Research, Theory, and Consulting for Human-AI Collaboration
Enactive AI is a research and prototype ecosystem for AI systems that participate in unfolding human activity—responding, adapting, regulating, and creating through interaction rather than merely generating outputs.
Most AI systems are evaluated by what they produce. Co-creative systems must also be understood through how they participate: when they respond, how they transform direction, whether they support coherence, and what emerges across time.
Dynamic collaboration is not a single prompt followed by an answer. It is an evolving relation among people, agents, actions, artifacts, environments, and histories. The central question becomes not simply what the AI made, but how the human–AI system sustained viable participation.
Each prototype approaches dynamic human–AI interaction from a different level: direct collaboration, enactive participation, trajectory measurement, motif discovery, lightweight online adaptation, and regulation under drift.
A quantified co-creative drawing agent that analyzes user marks, feedback, and interaction history to contribute on a shared canvas while modeling Creative Sense-Making.
Launch AI Drawing Partner →An interactive drawing AI designed as a participant rather than a generator—tracking emerging direction, sensing shifts in coherence, adapting participation, and producing session-level co-creative trajectory reports.
Explore Aether →A digital drawing environment that transforms raw interaction traces into states, trajectories, properties, events, chapters, interpretive outputs, CSV data, and structured PDF reports.
Launch the Laboratory →An online-regulating architecture for systems that must learn and act while environments change—detecting drift, organizing regimes, adapting across timescales, and sustaining coherent behavior.
Explore the architecture →A compact, browser-based proof that online adaptation does not require massive deep-learning infrastructure—showing continuous forecasting, skill tracking, plasticity, and drift response in a small interactive system.
Launch Tiny Emergence Machine →An expert-guided environment for discovering temporal motifs and reusable behavioral structures, supporting collaborative temporal science through human annotation, evidence review, and motif-centered interpretation.
Explore Kalyriel Scope →Co-creative intelligence unfolds as a trajectory. Human and AI participants continuously shape the possibilities available to one another, producing patterns that cannot be reduced to either side alone.
Marks, timing, motion, pauses, responses, and contribution sequences.
Exploration, coherence, drift, regulation, initiative, and stability.
Ordered histories of changing interaction states through time.
Attractors, transitions, returns, bifurcations, and emergent chains.
Chapters, reports, signatures, and research-ready process evidence.
The research presented across Enactive AI, Cognitive Trajectory Modeling, Enactive Drift Regulation, the Emergence Machine, Aether, and related projects is guided by a set of recurring principles. These principles are not final truths, but working commitments that organize inquiry into adaptive intelligence, human–AI interaction, and sense-making systems.
Adaptive systems are best understood not as repositories of information, but as patterns of organized activity unfolding through time.
From this perspective, intelligence is not something a system possesses. It is something a system does. What matters is not the accumulation of information alone, but how activity is coordinated, sustained, and reorganized in response to changing conditions.
Together, these principles motivate a shift from static representations toward dynamic organization, from optimization toward regulation, and from isolated cognition toward interaction-centered accounts of intelligence. They provide a common foundation for Enactive AI systems, cognitive trajectory models, adaptive forecasting architectures, and human–AI co-creative technologies.
A self-contained adaptive forecasting prototype implemented in approximately 300 lines of integrated HTML, CSS, and JavaScript. It runs entirely in the browser, requires no server connection, and combines online prediction, drift-sensitive adaptation, comparative skill measurement, and multi-scale attractor visualization in a single lightweight system that turns prediction into a visible process of regulation.
Upload a CSV or text signal—or use the generated stream—and watch the Emergence Machine reorganize as conditions change. Click Begin process to start.
Rolling forecast skill relative to a persistence baseline. Zero means equal error; positive values favor the adaptive model, while negative values favor repeating the previous observation. As a practical guide, 0–10 indicates only a slight advantage, 10–20 a modest advantage, 20–30 good skill, 30–40 strong skill, and values above 40 very strong skill. These ranges are interpretive rather than universal and should be considered alongside the signal, window length, and evaluation conditions.
How strongly the model is currently revising itself in response to drift and surprise.
How strongly the signal’s recent direction and pace are departing from its established local pattern.
This working prototype reflects the kind of research-driven system design, interaction modeling, and adaptive AI development available through Enactive AI consulting.
