The Active Inference Institute is a volunteer-led 501(c)(3) nonprofit dedicated to improving the accessibility, rigor, and applicability of Active Inference. We welcome people of all backgrounds, time zones, and familiarity levels.
Active Inference is the central scientific and practical framework around which the Institute organizes education, research, applications, and ecosystem support.
The Active Inference Institute is organized around a governing Board of Directors, executive Officers, a Scientific Advisory Board, and two operational units for education and research.
Activities are the Institute's live participation surface: learning groups, model streams, project meetings, public media, and recurring updates.
The Institute's history runs from a 2020 co-founder team meeting around a shared interest in Active Inference, through the Active Inference Lab, to the nonprofit Active Inference Institute it is today. This page traces that story year by year. For the current and upcoming years, see the 2025 and 2026 annual overviews.
InstituteOS keeps operational knowledge in the private docs and library trees, then exports only public-safe slices into this static website. This page makes the interface explicit: authored public copy, sanitized public tables, exported graphs, public project pages, and the checks that block private material before publication.
Institute Programs turn interest into participation through volunteer, internship, mentorship, fellowship, partnership, philanthropy, grants, and open-source pathways. All programs are open to people of any background and familiarity level.
Projects convert shared ideas into repositories, research outputs, learning infrastructure, media, events, and applied work.
Learning resources point newcomers and experienced participants toward living references, livestreams, Active Inference Insights, textbook groups, readings, code implementations, and domain-specific courses.
The Active Inference ecosystem includes challenge areas, user segments, information architecture, organizations, projects, and domains of application.
All backgrounds, time zones, and familiarity levels are welcome. Start with public channels, attend activities, and choose a contribution pathway that fits your context.
Volunteering is how most people begin participating in the Institute. There is no expertise threshold — contributions at every level of familiarity with Active Inference are meaningful and welcome.
An internship is a structured contribution pathway — work on real Institute projects, develop skills, and build relationships with researchers and educators in the Active Inference field.
Mentorship at the Institute runs in both directions — experienced contributors support the development of newer participants, and everyone benefits from the exchange of perspectives, context, and guidance.
Fellows develop research, applications, or public resources aligned with the Institute and Ecosystem, provide regular progress reports, and produce outputs that advance the field and remain publicly accessible.
The Institute supports grant proposals and collaborative funding efforts that advance Active Inference research, education, and application. Participants and projects can seek Institute support for proposal preparation, institutional affiliation, and co-investigation.
Partnerships align organizations and collaborators around concrete shared work — co-hosting events, co-authoring research, sharing infrastructure, or developing joint education and application programs.
The Active Inference Institute is a registered 501(c)(3) nonprofit. Donations from individuals and organizations directly fund public education, research infrastructure, community events, and ecosystem stewardship.
EduActive is the Institute's education arm. It hosts the Textbook Group, courses, ontology work, the Active Inference Journal, audio-visual production, seasonal school, and related learning initiatives. EduActive projects connect Active Inference to learning infrastructure, pedagogy, and public educational resources.
ReInference is the Institute's research and development arm. It hosts software projects, theoretical work, applied modeling, geospatial analysis, knowledge engineering, and interdisciplinary research under the Active Inference framework. ReInference projects range from widely-used open-source tools to specialized research initiatives.
Active Blockference develops Active Inference examples and tools applicable to decentralized and blockchain-adjacent systems. The project has produced a GitHub repository, blog posts, and video overviews, and serves as an integration point for multiagent modeling work from Active InferAnts.
AICACP is an Institute project organized around AI capabilities, alignment, and care practices — exploring how Active Inference principles inform responsible and care-oriented AI development. The project maintains a public information page and contributes to the Institute's research on beneficial AI.
The Applied Active Inference Symposium is a yearly event that convenes the global Active Inference community. It has run five times from 2021 through 2025, with the sixth symposium in 2026. The Symposium features research presentations, discussions, and collaborative sessions spanning computational neuroscience, AI, ecology, economics, health, and other domains.
Cognitive Agent Modeling is a ReInference Institute project focused on developing minimal cognitive agent models grounded in Active Inference. The work explores how perception, action, and learning can be formalized and implemented using the Active Inference framework.
