The Active Inference Institute & Active Inference Ecosystem
Abstract
This document surveys the current state of The Active Inference Institute and The Active Inference Ecosystem, in the context of our current and future directions. As embodied agents, we aim to update our decisions, goals and predictions as an institute by actively gathering (sampling) insights (observations) from our members. As Heraclitus once said “No one ever steps in the same river twice. For it’s never the same river and it’s never the same person”. In the same way, the Institute evolves with each new member, accumulating a variety of perspectives to drive improvement. The interactive and living form of this document can be found here.
Full text
Welcome The Active Inference Institute supports education, research, and applications of Active Inference is a participatory Open Science institute dedicated to improving the accessibility, rigor, and applicability of the framework. Active Inference Institute (AII) Active Inference ( ), containing pages and sections on and , such as: , , and . This is our main, interactive, living document chat with this document directly The Active Inference Institute The Active Inference Ecosystem Institute Programs Projects Ecosystem Support As of 2024 we are a 501(c)(3) educational non-profit organization ( and ).donate Philanthropy All backgrounds, time zones, and familiarity with Active Inference are welcome to in the and . Get Involved Institute Programs Activities Learn more the About Institute & Ecosystem See our playlist of for more on , including the and . quarterly updates through time The Active Inference Institute History of The Institute Institute Organization Join the for text and voice chats with others.Discord Check out and : Activities Institute Programs Engage in learning on your own time, and join synchronous activities when you can. Projects Ecosystem Projects Institute Projects for proposing projects at the InstituteProject ~ Preparation for reporting updates from Project ~ Measurement The Active Inference Ecosystem ◦ ◦ ◦ ◦ ◦ ◦ ◦
How to get started with ? Active Inference Explore this living Coda document See the pageStart See our ( ) past and upcoming Livestreams Videos and Podcasts Listen to the conversational Active Inference Insights podcast on or . YouTube Spotify Textbook Group Explore various roles and , which may fit you or others you know:Affordances , , Volunteer Internship Mentorship and Fellows Partnership , , Scientific Advisory Board Board of Directors Officers Readings and :Research The 2022 Textbook: “ ” by Thomas Parr, Giovanni Pezzulo, Karl J. Friston — See to learn this material in a collaborative setting. Active Inference: The Free Energy Principle in Mind, Brain, and Behavior Textbook Group “ ” by Maxwell Ramstead (October 2023)The free energy principle—a precis " " by Jared TumielSpinning Up in Active Inference and the Free Energy Principle “ ” by Beren Millidge FEP and Active Inference Paper Repository “ ”, Vyatkin et al. 2020Active Inference & Behavior Engineering for Teams A. Levenchuk, 2015 “ ”Towards a Systems Engineering Essence “An Active Inference Ontology for Decentralized Science: from Situated Sensemaking to the Epistemic Commons”, .Friedman et al. 2022 “ ” 2018 conversation-style interview with Karl Friston. Of woodlice and men: A Bayesian account of cognition, life and consciousness Code — see . Implementations of Active Inference Active Inference Institute is active on the following platforms: : Discord discord.activeinference.institute YouTube: youtube.com/c/ActiveInference X: https://x.com/InferenceActive BlueSky: https://bsky.app/profile/activeinference.bsky.social Podbean: https://activeinference.podbean.com/ Facebook: https://www.facebook.com/ActiveInference LinkedIn: https://www.linkedin.com/company/active-inference/ Email: [email protected] • ◦ • ◦ • • ◦ ◦ ◦ • ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ • • • • • • • •
Activities See for more background and context on the Institute.Welcome These events can be , and seen in the Events section. added to your calendar with this link Discord Email if you have questions about the activities or [email protected] This page shows the coming activities in the next 7 days, and information on all active projects of different types (e.g. , Institute Projects Feel free to drop in to any of these whenever it works for you — see the table for information on how to get involved with each project. Activities See for more information on proposing or measuring projects of your own.Projects Also see the and if you want to engage in a more structured way, and for specific contribution opportunities. Volunteer Internship Institute Programs Affordances Not synced yet The are and . Scroll down further to see by , members, , and . Projects at Active Inference Institute Institute Projects Institute Programs Projects Research Fellows Scientific Advisory Board Current Partners Ecosystem Projects Not synced yet ReInference 6 12/17/2025, 15:00 Active Inference GuestStream #125.1 ~ "From Charles Darwin’s “Root Brain” to Nikola Tesla’s “6G World Brain” and XAI-native 6G Networks" 12/18/2025, 16:00 ReviewStream 2025 12/19/2025, 18:00 2025 Quarterly Roundtable #4 CogNarr Ecosystem project Information on CogNarr Ecosystem project RxInfer.jl learning and development group Learn and apply RxInfer.jl in 2024 — building out multiscale se modeling. Knowledge Engineering This project seeks to alleviate the information burden in the Ac through information curation, organization, and condensationproductions (courses, livestreams, etc), enhancing the CRM, e Active Blockference We are applying Active Inference by building capacities & crea Date & Time (UTC) Event name Description Project Documentation Mission & Objectives Organizational Unit Activities at the Institute ~ Coming 7 days Projects at Active Inference Institute • • • • • ◦ ◦
EduActive 7 are organized by individuals in These projects are submitted via the form. Projects in the Active Inference Ecosystem The Active Inference Ecosystem Project ~ Preparation Projects by Research Fellows Not synced yet FarmWorks Develop minimal model of personalized agents Applied Active Inference Symposium To have a year-end Symposium, featuring applied Active Infere AICACP AICACP is a multi-year initiative designed to reshape the conve and regulation. Active Inference Ontology Maintain, improve, elaborate, extend, translate, educate, docu Ontology as core infrastructure for the Active Inference Institu Audio-Visual Production Produce accessible, rigorous, informative (epistemic value) an content, for example through Livestreams, Podcasts, and othe Active Inference Journal To develop evolving hybrid (AI+people) project architecture an Textbook Group (Parr, Pezzulo, Friston 2022) Improve the accessibility, rigor, applicability, and impact of the and Friston. Course Development Develop educational materials and experience to increase fam practice. Applied Active Inference Symposium Host an annual Symposium to highlight the state of the art in a Seasonal School Develop in-person experiences for education and developmen Symbolic cognitive robotics Explore the joint problem space of “symbolic active inference”, “societies of mind” and “morta computing”, with an emphasis on unsupervised learning. Using symbolic processing, build a rudimentary artificial agent (a LEGO rover robot) whose b fulfills the requirements of Active Inference Active Inference Cycle Book for Self-Knowing Perform a meta analysis of the “wellness” space through the lens of active inference highligh most impactful points for the larger population in an easily digestible format. Use this work t longer term collaboration and contribution to the larger AII community. Project Documentation Mission & Objectives Projects by Research Fellows
