Skip to content
Contact
☰
Color scheme: Dark.
About EIL

Lab

EIL is an independent experimental research lab for emergent intelligence, adaptive systems and Human–AI co-evolution.

Who stands behind EIL

Founder

Laura Serna Gaviria

Research & Intelligence Architecture

Laura Serna Gaviria founded EIL. She studied mechatronics with a focus on user experience and holds a Bachelor of Engineering. Her work centers on research architecture, intelligence design, multi-agent architecture, system reconstruction, hidden-state reasoning, behavioral/system analysis, causal and transition logic, evaluation architecture and experimental research design.

Research Perspective

Laura Serna Gaviria's research approaches intelligence, cognition and behavior through the processes that give rise to them. Her work focuses on the relationships between interaction, system structure, feedback, behavioral transformation and emergent capabilities.

Rather than treating observed outcomes as isolated properties, she investigates the conditions, dependencies and interaction histories through which those outcomes become possible. This perspective connects research on human cognition, emergence, Human–AI interaction, reconstruction and distributed intelligence.

RFI-IRFOS

Collaboration

RFI-IRFOS is the organization EIL works with. It contributes engineering and research infrastructure, and DINGIR, the World Model, was developed jointly.

Collaboration with RFI-IRFOS

EIL and RFI-IRFOS work together where research on emergence, reconstruction and distributed intelligence turns into concrete technical systems. Both institutions remain independent and bring different strengths.

Emergent Interaction Lab

EIL mainly contributes the research and interaction perspective:

  • Emergence in human cognition, human to AI interaction and complex systems
  • longitudinal human to AI research
  • Perspective Modeling & Cognitive Reconstruction
  • reconstruction and causal logic for the World Model (DINGIR)
  • research instruments, measurement systems and evaluation logic
  • multi-agent systems as distributed intelligence architectures
  • observation of drift, state development, memory and reintegration

RFI-IRFOS

RFI-IRFOS mainly contributes the technical and systemic development environment:

  • development of the World Model (DINGIR)
  • Ternary Intelligence Stack
  • model and systems engineering
  • agent and runtime architectures
  • data, audit and security structures
  • technical implementation, integration and testing of larger system networks
  • open science and research infrastructure

Together

The collaboration concentrates on systems where research logic and technical architecture interlock, in particular:

DINGIR, the World Model
Reconstruction of incomplete world states, relations, causal structures and uncertainties.
Distributed intelligence
Development and study of specialized agents, roles, handoffs and shared decision structures.
Research → system → research
Findings from the research are formalized and implemented technically. The resulting systems are then studied again as research objects.

EIL studies how such intelligence and reconstruction structures arise and change. RFI-IRFOS also drives their technical implementation and their embedding in larger system architectures.

Epistemic perspective

The Researcher's Perspective: How I Approach Complex Systems

Understanding how things become what they are

My research begins with a question that goes beyond describing what can be observed: What processes made this particular outcome possible?

Whether examining human behavior, cognitive transformation, complex interactions or artificial intelligence, I am interested in the conditions under which patterns emerge and relationships change. I also ask when systems develop capabilities or structures that were not previously present in the same form.

I approach these questions by examining relationships rather than isolated components, reconstructing possible developments from observable evidence, and identifying contradictions between assumptions, intentions, actions and outcomes.

A central concern is how interactions change the systems involved: how feedback modifies future behavior, how local actions produce broader consequences, and why similar initial conditions may lead to different developmental paths.

This perspective also shapes how I work. When existing analytical instruments cannot adequately represent a research problem, I develop experimental models, computational frameworks or intelligent systems to investigate it.

For me, understanding a system means more than describing its current state. It means investigating the processes through which that state became possible, the alternatives that may have existed and the conditions under which it may change again.

This text describes a scientific perspective and orientation. It is not empirical evidence of a personal trait, and it is not a finished or validated theory of science.

