FractalPerception.com
Exploring Aesthetic Perception in High-Dimensional Parametric Fractal Systems
1. Making Aesthetic Choices Observable to AI
FractalPerception is an independent research program at the intersection of artistic creation, parametric fractal systems, aesthetic perception and artificial intelligence.
The project has a distinctive feature: the images are not available only as pixels. They remain connected to their generative structure — mathematical functions, parameters, transformations, layers, gradients and blending modes.
This dual access makes it possible to connect:
generative structure ↔ visual rendering ↔ perception ↔ aesthetic decision
The compositions are non-figurative and non-referential. Analysis can therefore focus directly on their visual structures — forms, rhythms, densities, colors and balances — without first depending on the recognition of a represented subject.
FractalPerception seeks in particular to determine whether an artist’s repeated aesthetic choices — what is retained, rejected, transformed or combined — reveal regularities coherent enough to contribute to describing an aesthetic grammar.
In the longer term, this work aims to make the formation of human aesthetic judgment progressively more understandable to AI.
2. An Existing Parametric Artistic Corpus
FractalPerception is developed by Philippe Chevalier / Philoxerax, a digital artist and developer. The project takes as its starting point his work with fractals, begun in 2020.
The corpus currently includes:
- 376 fractal compositions
- 1,020 parameterized layers
- 1,396 shapes
- 116 unique functions
- 34,369 composition-level parameters
- a high-resolution bitmap rendering for each image
This first phase deliberately focuses on a coherent and identifiable artistic sensibility so that its choices of creation, transformation and selection can be analyzed as a whole.
A first public version of this historical corpus is currently being prepared for Hugging Face.
Visual Corpus Documentation
Structural Corpus Documentation
3. A Co-Creation Channel Between the Artist and AI
AI can read the parametric files that constitute the saved states of the artworks and thereby access the functions, parameters, transformations and structures underlying their visual rendering.
It can analyze them, compare choices made across different works and search for recurring patterns in the artist’s aesthetic decisions.
But the channel also works in the other direction: AI can directly generate new parametric files that can be used by the creation software.
Following a direction given by the artist, it can explore a function or a region of parameter space at scale and produce series of new forms. The artist observes the results, selects, rejects, transforms and decides on the next direction.
AI explores at scale; the artist directs the exploration.
FractalPerception thus establishes a bidirectional channel between the artist and AI through parametric space.
The dialogue no longer takes place only through words and prompts: AI can act directly within the generative material of the artwork, while the artist’s choices in turn become observable to AI.
4. A New Stage: From an Image Corpus to a Corpus of the Creative Process
The initial corpus makes it possible to study completed artworks and the parametric structures that constitute them.
The new bidirectional channel also makes it possible to document information that usually disappears when only the final artwork is archived:
- explored proposals;
- selections and rejections;
- transformations;
- successive versions;
- bifurcations and abandoned branches;
- the origin and evolution of shapes;
- aesthetic decisions between two states;
- the dialogue between the artist and AI.
An artwork can thus be described as a parametric genealogy of decisions.
FractalPerception now has the means to build a corpus of the creative process: documenting not only the aesthetic result that was retained, but also the trajectory of decisions that led to that result.
5. Two Complementary Partnerships
FractalPerception is now seeking to establish, in parallel, a scientific partnership and an institutional and artistic partnership around this new framework.
Scientific Partnership
The objective is to methodologically structure this corpus of the creative process so that it can become a genuine research resource usable by AI.
In particular, the aim is to study recurring patterns in aesthetic decisions and the mechanisms involved in the formation of judgment.
Relevant fields include:
- artificial intelligence;
- visual perception;
- computational aesthetics;
- generative systems;
- human–machine interaction.
Institutional and Artistic Partnership
The same process also produces material that can be made visible to the public.
An exhibition can show not only the final artwork, but also the AI’s proposals, the artist’s choices, transformations, bifurcations, successive states and their parametric evolution.
It thus becomes possible to observe from within a creative process between an artist and AI.
FractalPerception is seeking an institution with which to develop the most appropriate form for presenting this process to the public.
Philippe Chevalier / Philoxerax
Philippe Chevalier / Philoxerax is a digital artist and developer, and a graduate of EnsAD Paris. He has worked for more than thirty years at the intersection of art, code and digital systems.
He is the architect of FractalPerception, the author of the initial corpus and the person making the aesthetic decisions currently being studied.
In the longer term, the framework can be opened to other observers and then to other artists in order to compare different sensibilities and distinguish what is singular from what may correspond to shared regularities.