Explore consulting →Collaborative Temporal Science is a high-level paradigm for using AI to help scientific communities discover, interpret, and refine knowledge about dynamic systems as they unfold through time.
Most AI systems treat temporal data as a forecasting, classification, or optimization problem. Collaborative Temporal Science begins from a different premise: scientific understanding is built by discovering recurring phenomena, naming them, comparing interpretations, accumulating evidence, and developing shared vocabularies over time.
In this approach, AI contributes by organizing time-series data into inspectable temporal structures: motifs, transitions, regimes, and chapters. Human experts contribute the domain knowledge needed to determine what those structures mean by labeling patterns, adding notes, and refining interpretations. As these expert annotations accumulate, Kalyriel Scope uses them to enrich subsequent analysis, allowing newly discovered structures to be compared against an evolving motif library. The result is not automated expertise, but a higher-fidelity representation of the data shaped through collaboration between computational discovery and human interpretation.
High-drift variability followed by progressive stabilization. Often appears after a disruptive transition and may indicate recovery toward a more coherent regime.
This motif is consistently associated with a transition from unstable temporal organization into a more coherent pattern, making it useful for studying recovery, adaptation, and regime stabilization across domains.
Segmentation, regime detection, motif discovery, chapter clustering, and drift-sensitive adaptation are computed locally through lightweight mathematical operations rather than large remote classification machinery.
The same computed structures can be read statistically or semantically. Experts can name motifs, annotate examples, build motif libraries, and create readable reports that reflect human interpretation.
Motif libraries preserve stable computational identities while allowing meanings to evolve, making it possible for communities to share, revise, and expand domain-specific temporal vocabularies.
The interaction quality is the contribution: the AI continuously organizes change into candidate structures, while humans decide what matters. The result is a platform for collaborative knowledge co-construction across domains such as healthcare, neuroscience, climate science, industrial monitoring, finance, creative process research, and human-AI interaction.
Enactive AI translates interaction-centered theory into research instruments, adaptive architectures, and consulting frameworks for domains where process, regulation, and change through time matter.
Art therapy depends heavily on process, yet much of that process disappears once the final artwork is complete. Enactive AI develops research instruments that preserve the temporal organization of art-making—including exploration, hesitation, persistence, reorganization, regulation, and stabilization. AI Drawing Partner, Aether, and the Cognitive Trajectory Laboratory provide complementary environments for studying how creative engagement unfolds and changes through time.
Studies reciprocal human–AI contribution, feedback, uptake, and Creative Sense-Making during shared art-making.
Launch the platform →Investigates how a regulatory creative AI can sense changes in interaction, adjust its participation, and support coherent co-creative trajectories.
Explore Aether →Transforms drawing activity into measurable process traces, cognitive trajectories, events, chapters, therapeutic process indicators, data exports, and session reports.
Launch the Laboratory →This work supports research instrument design, longitudinal process measurement, clinical research partnerships, validation studies, therapeutic technology development, and responsible AI integration in creative and healthcare settings.
Many AI systems are trained under relatively stable assumptions but deployed in environments that continually change. The Emergence Machine is an experimental architecture for systems that must detect drift, reorganize behavior, learn across timescales, and preserve coherence without relying on a fixed operating regime.
Agents that revise strategies as tasks, tools, users, and operating conditions change.
Systems that regulate initiative, participation, and coordination over extended interactions.
Identifying changing regimes, emerging patterns, and breakdowns in distributed work.
Maintaining viable behavior under environmental drift, uncertainty, and shifting constraints.
Adapting to users without reducing personalization to static preference profiles.
Detecting structural change rather than treating every deviation as noise.
Enactive AI can help organizations apply these principles through adaptive-system strategy, architecture design, drift and regime analysis, simulation, prototype development, and evaluation frameworks for systems operating under continual change.
Not every adaptive problem requires a foundation model, an offline training pipeline, or continual access to remote inference infrastructure. Enactive AI explores systems that use computation proportionally—applying large-scale models where they are necessary while developing lightweight, locally adaptive alternatives where they are not.
The Tiny Emergence Machine illustrates this principle. It performs online forecasting, drift monitoring, and multi-scale input characterization directly in the browser, without a GPU cluster, persistent inference server, or repeated model API calls.
