Interaction as the Primary Unit of Analysis in Co-Creative AI Systems

-Nicholas Davis, PhD

From Isolated Intelligence to Interaction-Centered Intelligence

Traditional artificial intelligence has largely inherited its assumptions from classical information-processing models of cognition. Across much of AI research, the primary unit of analysis has historically been the isolated agent: a bounded system that receives inputs, performs internal computation, and produces outputs. Intelligence within this paradigm is typically evaluated through accuracy, optimization, prediction, planning, benchmark performance, or autonomous generation.

While this paradigm has produced major advances in machine learning and generative AI, it often treats interaction as secondary — merely a channel through which information passes between otherwise separate entities. Human users become external operators, prompts become static inputs, and collaboration becomes reduced to a sequence of requests and responses. Intelligence is understood primarily as internal computation occurring inside isolated systems.

Co-creative AI research proposes a fundamentally different perspective.

Rather than treating intelligence as something contained entirely within humans or machines, interaction-centered approaches argue that cognition, creativity, meaning, and adaptive behavior emerge dynamically through interaction itself. From this perspective, the interaction between participants becomes the primary phenomenon requiring explanation.

This shift represents a major conceptual transition: from isolated intelligence to relational intelligence, from static outputs to evolving interaction dynamics, and from autonomous generation to participatory sense-making.

The infographic below situates this interaction-centered perspective within a broader historical progression of cognitive and AI frameworks, illustrating how different paradigms define their primary unit of analysis and what each framework contributes toward understanding co-creative AI systems.

Interaction as the Primary Unit of Analysis

The interaction-centered perspective proposed in co-creative AI systems synthesizes many insights from: distributed cognition, embodied cognition, enaction, participatory sense-making, computational creativity, and human-computer interaction.

However, it advances a more explicit claim: interaction itself is the fundamental unit in which intelligence and creativity emerge.

This shift has major implications. Rather than asking:

“What did the AI generate?”

interaction-centered approaches ask:

What emerged during interaction?
How did participants adapt to one another?
How did meaning evolve through collaboration?
How did interaction dynamics shape outcomes?
How was coherence sustained or disrupted through time?

Within this paradigm: intelligence is participatory, creativity is co-constructed, and cognition unfolds relationally. The interaction becomes: the site of emergence, the locus of adaptation, and the primary object of analysis.

Why This Matters for the Future of AI

As generative AI systems become increasingly powerful, many contemporary systems risk reducing humans to: prompt engineers, selectors, or evaluators of machine-generated outputs. Interaction-centered co-creative AI proposes a different future. Rather than replacing human participation, co-creative systems aim to: sustain collaborative engagement, support participatory sense-making, maintain adaptive coordination, and enable shared creative emergence between humans and AI systems.

This perspective increasingly suggests that the future of AI may not be defined solely by autonomous intelligence, but by: the quality of interaction intelligent systems can sustain with humans over time.

From this perspective, intelligence is not merely what a system computes internally.

Intelligence emerges through interaction.

Information Processing and Traditional AI

Classical information-processing approaches treat cognition as symbolic computation occurring inside an individual mind or computational system. Within traditional AI, the primary unit of analysis is typically: the internal model, the algorithm, or the isolated intelligent agent.

These approaches excel at: optimization, classification, prediction, planning, and autonomous generation. However, interaction itself is often under-theorized. Human-AI collaboration becomes reduced to:

input → computation → output

The strength of this perspective lies in its tractability and formal precision. Yet it frequently struggles to account for: situated interaction, collaborative emergence, improvisation, participatory creativity, and the evolving dynamics of human-AI relationships through time.

In co-creative systems, many important phenomena cannot be explained solely through isolated internal computation because meaning emerges dynamically during collaboration itself.

Distributed Cognition

Distributed cognition expanded the unit of analysis beyond isolated minds by proposing that cognition distributes across: people, tools, artifacts, workflows, and environments. This represented a major conceptual advance because cognition was no longer viewed as residing solely inside individual agents. Instead, intelligent activity emerged through coordinated systems involving humans, technologies, representations, and environmental structures.