Projects by members Scientific Advisory Board Not synced yet Projects by ( organizations) Current Partners Partnership Not synced yet CogNarr Ecosystem: Facilitating Group Cognition at Scale The initial mission is to advance the CogNarr project from its current incubation phase into a concept demonstration, followed by a minimal viable product. In concept, the CogNarr ecosystem of software and tools is designed to serve as a compone group’s cognitive architecture. Model-Centric cognition Develop the central idea, raise awareness; assess whether this departure from the brain-asparadigm is needed. Some pushback is expected, even hoped for. The project is a developm existing wave hypothesis. Humanity’s Story of an Uncertain Self Producing an academic paper or blog that containsa set of equations, computer simulations ultimately a framework that explains the core components of humanity’s sociological-narrativ framework. Specifically, breaking down a few pieces of say, ancient epics, along with an set economic and civic institutions, would allow us to model to simulate, predict, and give maste otherwise seemingly intractable world of humanity’s cultural niche. Project Development for “Solving the Tower of Babel Problem: UniFysica Philo-sophia” To outline, draft, a collection of papers with the title “An Inclusive System of Communication Shared Meanings and Cognition: From Blombos to Friston and Fields” Creativity and creators under the light of the Free Energy Principle Design and run experiments to answer the key questions The Three Mosqueteers Create a livestream aimed at disseminating science and helping people without a scientific b to adopt a more critical attitude toward the information they receive. Numinia First mission would be to make sure we are implementing Active Inference in the game prope well explained, another mission would be to ensure that the design of the incentives aligned values of Numinia and the AII. Project Documentation Mission & Objectives Project Documentation Mission & Objectives Projects by SAB members Projects by Partner organizations
Ecosystem Projects Not synced yet Active Inference Account of Belief Updating in PTSD Write a theoretical paper in the style of Parr et al. chapter 6 Improving RxInfer.jl’s Model Visualization Capabilities Our mission is to equip RxInfer.jl - and its relevant component libraries - with a host of model visualization modalities that prove useful to those who wish to use, and/or to develop RxInfer To that end, we anticipate measuring the initial quality of our contribution/s by their reception RxInfer.jl’s core developers: TU/e’s BIASlab. All our objectives must therefore take the approva BIASlab as their proverbial North Star. Neurodivergent Learning Sessions Neurodivergent learning is focused on outreach and spreading awareness geared towards th struggle with standardized curriculum environments when it comes to public and higher educ milestones... as a number of people with neurological conditions not limited to autism spectru can struggle in varying ways with learning and being in the right environment in which inform presented to them in a manner which is coherent. The Unordinary Bible Study (abbreviated as TUBS) Hosting once a month sessions that focus on cross-referencing biblical verses but not spend much time digging into scripture as opposed to focusing on inter-faith and contemporary per focused dialogue. The Einstein Model of a Solid as a Model of the Mental Apparatus from the Economic Perspective of Psychoanalytic Theory. Bridging Psychoanalysis and Thermodynamics with applications to Artificial Intelligence. App AI. Project Sweet (Sus) Dogg To Help Warm-up or Prepare a Plausibly Notable Aspect of Agent Based Alignment By Social Active Inference for Built Environments & CooperActive Systems To advance the application of Active Inference in designing, managing, and evolving built env that prioritize the flourishing of all life on this planet. While humans possess unique cognitive capabilities, we recognize that excessive anthropocentrism blinds us to the needs of other liv organisms. Our work centers on life prosperity as the foundational principle for all built envir decisions. We seek to develop adaptive, nature-integrated solutions through distributed intelligence, di technologies, and decentralized decision-making that serve the broader web of life while me human cooperative living needs. Project Documentation Mission & Objectives Projects in the Active Inference Ecosystem
Affordances Specific opportunities for your contributions Check out the (opportunities for action) table below, and email , or follow specific instructions, if you are interested in exploring more: Affordances [email protected] Not synced yet Livestream and Podcast organizer/contributor Have you enjoyed the Active Inference Institute videos/podcasts? I would GREATLY enjoy the collaboration of 1 or more people in planning and implementing the video production for 2025. for curating and inviting guests, on through implementing the recording or stream. Expertise in Active Inference is not required. Truly this is a great opportunity for people of any background, who want to learn more about the space, connect with the authors/researchers personally, and have a big impact in increasing the visibility and accessibility of Active Inference. If you might like to join on this journey in 2025 -- Email [email protected] with subject [PRODUCTION] https://video.activeinference.institute/ https://www.youtube.com/@ActiveInference/ There is a checklist Contribution sought Details Mo Affordances