Seven dimensions of the way of working

  1. Generative Inquiry. How did this become possible? Describing a state and investigating its origin are two different tasks.
  2. Relational Reasoning. Which interactions and dependencies shape the observed system? The relations between people, components, environments, states and actions are examined.
  3. Causal Reconstruction. Which sequence of events and conditions could have produced this outcome? Temporal reconstruction that takes unobserved intermediate steps and alternative explanations into account.
  4. Adversarial Epistemology. What does not fit the current explanation? Inconsistencies, implicit assumptions, counter-hypotheses and gaps in explanation are examined.
  5. Dynamic Systems Thinking. How do systems transform through feedback and interaction? Feedback, path dependence, stabilization, destabilization and possible qualitative transitions are analyzed.
  6. Perspective Reconstruction. How does the observed situation change when examined from another position or explanatory model? Different actor perspectives are reconstructed without presenting inferred thoughts or intentions as certainties.
  7. Experimental System Building. What must be built to investigate what existing tools cannot adequately explain? Models, measurement procedures, agent systems and experimental environments follow from scientific questions.

Four recurring guiding questions

  1. Why do similar initial conditions lead to different outcomes? Connects emergence, nonlinearity, context dependence, path dependence and developmental trajectories.
  2. Why do intentions and actual outcomes diverge? Connects human behavior, social dynamics, agent interaction, coordination and systemic feedback.
  3. How do interactions transform the entities involved? Connects human cognition, relationships, human-AI co-evolution, distributed intelligence and emergent system structures.
  4. How can an invisible process be reconstructed from its observable consequences? Connects Behavioral Intelligence, World Models, longitudinal analysis, perspective reconstruction, evidence assessment and causal hypotheses.

These four questions are not four new programs. They describe the perspective from which the research questions of the spaces and programs are developed.

A research environment for intelligence

The Emergent Interaction Lab (EIL) is an independent experimental research environment. It investigates how intelligence emerges, changes, organizes itself and becomes capable of action within complex interacting systems.

The lab studies general processes of emergence and also the specific formation of intelligent and cognitive capabilities. The two are connected, and neither is treated as a special case of the other.

Its subject reaches beyond artificial intelligence. EIL studies intelligence across the boundaries between

  • humans
  • artificial intelligence
  • cognitive processes
  • technical systems
  • autonomous agents
  • groups of agents
  • information environments
  • adaptive interfaces
  • memory and knowledge systems
  • larger socio-technical structures
Central object of research

The central object of research is intelligence as a dynamic system phenomenon. A particular model, tool or AI architecture is one case within it. EIL asks what happens when perception, memory, reasoning, interaction, adaptation, feedback and action are connected over time.

Boundaries: what appears in the lab and what stays separate

EIL focuses on research into intelligence, interaction, cognition, emergence, adaptation, behavioral systems and Human–AI systems.

What appears inside the lab

Many things appear inside EIL: chatbot experiments, single frameworks, collections of agents, the Observatory dashboard, single Human–AI studies, world-model work and software products. Neurotechnology is a planned direction and not yet implemented. Each has its place in the lab. None of them defines EIL on its own. The laboratory analyzes existing intelligent systems and also builds experimental systems that make intelligence observable.

AI consultancy work belongs to the applied field and has its place on the Applied Research page. It complements the research of the lab.

What stays separate (5)

Concept

  • EIL is the overarching interdisciplinary research environment. Frameworks, metrics, systems and agent runtimes are parts of it.
  • The entries in the phenomena register form a registry and an ontology. They are candidates. Evidenced findings do not exist yet.
  • A recurring pattern counts as emergence only once evidence, recurrence, context diversity, functional consequence, counterevidence and provenance are in place.
  • A publication discloses no implementation details that are meant to stay protected.
  • Private psychological, biographical or health-related raw data stays unpublished even though it exists in the corpus.

The interaction space develops structure

AI research often looks at the model, HCI at the interface, cognitive science at the human, software engineering at the technical system. EIL looks at the interaction space between them. That space can develop structure of its own.

Interaction is more than an exchange of properties the participants already had. It can bring forth relations, roles, structures and possibilities of behavior that none of the participants held before.

Interaction is part of the mechanism through which intelligence develops.

Phenomena: what repeated interaction can produce

  • behavioral adaptation
  • persistent strategies
  • role differentiation
  • coordination patterns
  • shared representations
  • semantic conventions
  • feedback loops
  • memory structures
  • state transitions
  • emergent workflows
  • new problem-solving strategies
  • changes in the human participant

Observation first. Interpretation second. Claim last.