Within distributed cognition: tools become part of cognition, environments scaffold reasoning, and collaboration becomes central to understanding complex activity. This framework strongly influenced early work in creativity support systems and collaborative creativity research.

However, distributed cognition often emphasizes the structural distribution of cognition across systems more than the evolving temporal dynamics of interaction itself. It explains where cognition occurs across socio-technical systems, but less frequently models: moment-to-moment coordination, improvisational adaptation, interaction rhythms, or the unfolding emergence of meaning during collaboration.

Embodied Cognition

Embodied cognition shifted attention toward the role of the body in shaping cognition. Rather than treating intelligence as abstract symbolic reasoning detached from physical experience, embodied approaches argue that cognition depends fundamentally upon: perception, action, sensorimotor coupling, and bodily interaction with the world.

Embodied cognition demonstrated that: how we move, what we can perceive, and how we physically engage environments deeply shape cognition and meaning-making. This framework helped challenge purely representational models of intelligence and influenced later work in robotics, human-computer interaction, and enactive cognition.

However, embodied cognition often remains focused primarily on the embodied individual rather than the collaborative interaction dynamics between multiple participants. While embodiment is essential for understanding situated cognition, it does not fully explain how shared meaning emerges relationally during co-creative interaction.

Enaction

Enactive cognition further expanded these ideas by proposing that cognition emerges through ongoing interaction between agents and environments. Rather than representing a pre-given world internally, agents enact meaningful worlds through continuous perception-action coupling.

Within enaction: cognition is fundamentally relational, meaning emerges through interaction, and intelligence unfolds dynamically through adaptive engagement with the environment. Enaction introduced several concepts highly relevant to co-creative AI: autonomy, sense-making, embodiment, emergence, and participatory interaction.

Importantly, enaction shifted attention toward dynamic coupling between agents and environments. Creativity, cognition, and meaning were increasingly understood as emergent properties of interaction rather than isolated internal processes.

However, while enaction strongly emphasizes interaction philosophically, many enactive frameworks remain qualitative and less operationalized computationally. They often describe interaction conceptually without fully modeling or quantifying the evolving dynamics of collaboration through time.

Participatory Sense-Making

Participatory sense-making extended enaction into social interaction by arguing that meaning can emerge between participants through coordination itself.

This represented a major shift.

Meaning was no longer treated solely as something generated internally and exchanged externally. Instead: interaction itself becomes constitutive of cognition, coordination dynamics become meaningful, and collaborative interaction develops partially autonomous structures over time.

Participatory sense-making became highly influential for co-creative AI research because it directly addressed: collaboration, turn-taking, improvisation, mutual influence, and relational emergence. This framework helped establish interaction as a central phenomenon in human-AI collaboration research. Yet participatory sense-making often remained difficult to operationalize computationally. Many of its concepts were theoretically rich but lacked concrete frameworks for modeling interaction dynamics quantitatively during co-creative activity.

Creative Sense-Making and Quantified Co-Creation

Creative Sense-Making (CSM) emerged as an attempt to operationalize interaction-centered cognition computationally. Rather than evaluating only final artifacts, Creative Sense-Making focused on: interaction dynamics, activity traces, creative trajectories, sense-making curves, and collaborative emergence unfolding through time.

Within this framework, the primary unit of analysis becomes: interaction itself.

Creativity is no longer viewed as: located solely inside humans, produced autonomously by AI, or reducible to isolated outputs. Instead, creativity emerges dynamically through: reciprocal influence, adaptive coordination, participatory interaction, and evolving collaborative trajectories. This interaction-centered perspective allows researchers to study: how collaboration evolves, how meaning emerges, how interaction rhythms stabilize or destabilize, and how co-creative systems sustain participatory engagement over time.

Importantly, interaction leaves observable traces that can be modeled computationally: timing, turn-taking, interruptions, conceptual divergence, adaptation, coordination patterns, and collaborative trajectories.

This makes interaction not only philosophically important, but empirically measurable.

Page updated

Google Sites

Report abuse