Weekly Update Announcements for week of December 15, 2025 Greetings. The Institute is enjoying a winter break until January 2026. Read on for some of the last updates of the year, ways for your end-of-year updates to get visibility, and areas of contributions for next year. 1. Submit Your End of Year Updates We would love to include your updates in the December Newsletter and the Quarterly Roundtable. Please use the for providing updates on your (research, learning, application) work, by December 17th. Measurement form https://measure.activeinference.institute 2. Applications open for 2026 Scientific Advisory Board (SAB)! The is a collaborative group of professionals who informally advise the Institute and serve as reviewers, mentors, contributors, and co-creators. For the coming year of 2026, we welcome applicants with backgrounds in Active Inference, as well as more broadly in education, research, open source, technology, and professional service. Membership on the SAB requires a modest time commitment (0–few hours per month), communication skills, and a shared enthusiasm for advancing the Institute’s mission and our broader field. We work to make the experience meaningful and streamlined. SAB If you would like to be considered, please complete before the end of December 2025. Additionally, if you know someone who would be great for this position, feel free to pass this opportunity along to them. All information: this form https://sab.activeinference.institute/ 3. Upcoming Livestreams: GuestStream #125.1 ~ at UTC with Nika Hosseini, Osman Tugay Bosaran, and Martin Maier From Charles Darwin’s “Root Brain” to Nikola Tesla’s “6G World Brain” and XAI-native 6G Networks 12/17/2025 15 https://www.youtube.com/live/LmgPFAlNHNQ ReviewStream 2025 ~ 12/18/2025 at 16 UTC 2025 Active Inference Livestream Review https://www.youtube.com/live/gS-qhMNFm84 2025 Quarterly Roundtable #4 ~ at 18 UTC 12/19/2025 https://www.youtube.com/live/09FfbL1YYOI See for more information if you would like to contribute to the scheduling and production of future livestreams. https://youtu.be/TaFwI2zr_lE 4. More: Join the Discord: http://discord.activeinference.institute/ Learn more about the Active Inference Ecosystem http://ecosystem.activeinference.institute/ • • • • • •
Ryan Henry Yale University 0000-0002-0706-6841 Sandeep Ramesh Panopticon Ventures; Primordia Co. 0009-0006-4976-3326 Scott David University of Washington: Applied Physics Laboratory 0000-0003-0679-3286 Sebastian Alvarado Queens College + City University of New York, Biology Department 0000-0001-5866-4043 Zach Baker University of Colorado 0009-0006-7283-9392
Active Inference What is Active Inference? is an integrated physics-based approach to modeling cognition and behavior as the active minimization of prediction error. Arising from the empirical study of cognitive systems (those involved in perception and action), Active Inference now is being explored across many . Active Inference Domains of Application The formal aspects of the framework describe in mathematical terms the tendency of complex adaptive systems to selforganize as to maintain low-surprise states (formally, through minimization of Variational Free Energy). Active Inference treats this tendency as the basic process, enabling the modeling of perception and behavior in various kinds of cognitive agents, including but not limited to humans. For those encountering this term for the first time, this can sound technical and obscure, but Active Inference can also be first understood more conceptually and practically as a framing for analysis that is broadly useful towards addressing or gaining perspective in a wide variety fields that formerly seemed unconnected. At its most basic level, Active Inference can be compared to the guessing game called “20 Questions,” a game in which one person is challenged to guess the identity of an object imagined by another person. In the game, each addition question asked is the “active” part of active inference, and the responses serially constrain the next question as the person guessing. Through this process surprise (bounded by “Variational Free Energy”) due to the differences of an observers “internal model” and the outside reality is reduced offering advantages to the cognitive/behavioral system, whether that system is a cell, an organism, a human organism, or an organization. For example, Active Inference finds application as diverse as and ecology (see ). It is not surprising that Active Inference framing is broadly useful in structuring a deeper understanding of the information flows associated with human cognition and bio-social behaviors in a variety of interaction settings and contexts, since Active Inference first emerged from the study of information flows in nature, where the organizing effects of its thermodynamic underpinnings are expressed most freely. mental health Domains of Application When Active Inference analysis is directed toward human social and organizational structures and behaviors, it reveals how relevant these bio-physical imperatives are when reflected and expressed in our everyday world. Greater awareness of this foundation, both in individual and organizational contexts, could enhance the overall effectiveness of a variety of information and communication systems and structures, many of which have never enjoyed a “spring cleaning” since their respective historical inceptions. We could, for example, define Active Inference in everyday language as an approach to understanding our interaction with the world and with those around us, how can we create some sort of model to understand how or why we behave as we behave and then apply this understanding to improve how we self-manage our shared models in a quickly changing reality. More formally, one could rather approach it through mathematics, and explore the foundational aspects of the operation of Active Inference. The advice to new members of the community who are looking for the best way to begin interacting with the broad range of materials and use cases impacted by active inference analyses is to seek the papers, discussions, materials that present the most familiar vocabulary, narratives and metrics as the starting point, and then to explore from there. To look for terms and keywords across the Active Inference resources that appeal to you most readily and start your journey by following those threads. . For background readings related to the theoretical basis of , see: The 2022 Textbook: “ ” by Thomas Parr, Giovanni Pezzulo, Karl J. Friston (focus of the ), “ ” by Maxwell Ramstead (October 2023), " by Jared Tumiel (October 2020), and “ ” 2018 conversation-style interview with Karl Friston. Active Inference Active Inference: The Free Energy Principle in Mind, Brain, and Behavior Textbook Group The free energy principle—a precis Spinning Up in Active Inference and the Free Energy Principle Of woodlice and men: A Bayesian account of cognition, life and consciousness
