Quantified Co-Creative AI Systems

Unlocking the Black Box of Artificial Media with Explainable and Quantified Co-Creative AI

As generative AI systems become increasingly integrated into creative practice, a central challenge emerges:

How do we understand, explain, and evaluate the collaborative processes unfolding between humans and AI systems during co-creation?

Artificial media is often treated as a black box of combined creative effort. While both the human user and AI system contribute to the final artifact, their respective roles, decisions, interaction patterns, and influence on the creative trajectory are frequently obscured within the completed work itself.  

This research introduces the concept of:

Quantified Co-Creative AI Systems

Quantified co-creative AI systems are co-creative systems that:

• quantify,

• model,

• visualize,

• and explain

the interaction dynamics unfolding during human–AI collaboration.  

Rather than evaluating only the final creative output, quantified co-creative systems analyze the collaborative process itself — including interaction rhythms, communication patterns, turn-taking behavior, cognitive dynamics, and evolving creative trajectories across time.

This work proposes that the interaction history behind artificial media can function as a form of:

creative provenance.

Unlocking the Black Box of Artificial Media

Traditional generative AI systems often conceal the interaction processes responsible for producing a creative artifact.

This paper proposes that explainability in co-creative systems should extend beyond model transparency and include:

• interaction transparency,

• collaborative transparency,

• and creative process transparency.  

By quantifying co-creative interaction, it becomes possible to examine:

• who contributed what,

• when contributions occurred,

• how collaboration evolved,

• how ideas emerged,

• and how human and AI participation shaped the final artifact.

The paper describes quantified co-creative systems as providing:

“an x-ray into understanding the creative product.”  

The Co-Creative Sense-Making (CCSM) Framework

At the center of this research is the:

Co-Creative Sense-Making (CCSM) Framework

The CCSM framework is introduced as a cognitively grounded framework for quantifying and explaining co-creative interaction dynamics in human–AI collaboration.  

The framework integrates concepts from:

• participatory sense-making,

• creative sense-making,

• improvisation theory,

• enactive cognition,

• and co-creative AI research.  

CCSM organizes co-creative interaction into four primary dimensions:

Cognitive Dynamics

Cognitive dynamics describe the modes of cognition unfolding during co-creation, including:

• creative flow,

• exploration,

• communication,

• interface manipulation,

• and creative execution.  

The framework introduces the:

Creative Sense-Making Curve

This curve visualizes cognitive fluctuations and interaction patterns unfolding across time during collaboration.  

The resulting interaction traces provide insight into:

• engagement,

• cognitive regulation,

• interaction rhythms,

• and collaborative flow states.

Interaction Dynamics

Interaction dynamics capture:

• turn-taking,

• timing,

• communication strategies,

• interaction couplings,

• feedback,

• synchronization,

• and coordination patterns.  

The framework examines how:

• human and AI contributions become structurally linked,

• collaborative momentum emerges,

• and participatory sense-making develops across interaction.

Interaction dynamics provide a computational lens into:

how collaboration unfolds.

Collaboration Dynamics

Collaboration dynamics analyze:

• improvisational offers,

• acceptance and rejection patterns,

• elaboration behavior,

• leadership and follower dynamics,

• and collaborative influence.  

This dimension treats co-creation as an evolving relational process in which:

• ideas are proposed,

• transformed,

• negotiated,

• and expanded collaboratively.

Domain Dynamics

Domain dynamics capture:

• action histories,

• interaction logs,

• creative output metrics,

• interface behavior,

• and domain-specific creative actions.  

In the domain of drawing, for example, this includes:

• line production,

• drawing duration,

• sketch transformations,

• object generation,

• and interface interactions.

These dynamics provide measurable insight into:

how creative activity unfolds within a specific creative medium.

Explainable Co-Creative AI

This work argues that quantified co-creative systems are inherently:

explainable co-creative systems.

Rather than merely explaining model outputs, explainable co-creative AI systems can explain:

• interaction history,

• collaborative reasoning,

• creative trajectories,

• communication patterns,

• and participation structures.

The framework enables AI systems to contextualize:

• why certain creative decisions emerged,

• how collaboration evolved,

• and how interaction shaped the final artifact.

This approach connects explainability directly to:

• Human–AI Interaction,

• participatory cognition,

• enactive social interaction,

• and collaborative creativity.

Adaptation and Co-Evolution

The paper also proposes that quantified interaction modeling enables:

adaptive co-creative systems.

By analyzing interaction dynamics across time, AI systems can:

• model user preferences,

• anticipate creative needs,

• balance novelty and predictability,

• and adapt their collaborative behavior dynamically.

This creates the possibility for:

• mutual learning,

• co-evolution of human–AI partnerships,

• adaptive participation,

• and long-term collaborative growth.  

AI Drawing Partner

To demonstrate quantified co-creative AI in practice, the paper presents:

AI Drawing Partner

A quantified co-creative drawing system designed to:

• collaborate with users dynamically,

• model interaction patterns,

• visualize co-creative dynamics,

• and quantify collaboration during drawing sessions.  

The system captures:

• cognitive dynamics,

• interaction couplings,

• collaboration styles,

• drawing behaviors,

• and communication strategies

in real time during co-creative interaction.

The AI Drawing Partner serves as:

• a research platform,

• an experimental co-creative system,

• and an implementation example of quantified co-creative AI.

Quantified Co-Creative AI and Emergent Aesthetics

The framework also explores how quantified interaction data may contribute to understanding:

emergent aesthetics in artificial media.

By preserving interaction dynamics as a form of creative provenance, quantified co-creative systems may function as:

• retention mechanisms,

• interaction archives,

• and computational records of aesthetic emergence.

This creates opportunities for:

• preserving creative processes,

• analyzing stylistic evolution,

• studying collaborative novelty,

• and understanding how interaction dynamics shape artistic outcomes.

Research Significance

This work contributes toward a broader vision of AI systems designed not merely to generate artifacts, but to:

participate meaningfully within collaborative creative interaction.

Quantified co-creative AI systems reposition artificial intelligence from:

• isolated generation systems,

toward:

• interactive,

• adaptive,

• explainable,

• and participatory creative partners.

The long-term goal is to develop computational frameworks capable of:

• modeling co-creative interaction,

• sustaining meaningful collaboration,

• and understanding creativity as an emergent relational process unfolding between humans and intelligent systems.

Read the Full Paper

Unlocking the Black Box of Artificial Media with Quantified and Explainable Co-Creative AI Systems