Every statement in the lab passes through this order. Emergence (novel structures, patterns and capabilities that arise through repeated interaction) counts as a finding once longitudinal evidence exists, that is, evidence from longer periods.

Research and system building are linked in both directions. Observations lead to questions, questions to models and instruments, and the instruments change what can be observed. Systems built this way are partly a result, partly an instrument and partly a subject of the research itself.

  1. Systems before isolated outputs

  2. Longitudinal before snapshot

  3. Interaction is part of the system

  4. Evidence before narrative

  5. Provenance before confidence

  6. Unknown is a valid state

  7. Humans remain part of the system

  8. Intelligence can be distributed

Working principles: how EIL works with AI systems over time

  • Systems before isolated outputs

    Individual model responses matter less than the structures that produce behavior over time.

  • Longitudinal before snapshot

    Intelligence changes. Research methods must be able to observe that change.

  • Interaction is part of the system

    The interface between components shapes the outcome. Interaction can create new structure.

  • Evidence before narrative

    Interesting interpretations wait for evidence.

  • Provenance before confidence

    A confident answer without traceable evidence is weak research material.

  • Unknown is a valid state

    The system should be able to represent uncertainty and leave the gap open.

  • Humans remain part of the system

    Human cognition, interpretation and decision-making are part of what Human–AI research studies.

  • Intelligence can be distributed

    Capabilities can live in networks of humans, agents, tools and memory systems.

  • Intelligence Design

    Designing how an intelligent system decomposes a problem, maintains relevant state, generates and rejects hypotheses, handles uncertainty, recognizes contradiction and updates its next actions.

  • Multi-Agent Intelligence

    Designing specialization, handoffs, coordination and shared research state across larger agent structures.

  • Research Continuity

    Preserving constraints, evidence, decisions and prior correct states across long-running research processes.

Research hypothesisExploratory

Intelligence should not be studied exclusively as a property of individual minds or individual models. Some forms of intelligence may only become visible when cognition, memory, interaction, tools, agents and environments are studied as a coupled system evolving over time. EIL exists to investigate that possibility.

How EIL describes intelligence as a system

Phenomenon, theory, reconstruction, measurement and experimental system belong to one chain. A phenomenon raises a theoretical question. Reconstruction asks how it could have arisen. Measurement decides what counts as evidence, and the experimental system makes the next observation possible.

The research loop

Human, AI and infrastructure connect in six stations that lead back to the first.

The research loop: how it all connects

The loop explains why methods, systems, agents, memory, longitudinal measurement, the Observatory, SELFGRAPH and the Kernel exist: each of them is a step or a measuring instrument of this loop.

The EIL research loopSix stations in a cycle that leads back to the first: 1. Human, 2. Interaction, 3. Emergent Patterns, 4. Externalization, 5. Artificial System, 6. ReintegrationResearchloop01Human02Interaction03Emergent Patterns04Externalization05Artificial System06Reintegration
  1. Human
  2. Interaction
  3. Emergent Patterns
  4. Externalization
  5. Artificial System
  6. Reintegration

The same loop is drawn here with its two carriers, the human and the artificial system.

Human and AI are in persistent interaction. Patterns that arise are held externally and continue inside artificial systems. The infrastructure carries this loop by preserving state, memory and evidence.

The six stations in detail

  1. Human

    behavior / cognition / state

    A person works with a language model for years. Behavior, decisions and corrections are observable.

  2. Interaction

    persistent coupling

    Human and AI change each other. Roles, meanings and expectations develop across sessions.

  3. Emergent Patterns

    recurrence / stabilization

    Some patterns recur, also in other contexts, and stabilize.

  4. Externalization

    method / representation / skill / agent / model

    A stable pattern is held externally: as a named method, a model, a memory structure or an agent role.

  5. Artificial System

    memory / reasoning / orchestration

    The externalized pattern continues inside an artificial system, with state, memory and coordination.

  6. Reintegration

    future human–AI interaction

    The system feeds back into the next interaction. The human meets a changed AI and has become someone different.