For seeing specific applications of Active Inference, see , as well as & . In short — read on! Domains of Application Institute Projects Ecosystem Projects What are key claims and aspects of Active Inference? Active Inference is scale-free as both a theoretical framework and a modeling approach. It characterizes all [information processing?] systems [of interacting components?] as behaving in a way that satisfies a single, fundamental goal: every systems acts so as to maintain the distinction between it and its environment. It characterizes all systems as employing the same strategy to achieve this goal: maximizing their ability to predict how their environment will next impact them. Active Inference thus characterizes all systems - from elementary particles to planetary ecosystems - as agents that both observe (accept input from) and act on (transfer output to) their environments. This information transfer is defined at the agent-environment boundary. For any agent, preserving its distinction from its environment is preserving its boundary, which preserves its identity. The Active-Inference process is, therefore, sometimes referred to as “self-evidencing”: any Active Inference agent continually provides its environment with evidence of its existence. By treating all systems at all scales as agents, Active Inference embraces a minimal, physical definition of “freedom”: an Active Inference agent is “free” in the sense that its next action is not causally determined by its environment. One can also put this as: the current state of an Active Inference agent is not causally determined by any, or all, of its environment’s past actions on its boundary. Freedom in this sense - freedom from local, causal determinism - is guaranteed to all physical systems by the Conway-Kochen theorems ( , ), which show that local, causal determinism in inconsistent with special relativity, which requires that causal processes take time, and quantum theory, which forbids the state of any system to be fully characterized by a single measurement. Hence Active Inference agents have internal states, and internal processes, that are “protected” from their environments by their boundaries. “Self-evidencing” is, therefore, also “maintaining one’s freedom of action”. 2006 2009 The generality and action orientation of Active Inference makes it a natural bridge between descriptive approaches to systems, and prescriptive approaches to implementation of artificial intelligence (e.g., machine learning) and design (e.g., user experience, communication, policies, , , etc.). Active Inference therefore enables a principled account of composition and decomposition, construction and de-construction, in complex adaptive systems. This generality provides a unified conceptual and pragmatic approach towards establishing a foundation for modeling, designing, and implementing various information processing systems across scales, disciplines, and settings. Active Inference is, therefore, intrinsically a trans-disciplinary framework both for theory and for modeling. As such, it provides a powerful common language into which discipline-specific languages can be translated. BOLTS requirements Active Inference leverages Bayesian principles, couching how systems perceive, learn, and act in their environments. It thus treats “knowledge” or “belief” as expectation or prior probability. It treats all agents as Bayesian satisficers, “doing the best they can do” in their environments given how they expect their environments to behave towards them. Over the last several decades, has been attracting increased attention as a quantitative and cognitive framework capable of acting as a common bridge, or Rosetta Stone, among various domains, and is gathering support across . Some citation search measures of this growth in popularity for “Active Inference” and “Free Energy Principle” are shown. Deeper is needed to make stronger inferences about the growth and change of the ideas and their applications, in the research literature and beyond. Active Inference Domains of Application Knowledge Engineering “Active Inference” on & PubMed arXiv “Free Energy Principle” on & PubMed arXiv for “Free Energy Principle” and “Active Inference” Google Books N-Gram viewer What is this excitement and growth about?! Read on to learn about the , the , and explore the depth and The Active Inference Institute The Active Inference Ecosystem
breadth of the work ongoing.
The Active Inference Institute The Active Inference Institute is a registered non-profit organization (Delaware, USA) which identifies, establishes, scaffolds, and supports the sustainable implementation of: Education and Research services. We learn and teach Active Inference We host and Institute Programs Institute Projects We provide visibility and opportunities for Ecosystem Projects Participation, communication, advisory, governance, and meta-governance affordances within the Institute and The Active Inference Ecosystem Publishing, and licensing protocols that establish , fair use, and effective dissemination of community products within and beyond the Ecosystem. Open Source services such as , , , and operation of cyber and cognitive ˜security systems aligned with our Ecosystem Support Communications Grants Partnership Mission, Vision, Values, and Principles The rest of this section covers: since founding in 2021History of The Institute Mission, Vision, Values, and Principles in terms of ongoing challenges (”where you find the challenge is where the learning/solution is done!”) Focus Areas for the Institute we are taking in light of the focus areas. Directions for the Institute , or morphology, in terms of roles and positions. Institute Organization and avenues for participation, such as , , , , , s. Institute Programs Volunteer Internship Fellows Philanthropy Grants Partnership hosted by the Institute Projects Organizational Units 1. a. b. c. 2. 3. 4. • • • • • • •
History of The Institute 2020 The begins in the co-founder team meeting in 2020 around a common interest in . This resulted in productive collaboration and the publication “Active Inference & Behavior Engineering for Teams” in September 2020 ( ). The group was then known as “Team Comm”. Check out , on July 28, 2020. History of The Institute Active Inference Vyatkin et al. 2020 our first livestream, ActInf Livestream #001.1 ~ “Narrative as active inference" Following the 2020 publication, discussions turned towards exploring approaches that could catalyze the accessibility, rigor, and applicability of Active Inference, and how to merge the developing framework with the and . Out of these discussions an “Active Inference Lab” (or ActInfLab) was formed and began operations in 2021. Systems Approach Open Source 2021 Over the first year of our operations, dozens of individuals from around the world engaged with ActInfLab through various projects such as educational , publishing, collaborative research projects, focused learning groups, , and initial developments of the . Production Open Source Active Inference Journal Active Inference Ontology Since the first quarter of operations in 2021, the ActInfLab hosted for communicating quarterly expectations and results to the community, a tradition that we continue to this day. Quarterly Roundtable livestreams 2022 Beginning in 2022, a cohort-based (SAB) was established to connect the ActInfLab to cuttingedge theoretical work as well as various domain-specific applications. As interest in both the ActInfLab’s activities and Active Inference itself began to grow, ActInfLab soon emerged as a key facilitating organization in what was then a primarily academic community working on the underlying theory and potential implications for Active Inference. Scientific Advisory Board The first Active Inference textbook comes out in 2022 ( ), and the Institute begins hosting a (ongoing through 7 cohorts in 2024). The Textbook Group is an important ecosystem service, as there are few academic/institutional locations where learners can be supported through the curriculum of the textbook and beyond. Additionally, the Institute has curated and categorized learning materials that learners