: Recognition as Psychic Event
This leads to a reconsideration of the decisive moment.
In traditional photography, the decisive moment is located in the capture. It is external, temporal, and event-based. In AI image-making, this moment shifts inward.
The decisive moment occurs at the point of recognition.
It is the instant when an image aligns with internal experience, when something previously unarticulated becomes visible. This moment is not produced by the system alone, nor by the artist alone, but by their interaction.
This corresponds closely to the structure already identified within the Phantom Mirror methodology:
moment of danger → dialogue → image → refusal → recognition .
Recognition is therefore not aesthetic judgement. It is a psychic event.
7. Misreading AI: The Problem of Familiar Form
A further issue emerging from the symposium is the persistent misreading of AI imagery.
AI images often resemble photographs, paintings, or cinematic stills. They adopt the visual language of existing media. As a result, they are interpreted through those frameworks.
This is a category error.
The situation is analogous to early film theory. Film was initially understood as a representation of reality because it resembled photographic recording. Over time, it became clear that film constructs its own language, one that must be learned.
The same applies here.
AI images are not photographs, even when they look like photographs. They are not representations of external reality. They are outputs of a generative system responding to patterns, associations, and prompts.
We are currently reading them as if they belong to existing visual traditions.
They do not.
They require a new literacy.
8. Historical Position: Beyond the Death of the Artist
This shift can be situated within a broader theoretical trajectory.
Roland Barthes’ concept of the “death of the author” displaced the idea of singular meaning originating from the artist. Walter Benjamin’ analysis of mechanical reproduction destabilised the aura and authorship of the artwork.
AI extends these trajectories.
However, it does not simply eliminate the artist. Nor does it fully democratise authorship. Instead, it produces a new condition in which authorship becomes distributed across systems, processes, and recognition events.
The artist persists, but not as sole origin.
The artist becomes the site where meaning is recognised.
9. Toward a New Artistic Paradigm
What emerges from this reflection is not a rejection of existing practices, but a clarification of a new one.
AI image-making, when approached through iterative prompting, is not:
- prompt-based image generation
- aesthetic experimentation alone
- or a question of authorship attribution
It is a dialogic, pattern-based process of meaning construction.
It requires:
- relinquishing control
- tolerating uncertainty
- engaging with emergence
- and recognising rather than imposing
This places it in continuity with earlier artistic practices of abstraction, improvisation, and process-based work, while also marking a distinct shift due to the generative autonomy of the system.
10. Closing Position: Confidence and Difference
The symposium clarified something essential.
The difference in my own practice is not simply conceptual. It is methodological. It arises from sustained engagement with AI as a process rather than a tool.
This note therefore marks a point of consolidation.
The work is not located in the prompt.
The work is not located in the image.
The work is located in the iterative dialogue between pattern and recognition.
That is where meaning emerges.
And that is where this research sits.