This does not make computation impact-free. The system still uses energy on the local device. Its value lies in demonstrating that useful adaptive intelligence can sometimes be achieved through smaller, task-specific architectures that improve portability, privacy, responsiveness, inspectability, and computational efficiency.
Reorganizes during use rather than relying on periodic retraining jobs.
Prediction and uploaded-data processing remain on the local device.
Computational scale is matched to the actual task.
The Tiny Emergence Machine could serve as a lightweight, online regulatory layer for modeling how neural activity changes during human–AI creative interaction. Rather than merely classifying EEG after a session, it could track the unfolding process in real time.
Can changes in neural dynamics be modeled alongside the evolving structure of human–AI co-creation?
This direction builds on the Co-Creative Sense-Making (CCSM) framework, which organizes the co-creative process across four interdependent domains: cognitive dynamics, interaction dynamics, collaboration dynamics, and domain behaviors. Together, these domains provide a structured account of how creative activity unfolds within and between human and AI participants.
Sense-making curves and other interaction traces exported from co-creative drawing systems could be analyzed by the Emergence Machine as evolving interaction and cognitive trajectories. Rather than treating the creative process as a sequence of isolated actions, this architecture would model how artistic participation develops through time and use those dynamics to regulate the AI’s behavior during creation.
Track strokes, pauses, timing, spatial movement, turn-taking, response latency, and patterns of engagement across the unfolding session.
Model exploration, consolidation, drift, reorganization, persistence, attractor formation, and transitions within the creative trajectory.
Estimate coupling, participation balance, initiative, responsiveness, influence, and the changing relationship between human and AI contributions.
Incorporate drawing-specific patterns such as motif recurrence, line extension, color use, compositional development, and medium-sensitive actions.
Sense-making curves provide a temporally organized representation of how creative participation shifts between exploratory, stabilizing, and action-oriented modes during co-creation. Introduced in the 2017 Creativity and Cognition paper Creative Sense-Making: Quantifying Interaction Dynamics in Co-Creation, the framework established a method for visualizing and quantifying the changing interaction dynamics of open-ended creative activity through time.
Move across the curve to inspect different forms of sense-making. Large fluctuations indicate relatively unclamped participation; activity near the centerline indicates tighter coupling with the immediate environment.
Broad, rapid shifts between internal ideation and embodied engagement as the human–AI system opens a field of possibilities.
Nicholas Davis, Chih-Pin Hsiao, Kunwar Yashraj Singh, Brenda Lin, and Brian Magerko · ACM Creativity & Cognition 2017
Machine-learning and time-series processes could operate directly on the evolving sense-making curve while a creative session is underway. Co-creative AI systems could use the detected trajectory to estimate the current phase of participation, recognize emerging transitions, and regulate when, how, and how strongly the AI contributes.
After a session, the complete curve could support statistical analysis, event classification, and comparison across people, systems, tasks, and creative conditions. This would allow researchers to study not only what was produced, but how the collaborative process developed through time.
Quantified Co-Creative AI treats the creative process itself as an analyzable and adaptive temporal phenomenon. By linking sense-making curves with interaction traces, cognitive trajectories, collaboration dynamics, and domain-specific behaviors, it provides a foundation for co-creative systems that can measure participation, explain their adaptations, and regulate their behavior in response to the unfolding creative process.
Explore a capability to see the organizational outcomes, engagement formats, systems, methods, and deliverables that Enactive AI can bring to a project.
Design human–AI systems around reciprocal participation, shared control, interpretability, and the evolving quality of interaction rather than output performance alone.
Strategic advisory and research scoping clarify the problem, stakeholders, operating context, and success criteria.
Enactive AI supports research programs that need more than abstract theory or isolated model evaluation. The work combines conceptual framing, technical probe construction, empirical study design, and process-sensitive analysis to investigate how people and intelligent systems coordinate, adapt, and make meaning through time.
Enactive AI can help transform an early research idea into a working, testable system. This includes conceptual prototyping, concept testing, interaction design, rapid prototype development, and technical probe construction for questions that are difficult to study through surveys or static demonstrations alone.
Enactive AI can design and conduct empirical investigations of human–AI interaction, co-creation, adaptive regulation, and cognitive trajectory dynamics. The focus is on studying unfolding processes: how people respond to systems, how systems participate in activity, and how coordination changes under ambiguity, drift, and creative exploration.