create while participating in the group, including questions and discourse. Parr, Pezzulo, Friston 2022 Textbook Group The Institute begins the program to scaffold and support the learning journey of learners. Interns come from different backgrounds — including high school, college, and graduate students on academic tracks, as well as professionals and others outside of academia. Interns, with their mentors, develop a personalized education and research curriculum which lasts months-years. Internship In mid-2022, ActInfLab made the developmental leap to become , a non-profit organization registered in Delaware, USA with the intention of making its facilitatory role in the community impactful and sustainable. As part of the requirements for a non-profit, we also laid out the , comprised of the : , , and . The Active Inference Institute Institute Organization Organizational Units Administrative EduActive (Education) ReInference (Research) At the end of 2022, the has its first meeting. The Board continues to meet on a quarterly basis. Board of Directors 2023 and continue, including the first two full course offerings: and . These courses span months, and include office hours with the lecturer and teaching assistants. Institute Projects Institute Programs Physics course Social Science course In addition to continuing livestream on YouTube (GuestStream, ModelStream, PaperStream, etc), the Institute hosts the popular . Production Active Inference Insights podcast
During the year, we begin researching and applying for private and government .Grants 2024 Organizationally, the Institute receives official recognition as a 501(c)(3) non-profit organization, supporting our efforts. We were able to achieve this milestone with the pro bono support of the law firm. Philanthropy Fried Frank The largest cohort to date of the makes many diverse contributions across projects. Scientific Advisory Board The program begins to highlight and scaffold the work of Ecosystem member. As of November 2024, there are 5 Research Fellows have joined. Fellows represent members of the Ecosystem who have contributed substantially to the ecosystem through publications and presentations. Fellows To meet the needs of trainees and Interns for one-on-one guidance with projects, we introduced the program. Members of the and select other individuals, volunteer to mentor and connect with individual trainees. Mentorship Scientific Advisory Board Following the Quantum Active Inference Prepare-Measure cycle described by Chris Fields in the 2023 , we implemented a “ ” system for and . Prepare and Measure allows people to set goals and report back when they have reached them. can be provided by anyone about different for . In contrast describes what someone is preparing to do, whether they are just letting us know, Physics course Prepare and Measure Institute Projects Ecosystem Projects Project ~ Measurement Domains of Application Active Inference Project ~ Preparation These always-open reporting systems are used to gauge the ongoing projects and work done by community members, and provide visibility to these updates in the .newsletter Work during this year remains all-volunteer. support begins to come in, supporting some operational software costs. We applied for several (such as and related to with the ). Philanthropy Grants FarmWorks AI safety RxInfer.jl Learning Group The collaborated on this leading up to the 4th on November 13th, 2024. Authors Institute & Ecosystem Applied Active Inference Symposium 2025 See for more information on the ongoing year!2025 •
2025 January and begin for the yearInstitute Programs Activities is now hosted at the Institute. Theoretical Neurobiology (TNB) Group February New with , collaborating on the and elsewhere. Partnership @Lazy Dynamics RxInfer.jl Learning Group March March 28 — 2025 Quarterly Roundtable #1 April May June June 27 — 2025 Quarterly Roundtable #2 July Summer break! Complete to have stay in active state.Project ~ Measurement Projects August September September 26 — 2025 Quarterly Roundtable #3 October November November 12-14th — 5th Applied Active Inference Symposium Recruitment for next year and ,Board of Directors Scientific Advisory Board mostly finish up after the Projects Applied Active Inference Symposium December December 19 — 2025 Quarterly Roundtable #4 We continue to review , , , , , and more. Strategy Mission, Vision, Values, and Principles Ecosystem Support Philanthropy Grants We select the for . Board of Directors 2026 • • • • • • • • • • • • • •
We have applications open for the for Board of Directors 2026 •
2026 2026 as it happens January and begin! Institute Programs Activities February March April May June July August September October November December •
Turnover rate in engagement and participation (e.g., direct participant engagement with Institute releases and material, and annual involvement in collaborative activities) Number of individuals enrolled in educational courses Turnover and completion rate in educational courses Turnover rate in partnerships (e.g., research and education partnership decisions to renew, maintain, or dissolve) Social media analytics (e.g., views, watch time, audience diversity) • • • • •
Communications Internal Communications Plan Institute participants, , s, and other roles communicate with one another and with members of the community as follows: Officers Volunteer Email serves as the primary means of communication for internal announcements, updates, sharing important documents, and any other professional communications where record keeping is of interest. Regular Synchronous Officer Meetings are held to keep communication lines open, address questions, and discuss progress on projects. The meets regularly in an open discussion format. The meets quarterly to respond to the quarterly roundtable update and address any other issues or concerns. Scientific Advisory Board Board of Directors Shared Calendars are used to schedule meetings, appointments, and events, ensuring everyone is aware of each other's availability. The Institute-operated Server is the primary location for asynchronous discussion and synchronous project meetings. Currently there are over 1000 people in the server, and we strive to keep it an accessible entry point for learning and applying Active Inference. Discord Organizational Communication The Institute communicates with potential partners, sponsors, and relevant constituencies through channels including: Livestreams and provide exciting avenues for live community engagement.Production Content Announcements via X , , , , , Bluesky @inferenceactive Discord Facebook Newsletter LinkedIn @activeinference.bsky.social come from and reflect: completed projects, recent publications, collaboration and other project opportunities, new releases of educational materials and tools, etc Measurements The Active Inference Ecosystem Target Audiences Curious and exploratory learners from all backgrounds and levels of familiar with different subjects/skills. Professionals and Academics: Individuals with an interest in cognitive science, machine learning, philosophy, physics, linguistics, computer