FarmWorks develops miniature Active Inference models and applications for agricultural and ecological contexts. The project has produced a public FarmWorks page and a 2024 publication describing the approach of treating farm and soil systems as active inference agents.
Generalized Notation Notation (GNN) is a text-based notation project for communicating and specifying generative models. It provides a shared language for describing Active Inference and related models in a way that is both human-readable and machine-processable, enabling interoperability across tools and codebases.
GEO-INFER develops geospatial modeling methods grounded in Active Inference. The project integrates spatial data analysis with the Active Inference framework, opening applications in ecology, urban planning, resource management, and other domains where geographic context matters.
The Graphical Interface project develops visual and interactive tools for working with Active Inference models. It focuses on making model structure, dynamics, and outputs more accessible through well-designed graphical interfaces and visualization layers.
Knowledge Engineering develops and maintains the public knowledge infrastructure for the Active Inference Institute, including the public frontend, literature meta-analysis, and organizational knowledge systems. The project connects Institute outputs to the broader literature and makes them machine-readable and navigable.
Active InferAnts is a multiagent modeling project that applies Active Inference to collective behavior, inspired by ant colony dynamics. It has produced a GitHub repository, a 2021 paper, and code developed within the Active Blockference project, with multiple realizations across different modeling contexts.
The RxInfer.jl Learning and Development Group supports learning and extending RxInfer.jl, a Julia package for reactive Bayesian inference. The group produces learning resources, code examples, and contributes to the broader development of RxInfer as a platform for Active Inference.
The Theoretical Neurobiology Group (TNB) supports theoretical work connecting Active Inference to neurobiology. It hosts regular meetings and discussions for participants working on mathematical and conceptual models of neural systems, bridging the free energy principle with empirical neuroscience.
Active Inference for Social Sciences develops courses, curricula, research, and writing connecting Active Inference to the social sciences. The project produced a 2023 course and continues to build educational and research resources for social scientists engaging with the Active Inference framework.
The Active Inference Journal is an Institute publication and community knowledge channel. It supports the development and dissemination of Active Inference research, discussion, and learning through volunteer-led editorial work and open publishing.
The Active Inference Ontology project maintains and extends the public ontology for the Active Inference framework. It provides structured, machine-readable definitions of concepts and their relationships, supporting decentralized science, reproducibility, and knowledge reuse.
Audio-Visual Production is a sustained Institute project responsible for planning, recording, and publishing the Institute's livestreams, podcasts, video events, and educational recordings. The project has produced a continuously updated table of all livestreams and videos from 2020 onward.
Educational Course Development is a sustained Institute project producing structured courses in Active Inference and related topics. It maintains an Obsidian repository and course catalog, developing educational materials for a range of backgrounds and learning goals.
The Physics Course is an Institute education project developing course materials on the physical foundations of Active Inference and the Free Energy Principle — covering thermodynamics, information theory, and the physics underlying biological self-organization.
The Seasonal School is an Institute educational program providing intensive, structured, and in-depth engagement with Active Inference theory, modeling, and applications. It has run multiple cohorts and developed a track record as a concentrated learning experience for participants from varied backgrounds.
The Textbook Group is a sustained Institute educational program running structured cohort-based learning through Active Inference textbooks. Since 2022 it has run 9 cohorts on the 2022 textbook 'Active Inference: The Free Energy Principle in Mind, Brain, and Behavior' by Thomas Parr, Giovanni Pezzulo, and Karl J. Friston. As of mid-2026 the group is live in its first cohort on the 2026 textbook 'Fundamentals of Active Inference: Principles, Algorithms, and Applications of the Free Energy Principle for Engineers' by Sanjeev V. Namjoshi.
The Video Improvement Project focuses on enhancing the quality, organization, and accessibility of the Institute's extensive video library — covering hundreds of livestreams, educational sessions, and project recordings from 2020 onward.
This Ecosystem project develops a formal Active Inference account of belief updating in PTSD. It applies predictive processing and free energy frameworks to model how traumatic experience disrupts normal belief updating, offering a principled theoretical basis for understanding and potentially treating PTSD.
This Ecosystem project develops an Active Inference agent model of human translation processes. It applies the free energy framework to the cognitive and linguistic tasks involved in translating between languages, producing models of human translation behavior.
