Artificial intelligence has traditionally been understood through the lens of computation.
Within classical AI paradigms, intelligence is often modeled as symbolic manipulation, information processing, internal representation, optimization, and goal-directed reasoning. These approaches have produced powerful systems capable of classification, prediction, generation, and strategic planning. Yet they often struggle to fully explain one of the most fundamental aspects of human cognition: meaningful interaction.
Human intelligence does not emerge solely through detached computation. Instead, it emerges through embodied action, environmental engagement, social participation, improvisation, and continuous interaction with a changing world.
This insight lies at the heart of Enactive AI.
Enactive AI is an interaction-centered approach to artificial intelligence grounded in the cognitive science paradigm of enaction. Rather than treating intelligence as isolated internal processing alone, enactive AI proposes that cognition emerges dynamically through participation between agents and environments.
Within this framework, intelligence is not merely computed.
It is enacted.
The theory of enaction emerged through the work of Francisco Varela, Evan Thompson, Eleanor Rosch, and others seeking alternatives to classical representational models of cognition.
Enaction proposed that cognition should not be understood as passive internal modeling of an external world. Instead, organisms actively construct meaning through engagement, perception is inseparable from action, and cognition emerges through adaptive interaction with environments.
Meaning therefore does not exist fully formed “out there” waiting to be represented internally. Nor is it fully generated internally independent of the world.
Meaning emerges through sense-making.
Agents continuously regulate their relationship with environments through action, perception, and adaptation.
This interaction-centered view of cognition became deeply influential in later work on co-creative AI and human-AI collaboration.
One of the earliest formal articulations of enactive AI within computational creativity appeared in:
Building Artistic Computer Colleagues with an Enactive Model of Creativity, presented at the International Conference on Computational Creativity in 2014.
The paper proposed a major conceptual shift in how creative AI systems should be designed.
At the time, many computational creativity systems focused primarily on autonomous generation, artifact production, or symbolic creativity models. The 2014 work argued that these approaches were insufficient for modeling the full dynamics of human creativity because they neglected improvisation, interaction, environmental coupling, and participatory collaboration.
Instead, the paper proposed the enactive model of creativity.
This framework synthesized enactive cognition, ecological psychology, creative cognition, improvisation theory, and collaborative interaction into a new theoretical foundation for co-creative AI systems.
The goal was not merely to build creative generators.
The goal was to build artistic computer colleagues.
The phrase “artistic computer colleague” represented a profound shift in AI design philosophy.
Rather than treating AI as a passive tool, a generator, or an autonomous replacement, the framework proposed that AI systems could function as collaborators, improvisational partners, adaptive participants, and co-creative agents.
This idea anticipated many later developments in human-centered AI, co-creative systems, interactive machine learning, and participatory AI.
The enactive approach proposed that creative systems should participate dynamically within interaction loops rather than merely generating isolated outputs.
Creativity therefore became an emergent property of interaction.
The primary implementation of these ideas became:
The Drawing Apprentice was an improvisational co-creative drawing system designed to collaborate with human users in real time.
Unlike conventional drawing software or modern prompt-based image generators, the system actively participated during the creative process itself.
The Drawing Apprentice interpreted user sketches, generated responsive visual contributions, adapted dynamically to interaction context, and engaged in reciprocal turn-taking with human collaborators.
The system therefore functioned less like a static tool and more like an interactive creative participant.
This was one of the earliest practical demonstrations of enactive AI principles in action.
One of the most important concepts introduced in the enactive model of creativity was perceptual logic.
Perceptual logic describes the way cognitive systems selectively organize and attend to affordances within an environment. Rather than perceiving everything equally, agents dynamically foreground certain possibilities while suppressing others based on interaction context, intention, and ongoing activity.
The framework proposed multiple layers of perceptual organization: local perceptual logic, regional perceptual logic, and global perceptual logic.
Within the Drawing Apprentice, local perceptual logic analyzed immediate sketch features, regional perceptual logic identified relationships among nearby structures, and global perceptual logic evaluated the composition as a whole.
This allowed the system to participate dynamically within evolving artistic interaction rather than merely executing static generation routines.
The concept of perceptual logic later became foundational to broader theories involving creative trajectories, interaction dynamics, adaptive cognition, and perceptual attunement.
A core insight of enactive AI is that creativity and intelligence are fundamentally improvisational.
In the enactive framework, agents do not simply execute fixed plans, cognition unfolds dynamically through interaction, and meaning emerges recursively through participation.
The 2014 paper explicitly framed creativity as an enactive process that emerges through constant interaction with the environment and other agents within it.
This idea became central to later developments in participatory sense-making, creative sense-making, co-creative AI, and quantified co-creation.
Rather than treating intelligence as detached symbolic reasoning alone, enactive AI emphasized adaptive interaction, environmental coupling, improvisation, and emergence.
The enactive approach also challenged dominant assumptions within traditional AI research.
Classical information-processing models typically assume cognition occurs internally, representations mirror external reality, and intelligent behavior emerges through computation over symbols.
Enactive AI instead proposes that cognition emerges through action, perception is active, and intelligence arises through participation within environments.
This distinction becomes especially important in collaborative systems.
In co-creative interaction, humans adapt to AI systems, AI systems adapt to humans, interaction histories shape future possibilities, and meaning evolves continuously through participation.
These dynamics are difficult to fully explain through static representational models alone.
Enactive AI provides a framework capable of modeling these relational processes.
Many recent developments in AI increasingly move toward interaction-centered paradigms, including conversational agents, adaptive copilots, collaborative design systems, interactive machine learning, and hybrid intelligence systems.
In many ways, enactive AI anticipated this broader transition.
The early work on artistic computer colleagues explored questions that are now becoming increasingly central:
How should AI systems collaborate with humans? How do interaction dynamics shape intelligence? How does meaning emerge through participation? What kinds of AI systems support creative improvisation?
These questions continue shaping modern human-centered AI research.
Enactive AI ultimately proposes a fundamentally different vision of artificial intelligence.
Rather than building systems that merely predict, optimize, classify, or generate, the enactive paradigm explores systems capable of participating, adapting, improvising, collaborating, and co-constructing meaning dynamically with humans.
This represents more than a technical shift.
It represents a shift in how intelligence itself is understood.
From an enactive perspective, intelligence is relational, cognition is participatory, and creativity emerges through interaction between coupled systems engaged in shared processes of sense-making.