This research program has developed through an iterative process of building prototypes, observing interaction, and using those observations to construct new theories and models. Early creativity-support tools examined how computation could help individuals engage in creative work. Co-creative systems then shifted the focus toward creativity as an emergent process between human and computational partners.
Each prototype functioned as a technical probe, making interaction dynamics visible and generating empirical questions that existing theories could not fully explain. Across successive systems, the research moved from artistic computer colleagues and participatory sense-making to Creative Sense-Making, quantified co-creation, hybrid intelligence, and explainable co-creative AI. This trajectory ultimately led to Interaction-Centered Intelligence and Cognitive Trajectory Modeling, which understand creativity and intelligence as relational, temporal processes that emerge and reorganize through interaction.
Select a publication to view its abstract and online paper.
Proceedings of the 8th ACM Conference on Creativity and Cognition, pp. 185–194.
Proceedings of the 8th ACM Conference on Creativity and Cognition, pp. 207–216.
Proceedings of the 9th ACM Conference on Creativity & Cognition, pp. 13–22.
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 651–660.
Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, 9(6), pp. 9–12.
Proceedings of the International Conference on Computational Creativity, pp. 38–45.
Proceedings of the 2015 ACM SIGCHI Conference on Creativity and Cognition, pp. 275–284.
Proceedings of the 2015 ACM SIGCHI Conference on Creativity and Cognition, pp. 185–186.
Creativity in the Digital Age, Springer London, pp. 109–133.
Proceedings of the 21st International Conference on Intelligent User Interfaces, pp. 196–207.
ACM Transactions on Interactive Intelligent Systems, 7(4), pp. 1–25.
Proceedings of the 2017 ACM SIGCHI Conference on Creativity and Cognition, pp. 356–366.
Proceedings of the International Conference on Computational Creativity, pp. 136–145.
Proceedings of the 2025 Conference on Creativity and Cognition, pp. 560–572.
In Artificial Media: Emerging Trends in Narratives, Education and Creative Practice, Springer Nature Switzerland, pp. 21–48.
Handbook of Human-Centered Artificial Intelligence, pp. 833–889. Singapore: Springer Nature Singapore.
arXiv preprint arXiv:2606.00807v2 [cs.AI].
arXiv:2606.15358 [cs.HC].
arXiv:2607.03834v1 [cs.HC].
arXiv:2606.22057
Art therapy has demonstrated effectiveness across diverse clinical populations, and its theoretical traditions have generated valuable perspectives on symbolism, expression, narrative reconstruction, meaning-making, physiological responses, and neurobiological processes. While these approaches provide important accounts of therapeutic experience and change, they have placed comparatively less emphasis on how cognition, regulation, and interaction dynamics evolve during the creative process itself, making it difficult to analyze how creativity and therapeutic outcomes emerge through time. As a result, art therapy research continues to rely heavily on qualitative interpretation, outcome measures, and retrospective self-report, while the dynamics of therapeutic change remain difficult to quantify. This paper proposes an enactive, dynamical framework for understanding and measuring cognitive change in art therapy through the analysis of creative interaction dynamics over time. Within this framework, therapeutic change is hypothesized to be reflected in cognitive trajectories, temporally unfolding patterns of engagement that reveal shifts in stability, exploration, and adaptation. To operationalize this framework, the paper introduces the Cognitive Trajectory Laboratory (CTL), an instrumented drawing environment that transforms interaction traces into cognitive trajectories unfolding through time, enabling the identification of emergent properties, significant events, and overarching chapters of the creative process. By making the dynamics of creative engagement measurable, the proposed framework and accompanying laboratory provide new methodological tools for art therapy assessment and research while creating opportunities for longitudinal analysis of therapeutic change. Implications are discussed for process-oriented evaluation and computational modeling of creative engagement.
Read on arXivThis work builds on more than a decade of research into distributed creativity, artistic computer colleagues, participatory sense-making, Creative Sense-Making, quantified co-creation, the Five Pillars of Enaction, explainable co-creative AI, Interaction-Centered Intelligence, Cognitive Trajectory Modeling, and regulation under drift.
Enactive AI serves as a public home for experimental systems, foundational publications, research frameworks, and research-driven consulting in dynamic human–AI interaction. It brings together theory, empirical inquiry, and prototype development to support organizations designing, evaluating, and advancing human–AI collaboration.