Anima is an Ecosystem project exploring how Active Inference principles apply to interactions with large language models. It examines how policy-based, blanket-aware Active Inference architectures can inform the design and understanding of current AI systems.
Artificial Sentience is an Ecosystem project examining the theoretical conditions for sentience in artificial systems. Drawing on Active Inference, it explores what it would mean for a machine to be sentient, using the free energy principle as a framework for understanding experience and self-organization.
Clinical Waveform Data Based Agent develops Active Inference systems for bedside clinical monitoring. It applies the free energy framework to real-time waveform data — EEG, ECG, and other physiological signals — to support clinical decision-making.
CogNarr (Cognitive Narrative) is an Ecosystem project developing infrastructure for facilitating group cognition at scale. It builds tools and frameworks that enable communities to coordinate shared understanding through structured narrative and cognitive scaffolding, with an initial focus on minimal viable incubation.
Energy Modeling Human Brain Metabolism applies Active Inference and the Free Energy Principle to model metabolic processes in the human brain. The project develops quantitative models of how the brain manages energy resources as an inference problem.
This Ecosystem project turns a robotic microscope into an Active Inference-driven intelligent agent for managing soil biology at scale. It develops AI systems that autonomously operate microscopy hardware, acquire data, and make decisions — treating the instrument as an inference agent in a biological environment.
Geometric Inquiry Theory develops a geometric framework for understanding inquiry as a structured dynamic process — establishing Q-State Dynamics and the structural basis of inquiry as a coherent mathematical theory, drawing on a multi-decade background spanning paramedicine, network engineering, and culinary arts as informal laboratories.
Graphspeak / Blorbbe is an Ecosystem project developing open-source tools for graph-based communication and knowledge representation, aiming to make structured, relational knowledge more accessible and usable for broad audiences.
Humanity's Story of an Uncertain Self is an Ecosystem project developing a broad account of human self-knowledge as an inference problem. Drawing on Active Inference, it examines how humans construct and maintain stories about themselves as agents in an uncertain world.
This Ecosystem project improves the model visualization capabilities of RxInfer.jl — developing tools that make it easier to see, explore, and understand the structure and dynamics of generative models built with the RxInfer system.
HDPLS-TARS is a speculative and exploratory Ecosystem project testing the limits and foundations of M-Theory through the lens of hyper-dimensional field interconnections. It examines new functional and experiential insights by treating M-Theory as a function of prismatic light scattering dynamics.
Model-Centric Cognition is an Ecosystem project developing a model-centric theory of cognition anchored in the Wave Hypothesis. It examines how cognitive systems represent and update internal models, drawing on Active Inference and existing wave-based theoretical traditions.
Project Sweet (Sus) Dogg applies Active Inference principles to understanding and improving human-animal relationships — particularly focusing on alignment and trust-building in interactions between humans and domestic animals.
Symbolic Cognitive Robotics is an Ecosystem project applying Active Inference to robotic systems that combine symbolic reasoning with embodied action. Work draws on papers in robotics and embodied cognition, and includes implementation on physical robot platforms.
This Ecosystem project develops a formal bridge between the Einstein model of a solid and the mental apparatus, viewed through the economic perspective of psychoanalytic theory. It applies Active Inference to bridge psychoanalytic concepts with modern neuroscience and artificial intelligence.
The Universal Basic Income Experiment applies Active Inference to study Universal Basic Income (UBI) — using token economics, policy simulation, and behavioral modeling to examine UBI's effects on human flourishing and economic dynamics.
Action Research on Collective Foraging (Negotiation Affordances) applies Active Inference to collective foraging behavior — studying how groups form coalitions, negotiate opportunities, and sustain value exchanges. The project has a focus on long-term sustainability and social dynamics.
The Active Inference Cycle Book for Self-Knowing develops a practical framework for personal growth and self-knowledge grounded in Active Inference. It uses the inference cycle — perceiving, modeling, acting, learning — as a scaffold for reflective practice and long-term personal development.
This Ecosystem project develops accounts of creativity and creative agents under the Free Energy Principle — examining how generative models and free energy minimization explain, predict, and enhance creative behavior.
