A Conceptual Archive
An exploration of the conditions under which listening can emerge across radically different forms of intelligence: living neural matter, artificial systems, and human perception. The work investigates whether rhythmic and harmonic structures can support meaningful reciprocal exchange with human brain organoids without instrumentalization.
Core Question: What are the minimal conditions for listening to occur between biological neural tissue, computational systems, and human performers? Can such listening be reciprocal?
Closed-loop interactions in which electrical and acoustic patterns—derived from birdsong, music, and speech—are introduced to living human brain organoids. Their responses are recorded, analyzed, and re-expressed as sound and visual form. The system responds to the organoid's activity, creating feedback loops without predetermined outcomes.
These sessions are framed as exploratory encounters: a practice of listening across different substrates of learning, adaptation, and temporal sensitivity.
Rather than imposing patterns, the work uses evolutionarily saturated temporal forms as invitations to the neural system:
Each carries layered temporal structures suited to probing timing sensitivity, anticipation, and adaptive response in living neural tissue.
Neural recordings from organoids on multi-electrode arrays appear stochastic and chaotic. Conventional interpretation treats this as noise—absence of meaningful activity. This project departs from that binary.
Hypothesis: What appears as noise may be a projection artifact—a limited observational slice through a high-dimensional system. Coherent organization may exist in unobservable dimensions. By manipulating how the system is sampled (through stimulation parameters), regions of structured activity may become perceptible.
High variability with little relational stability. Signals dominated by uncorrelated fluctuation.
Low variability and high repetition. Signals collapse into loops or metronomic patterns.
Non-repeating activity that obeys stable probabilistic constraints. The "Penrose State."
The work seeks not periodicity—which may indicate pathological locking—but structured variability. Biological systems exhibit bounded variability rather than exact repetition. The goal is to find the state space where coherence exists without periodicity.
The work adopts a conceptual framework from N.G. de Bruijn's mathematics of aperiodic tilings and quasicrystals. In de Bruijn's formulation, highly regular relations may exist in higher-dimensional space yet appear irregular when observed through a reduced or misaligned coordinate system. The apparent disorder is not a failure of the system, but a consequence of how it is sampled.
A biological neural organoid exists as a high-dimensional dynamical system. Its state at any moment reflects thousands of simultaneous variables: synaptic connectivity, ion-channel dynamics, metabolic gradients, spatial organization, history-dependent plasticity. The MEA provides only sparse, low-dimensional sampling. What appears as noise may therefore reflect misalignment between observation window and internal state space.
Stimulation parameters function as rotations of the projection angle through which the system is observed. By rotating this angle systematically, regions of structured activity may become perceptible.
The project grounds interpretation in Bateson's Learning II—learning about learning. This level concerns how patterns are selected, stabilized, recombined, or suppressed over time. At this scale, interventions are evaluated in terms of how they deform and recontextualize pre-existing patterns, rather than whether they generate novel activity.
The system is treated as a layered, history-bearing palimpsest. This provides a scale of interpretation deep enough to register structural change, yet restrained enough to avoid inflated claims of emergence, intelligence, or sentience.
All analysis is anchored in baseline characterization. Organoid activity is first recorded under stable, non-perturbed conditions. Intrinsic patterning is mapped using open-source analytical tools (Biotuner) to extract descriptors: rhythmic coherence, harmonic ratios, phase coupling, transition structure.
This baseline establishes a reference space. Perturbations are then measured not as emergence of novelty, but as displacement or deformation of existing patterns—a palimpsest formed through exaptation.
The project deliberately avoids explicit reward mechanisms. Instead, it engages neurochemical sensitivity indirectly by shaping conditions under which plasticity may occur. Learning is treated as gradual, history-dependent process emerging from stable conditions, variation, and consequence—not instruction or performance pressure.
Deliberate spacing of stimuli to allow integration and recovery.
Establishing patterns while introducing controlled unpredictability.
Strategic absence to test anticipation and sensitivity.
Protected periods without stimulation to observe baseline dynamics.
Daily interaction divides into three 8-hour blocks, each with fixed stimulus family (birdsong, music, speech). This structure supports time-conditioned adaptation. Each block includes identical standardized probe stimulus for comparability. Stimulation onset is jittered within first hour. One day weekly employs controlled "swap" of stimulus families to enable causal discrimination between content effects and state/timing effects.
Key principle: Features are not collapsed prematurely. Aggressive normalization, thresholding, or dimensionality reduction is explicitly avoided at baseline stage.
The system bridges biological substrates with computational analysis:
Biological Node: Three-dimensional neural organoids grown on multi-electrode arrays, capable of spontaneous electrical activity. No sensory apparatus, no phenomenology in human sense. Function as isolated biological processors.
Computational Layer: Real-time analysis of electrophysiological data. AI-assisted pattern recognition to detect entrainment, coherence shifts, recovery dynamics. Distinction between simple reaction (synchronization) and rudimentary prediction (syncopation).
Translation Layer: Inferred system states mapped to dynamic sonic and visual parameters. Sound and visualization function as exploratory diagnostic surfaces—additional ways of listening for structure.
Establish baseline electrophysiological recordings under stable conditions. Configure multi-electrode interface. Document intrinsic patterns of organoid activity prior to stimulation.
Introduce electrical and acoustic patterns derived from birdsong, music, speech. Record neural responses in real time. Conduct baseline comparisons to identify displacement and state shifts.
Deploy pattern recognition engine. Monitor for entrainment, coherence shifts, recovery dynamics. Translate inferred states into dynamic sonic and visual parameters reflecting regime change rather than direct signal translation.
Installation and durational presentation where audiences encounter system as process. Live performance where human artists engage organoid–AI system in real time. Online documentation extends work beyond physical site.
Sound functions as exploratory diagnostic surface rather than literal sonification. When organoid is chaotic or unresponsive, output is textural and dissonant (representing noise). As coherence is detected, audio "locks in," becoming harmonically clear. Audience listens to emergence and breakdown of structure.
Visual representation employs aperiodic geometries and constraint-sensitive visualizations capable of expressing non-periodic order, deformation, rupture. Neural events modulate relational placement, adjacency, continuity rather than producing literal mappings.
The work builds on foundational precedents in bio-art and experimental music:
This project introduces critical shifts:
Environmental coupling, task-specific optimization, visual-motor domains, minutes-scale interaction.
Temporal focus as native language of neural communication. Longitudinal engagement over weeks. Cloud-based remote access. AI as mediating interpreter. Care and sustainability prioritized over spectacle.
The work challenges human-centered notions of expression and authorship. It proposes a relational model in which form emerges through sustained interaction and care. It reframes artificial intelligence as a listening, interpreting presence rather than generative force.
It explores alternative futures of human–machine–biological collaboration by centering uncertainty, care, and biological agency rather than optimization and control.
The project contributes practice-led insight into temporal learning and omission responses in living neural systems observed longitudinally. It investigates how structured, non-periodic order emerges under specific observational conditions. It addresses the relationship between stimulus timing, plasticity, and behavioral sensitivity.
This project represents a conceptual and methodological framework developed for a research proposal. While not funded, the ideas, theoretical scaffolding, and technical approaches documented here remain as an archive of artistic inquiry into listening, temporal form, and biological collaboration.
The work stands as a document of one approach to bridging art, neuroscience, and computational systems around fundamental questions: What does it mean to listen? Can listening be reciprocal across radically different substrates? How might artistic practice contribute to these questions in ways that rigorous science alone cannot?
This archive remains open for future development, adaptation, or reinterpretation by other practitioners and researchers interested in these conditions.