The order of the research, from EIL to the outputs

  1. EIL
  2. Research spaces
  3. Research programs
  4. Research questions
  5. Methods / metrics / data
  6. Experimental systems
  7. Infrastructure
  8. Evidence
  9. Publications / research notes / applied research

Intelligence as a dynamic system

Ten connected processes, from perception to adaptation, describe intelligence as a dynamic system.

Intelligence as a dynamic system: ten connected processes

EIL approaches intelligence as a combination of interconnected processes. A useful decomposition follows the chain below. These processes can exist in humans, AI models, agents, teams, interfaces, organizations or hybrid systems.

  1. 01Perception
  2. 02Interpretation
  3. 03Internal State
  4. 04Memory
  5. 05Reasoning
  6. 06Prediction
  7. 07Decision
  8. 08Action
  9. 09Feedback
  10. 10Adaptation

The guiding questions (10)

  1. What information is available to the system?
  2. What does the system believe about its environment?
  3. What does it remember?
  4. What state is it currently in?
  5. How does previous interaction influence current reasoning?
  6. How does uncertainty propagate through the system?
  7. How are decisions produced?
  8. How does feedback modify future behavior?
  9. When does local behavior become persistent system structure?
  10. When does coordination between components produce capabilities that are absent from individual components?

This system-level perspective is the foundation of EIL.

Coupling stack

Intelligent systems develop through the coupling of seven layers, from the human to the environment. Each layer influences the others.

Coupling stack: seven layers that influence each other

EIL is increasingly converging on a broader research program, Emergent Intelligence Systems. The program investigates how intelligent systems develop through the coupling of the layers below. Each layer influences the others, so the resulting system often cannot be understood by analyzing one component in isolation.

  1. Human
  2. Interaction
  3. AI / Agents
  4. Memory
  5. Knowledge
  6. World Model
  7. Environment

Research position

The lab works at an intersection of three areas. This shows where EIL sits and where the research is heading.

Research position: three areas at one intersection

The intersection the lab works at

That intersection is the research space of the Emergent Interaction Lab.

Three areas at one intersection: where the research is heading

The long-term direction of EIL connects three areas that are usually separated. It is a research direction that has not been reached yet.

  • Intelligence Science

    Understanding how intelligent behavior emerges and changes.

  • Intelligence Engineering

    Building systems capable of memory, reasoning, adaptation and coordination.

  • Human Intelligence Interaction

    Understanding what happens when human cognition becomes tightly coupled to artificial cognitive systems.

The future research frontier lies at their intersection.

Development line

Eight steps lead from interaction to the neuroadaptive direction, each with its status.

Development line: from interaction to the neuroadaptive direction

EIL grew outward from Human–AI interaction and keeps it as a foundation. The last steps are research direction and not implemented today.

  1. 2023 to 2025: natural longitudinal human–AI interactionActive research
  2. Reconstruction: which changes and emergences are visible across the course?Exploratory
  3. Formalization: instruments and typologies such as CEI and UIPActive research
  4. Externalization: cognitive operations become agents and subsystemsExploratory
  5. SELFGRAPH: changes and relations become machine-readablePrototype
  6. Workbench: a persistent cognitive working environmentConcept
  7. Adaptive AI: the system responds to recognized human statesPlanned
  8. Neuroadaptive AI: additional physiological signals close the loopPlanned

filled dot: in progress or implementedopen dot and dashed card: planned or concept

From interaction research to intelligence research: twelve stations that build on each other.

  • Human–Machine Interaction
  • Human–AI Interaction
  • Longitudinal Human–AI Interaction
  • Emergent Interaction Patterns
  • Adaptive Human–AI Systems
  • Agent and Multi-Agent Systems
  • Distributed Intelligence
  • Memory / State / Knowledge Systems
  • World Models
  • Cognitive State Modeling
  • Neuroadaptive Human–AI Systems
  • General Research on Emergent Intelligence

The last stations are research direction and not implemented today.

Why EIL builds its own systems

From cognitive practice to the experimental system: why EIL builds its own systems

Why EIL builds its own systems

Longitudinal and distributed research needed tools that did not exist off the shelf: something that keeps trajectories, evidences patterns and makes reconstructions testable. The systems at EIL came out of that need and trace back to recurring practice that was made visible, formalized and externalized as a system, and that acts back on the interaction. They serve the research. The path from practice to system, its five steps and their limits are described under Systems.