Froebel's System studies the educational philosophy and methods of Friedrich Froebel — the inventor of kindergarten — through the lens of Active Inference and integral studies. The project captures and analyzes Froebel's approach as a cohort-based study, using Common Concepts as a prototyping platform.
Fundamentals of Active Inference is an Ecosystem project supporting the development and dissemination of the Fundamentals of Active Inference textbook — a comprehensive introduction to Active Inference principles, algorithms, and applications for engineers. The Institute hosts a Textbook Group cohort working through this book.
MathArt Conversations is an Ecosystem project creating a space for exploring the profound connections between mathematics and the arts. Through collaborative conversations, streams, and shared inquiry, it surfaces deep structural resonances between mathematical structures and artistic creativity.
Neurodivergent Learning Sessions develops Active Inference learning resources and sessions designed for neurodivergent participants — building curriculum, milestones, and community structures that support autistic, ADHD, and other neurodivergent learners in engaging deeply with Active Inference.
Numinia is an Ecosystem project developing an autonomous AI and educational adventure game grounded in Active Inference principles. The first mission embeds Active Inference in the values of the Numinia agent, creating an environment where players and agents co-learn through play.
This Ecosystem project develops a framework for solving the Tower of Babel Problem — the challenge of cross-domain knowledge communication and translation. Drawing on philosophical and physical principles, it outlines, develops, and implements UniFysica Philo-sophia as a universal conceptual bridge.
The Three Mosqueteers is an Ecosystem project creating a live science communication show for people without a scientific background. It develops a format that makes scientific information — including Active Inference and related work — engaging and genuinely accessible.
The Institute's strategy is organized around its motto — Act. Infer. Serve. — and pursued across four areas: Education, Research, Outreach & Engagement, and Methods. Each area is approached at three scales: the participant, the Institute, and the wider ecosystem. Together these give a single, legible frame for how the Institute advances Active Inference as a scientific, educational, and applied practice.
Project Measurement is how the Active Inference Institute tracks the health and progress of its projects — and the short Measurement form is how you contribute an update. Anyone can submit a measurement about a project or a Domain of Application for Active Inference; submitting one is the way to have your update included in Institute communications and to keep your project visible and active.
Project Preparation is how you let the Active Inference Institute know what you are setting out to do. Submitting the Preparation form is the way to get your project listed on the public site and to receive relevant support — it is the "Prepare" half of the Institute's Prepare-and-Measure system, paired with the Measurement form you use later to report progress.
Affordances is an Institute research thread that studies affordances — the action possibilities an environment offers an agent — through the lens of Active Inference and the Free Energy Principle. The concept originates in James J. Gibson's ecological psychology, where perception and action form a single coupled loop rather than separate stages. This thread reads affordances as relations defined jointly by agent and environment, and asks how an Active Inference agent comes to perceive and act on them. The work is part of the ReInference Unit's open research program.
The Wave Hypothesis thread gathers Robert Worden's work on the Brain Wave Hypothesis and the Projective Wave Theory of Consciousness, which he presented at the Active Inference Institute during 2024 through the GuestStream #082 series. These are Worden's proposals; the Institute hosts the presentations, links the underlying papers, and invites public commentary on the ideas.
The Active Inference Institute publishes an extensive library of videos and podcasts — recorded livestreams, learning-group sessions, interviews, lectures, and presentations. Public video is hosted on the Institute's YouTube channel, and audio episodes are distributed through podcast platforms. Produced as part of the Institute's open, participatory model, the library lets anyone follow the work as it happens or return to it later. The full curated catalogue is maintained on the video hub.
The Institute runs a weekly rhythm of public sessions, learning groups, and project meetings. The most current view of what's happening this week is the public Calendar, which lists upcoming sessions soonest-first — so this week's events appear at the top — and lets you search and filter to find a specific session.
2025 is the Active Inference Institute's annual overview for the year — a consolidated, month-by-month entry point to the activities, programs, and events that ran across the organization under the motto Act. Infer. Serve. The detailed record continues to be maintained on the 2025 hub.
2026 is the Institute's ongoing year — a light entry point to the various things happening across the organization. The best places to follow along are the Activities page and the Calendar, where recurring sessions, working groups, and public events appear as they are scheduled. The fuller record is maintained on the 2026 hub.