science, and related areas. Potential Partners: Government agencies, funding organizations, academic institutions, and other research-focused organizations. Active Inference Community: Researchers, academics and professionals who use and reference Active Inference and related approaches in their daily work. Broader Scientific Community: Researchers, academics, and professionals in compatible fields. Social Change Organisations: International Organisations, NGOs, civil society General Public: Individuals who may have a personal interest in cognitive science, machine learning, philosophy, physics, linguistics, computer science, and related areas. Research and Educational: Universities and academic institutions. Trade Associations and Think Tanks. Organizations which perform research about future industry trends, in addition to other communities of practice. Corporate: Companies with employees who would benefit from knowing Active Inference related approaches to business organization and operations. Government: Government agencies and funding vehicles. • • • • • • ◦ • • • • • • • • • • •
Private Donors: Individuals who understand the value and potential impact of this community of practice and its subject matter, and would be willing to help support it. Social Change Organisations: Taking basic underlying concepts and translating them into non-technical language and frameworks for organisations involved in change around large scale social issues (e.g., climate change, peace building) Approach The goal of our organizational communications plan is to provide the foundation for sustainable and accessible funding, and to work toward making Active Inference a household term, used as widely as “Machine Learning”, reflecting its demonstrable utility and impact in implementation. An ideal next step toward this goal is the professionalization of Active Inference core competencies and techniques and related competency and qualification standards. • •
Information Management The Institute hosts and disseminates information using , , , , and other platforms as needed. This stack of platforms streamlines specific levels of access to shared resources, and enhances overall productivity within the organization. We aim to ensure that participants are aware of the platforms being used and understand their purposes and functionalities. We regularly evaluate, communicate, and reinforce best practices for information storage, access, and organization. We implement security measures, such as strong passwords, 2-factor authentication, and appropriate access permission in order to protect sensitive information. We back up important data regularly to prevent loss due to technical issues or accidental deletion. We conduct periodic reviews and audits of the information storage systems to identify areas for improvement and optimization. The specific use of each platform is described below. Coda YouTube Discord Github Newsletter (Live Streaming and Video Hosting)YouTube YouTube is the primary platform for storing audiovisual content created for and by The Institute. Our designated YouTube Channel holds distinct playlists for courses, live streams, symposia, and other content that we host. We share and embed links within internal and external communication channels to provide easy access to relevant content. The content on YouTube is also backed up in a personal cloud storage service as well as in offline hard drives. (Forum and Instant Messaging)Discord Discord is our primary platform for engaging with the Active Inference Ecosystem and broader community. We use Discord for real-time communication, informal discussions, and team collaboration. Dedicated channels are used within Discord to categorize discussions based on topics or projects. Participants are encouraged to share relevant files, documents, or links within Discord channels, fostering easy access to shared resources. We regularly monitor and moderate Discord channels to maintain professionalism, and eagerly look to improve our protocols and guidelines here and elsewhere.
Discord Join the : Discord https://discord.activeinference.institute/ The Active Inference Institute (AII) maintains a server as its primary communication hub where all meetings, discussions, and collaborative activities take place. This digital workspace serves as the central nexus for the institute's diverse community of researchers, practitioners, and enthusiasts interested in active inference. Discord Server Structure Main Categories Research and education activities Project coordination Community discussions Voice chat rooms for meetings and livestreams Key Features The Discord server facilitates: Live voice meetings and discussions Project collaboration and coordination Access to educational resources Community engagement and networking Participation As with the Institute overall, the Discord server welcomes participants from: All backgrounds and experience levels Different time zones Various levels of familiarity with Active Inference The server can be accessed through the link . It serves as the primary venue for all institute meetings and collaborative activities, making it an essential platform for anyone interested in engaging with the Active Inference community. https://discord.activeinference.institute/ • • • • • • • • • • •
Coda Essentially all use system. Institute Projects Coda as a document Clicking through links and documentation of you will find many examples of links within and across documents — this was written collaboratively in Coda, and then exported for snapshot (whereas in 2023 version 1 we used a Google Document linear manuscript co-editing style). Institute Projects Institute & Ecosystem Coda is the primary platform for knowledge and project management at The Institute, Ecosystem, community, and individual scale. It organizes all information and content related to each project (or sub-project). Coda is version-controlled and accessrestricted, ensuring that all of our data is protected against accidental deletion and inappropriate user access. We use Coda for storing and organizing important documents, such as policies, procedures, project plans, and meeting notes. We follow best practices for Coda, including: (1) creating dedicated Coda “documents”, or work areas, for different departments or projects to ensure easy access and organization of relevant information, (2) implementing a clear folder and file structure within Coda to maintain document organization and version control, (3) archiving unnecessary and irrelevant pages, files, and folders, and (4) granting appropriate access permissions to users, allowing them to view, edit, or comment on documents as required. With adequate future support, Coda will be upgraded to an Enterprise License and consultants will assist in development of templates and low-code applications for streamlining support, records and knowledge management, and project management functions. Further, an Enterprise License will allow for a variety of new mechanisms for user-access control and permissioning, and for tracking of work activity and community engagement with hosted content.
Newsletter Since the initial activities of the Institute ( ), we have written a monthly . History of The Institute Newsletter See the archives https://activeinferenceinstitute.substack.com/ https://newsletter.activeinference.institute/
Directions for the Institute describe ongoing areas of activity and development at the Institute scale. Directions for the Institute The following table lists current developmental and connections with . Directions & Steps Focus Areas Below, we revisit the and outline some .Focus Areas for the Institute Directions for the Institute Research Advancement and Cross-disciplinary Expansion Seek for cross-disciplinary researchGrants Support core Active Inference research ( ) and educational ( ) development ReInference (Research) EduActive (Education) Explore implications in philosophy, social sciences, and other Domains of Application Facilitate collaboration with other cognitive models and research communities Research Advancement Support core Active Inference research; Explore theoretical implications in ; Examine group cognition functionality Philosophy Research papers; Theoretical frameworks; Computational models Deepened understanding of Active Inference; New insights at the intersection of multiple fields; Improved models of collective cognition Software Development Improve visualization capabilities; Enhance usability; Develop and curate examples of RxInfer.jl PyMDP Domains of Application Updated software tools; User-friendly interfaces; Application case studies Open Source More accessible and powerful Active Inference modeling; Increased adoption by researchers and practitioners; Practical demonstrations of Active Inference in action Educational Outreach Develop curricula for different languages and contexts; Provide courses and workshops; Increase efforts Communications Comprehensive curriculum; Industry-focused courses; Educational materials for various skill levels Wider accessibility of Active Inference concepts; Increased industry engagement; Growth of skilled Active Inference practitioners Cross-disciplinary Expansion Seek grants for cross-disciplinary AI research; Pursue features in popular science media; Focus outreach to social sciences and proposals; Media articles; Collaborative research projects Grants Broader adoption of Active Inference across disciplines; Increased public awareness; New applications in social sciences Community Growth Facilitate intern-mentor connections; Encourage SAB member interactions; Foster edge interactions within community Mentorship program; Enhanced community engagement; Collaborative projects Stronger, more connected Active Inference community; Knowledge transfer between experts and newcomers; Innovative crosspollination of ideas Public Engagement Translate concepts for broader public; Address societal challenges through Active Inference; Provide foundations for trust and ethics in AI Accessible content; Applied solutions to real-world problems; Ethical guidelines for AI development Increased public understanding of Active Inference; Real-world impact on societal issues; Responsible AI development informed by Active Inference principles Practical Application Develop policy appraisal methodologies; Consider ethical and cognitive security aspects; Research capabilities in various domains Policy frameworks; Ethical guidelines; Domain-specific applications Informed decision-making in policy; Enhanced cognitive security measures; Demonstration of Active Inference's versatility across fields Direction Method Deliverables Impact / Implication Directions & Steps 1. a. b. c. d.
Develop new policy appraisal methodologies with focus on ethical and cognitive security considerations Educational Outreach and Resource Development Develop a full academic curriculum for interdisciplinary audiences Create educational resources ( and Beyond) Fundamentals of Active Inference Provide courses on for industry professionalsImplementations of Active Inference Increase learning resources for coding Active Inference agents/simulations Develop foundations for trust, ethics, and education in the context of rapid AI advancement Software Development and Practical Applications With development, Improve and visualization capabilities and overall usabilityOpen Source RxInfer.jl PyMDP Develop real-world across Implementations of Active Inference Domains of Application Support multi-agent workflows (e.g. using )Active Entity Ontology for Science (AEOS) Create reliable and accurate models for engineers Community Growth and Engagement Facilitate connections with , , and membersMentorship Internship Fellows Scientific Advisory Board Foster edge interactions within the community and The Active Inference Ecosystem Implement automated feedback mechanisms Moderate community discourse to ensure compliance with culture and values Improve onboarding experience for new users Increase awareness and involvement from organizations outside the Institute Public Engagement and Knowledge Dissemination Translate Active Inference concepts for broader public understanding Develop strategies to disseminate knowledge to general public, and professional across areasCommunications Explore the intersection of with current global issues (social, economic, geopolitical, technological, environmental) Active Inference Continue to develop publishing and licensing support systems for contributorsOpen Source e. 2. a. b. c. d. e. 3. a. b. c. d. 4. a. b. c. d. e. f. 5. a. b. c. d.
Strategy https://www.activeinference.institute/strategy Active Inference Institute (AII) is on the path of open-endedness. Our considers learning and applying Active Inference for changes in the niche over multiple nested scales. Through time we increase the degree of hierarchical organizational complexity to overcome competing interactions and frustrated states. Strategy We engage in policy selection across multiple scales, reducing our uncertainty about realizing our expectations and preferences. We learn, finding epistemic value along the way, while pragmatically ensuring Institute persistence and development. The three scales that Active Inference Institute modifies and interact with: Participant as an agent. The Institute provides affordances and updates participants’ generative model via niche modification and offering of affordances. Institute as the agent. This is where we engage in Institute-level policies selection and evolve our shared generative model. as our epistemic niche.The Active Inference Ecosystem • • •
Organizational Units The of the Institute describe the main concentrations or nestings of Organizational Units for organizational and operational work Administrative for inquiry and learningEduActive (Education) for research and developmentReInference (Research)
Administrative The Institute Unit performs various support tasks within and the wider , such as project coordination, record keeping, graphic design, , project facilitation, preparation, and compliance, and other activities. Administrative The Active Inference Institute The Active Inference Ecosystem Grants Communications Administrative activities contribute to the development of core infrastructure to provide such support and automate or systematize and standardize tasks, and become the organizational umbrella for financial, human resources, , , security, community moderation and management, and related activities and organizational components. These tasks are currently assumed by The Institute’s , who will continue to provide oversight as the unit develops to include more contributors. Internship Volunteer Officers
EduActive (Education) The Institute’s Education Unit is named “EduActive” to highlight the active element of education. Projects of include: EduActive (Education) Active Inference Journal Active Inference Ontology Active Entity Ontology for Science (AEOS) Applied Active Inference Symposium Educational Standards & Qualifications Fundamentals of Active Inference Physics course Social Science course Textbook Group Production Not synced yet Ecosystem 7 Project Development for “Solving the Tower of Babel Problem: UniFysica Philo-sophia” To outline, draft, a collection of papers with th of Communication for Sapiens’ Shared Meanin Blombos to Friston and Fields” Numinia The Adventure of Curiosity (oncyber.io) https://app.charmver se.io/numinia First mission would be to make sure we are im in the game properly and is well explained, ano ensure that the design of the incentives aligne and the AII. Neurodivergent Learning Sessions Neurodivergent learning is focused on outreac geared towards those who struggle with stand environments when it comes to public and hig as a number of people with neurological condi spectrum disorder can struggle in varying way the right environment in which information is p manner which is coherent. Active Inference Cycle Book for SelfKnowing Perform a meta analysis of the “wellness” spac inference highlighting the most impactful point an easily digestible format. Use this work to k collaboration and contribution to the larger AII The Unordinary Bible Study (abbreviated as TUBS) Hosting once a month sessions that focus on c verses but not spending too much time diggin to focusing on inter-faith and contemporary pe Project title Examples of work Coda Mission objectives Type of project Active Projects ~ EduActive • • • • • • • • • •
Institute 7 Creativity and creators under the light of the Free Energy Principle Design and run experiments to answer the key The Three Mosqueteers Create a livestream aimed at disseminating sc without a scientific background to adopt a mo information they receive. Active Inference Ontology Public Active Inference Ontology website Maintain, improve, elaborate, extend, translate apply the Active Inference Ontology as core in Inference Institute & Ecosystem. Audio-Visual Production Table with all livestreams and videos from 2020Ongoing Produce accessible, rigorous, informative (epis (pragmatic value) audio-visual content, for exa Podcasts, and other formats. Textbook Group (Parr, Pezzulo, Friston 2022) 5.5 completed cohorts since 2022 (see )Coda Improve the accessibility, rigor, applicability, a by Parr, Pezzulo, and Fri2022 Active textbook Active Inference Journal See the Github To develop evolving hybrid (AI+people) projec volunteers team Course Development , Courses Obsidian Repository Develop educational materials and experience theory and practice.Active Inference Applied Active Inference Symposium Applied Active Inference Symposium Host an annual Symposium to highlight the sta Active Inference. Seasonal School Seasonal School Develop in-person experiences for education a
ReInference (Research) The Research unit at the Institute is named “ReInference” to highlight the perspective on scientific research and inquiry more broadly in terms of ( , ).Active Inference Pietarinen & Beni 2021 Balzan et al. 2023 The Institute’s ReInference unit focuses on Research activities such as: (i) the forming of fit-for-function interdisciplinary research teams, (ii) the development and execution of research proposals and projects aligned with the mission of The Institute and challenges faced by The Institute and Ecosystem at large. The ReInference unit is committed to hosting and sharing all relevant data, findings, publications, tools, and derivative artifacts under or similarly accessible licensing agreements wherever practicable and appropriate.Open Source Projects of include: ReInference (Research) RxInfer.jl Learning Group Active Blockference Active InferAnts Tech Tree Knowledge Engineering Generalized Notation Notation FarmWorks Not synced yet Institute 6 Knowledge Engineering and from end of 2022 Public frontend Literature meta-analysis This project seeks to alleviate the information b Ecosystem through information curation, organ summaries of institute productions (courses, liv Active Blockference , , video Github Blog post overview from 2022 We are applying Active Inference by building ca generative models. RxInfer.jl learning and development group See project overview Learn and apply RxInfer.jl in 2024 — building ou generative modeling. FarmWorks See and . FarmWorks page 2024 publication Develop minimal model of personalized agents Project title Examples of work Coda Mission objectives Type of project Active Projects ~ ReInference • • • • • • •
Ecosystem 9 Applied Active Inference Symposium 4 prior Symposia from 2021 through 2024 To have a year-end Symposium, featuring appli the world AICACP AICACP AICACP is a multi-year initiative designed to res capabilities, alignment, and regulation. Active Inference Account of Belief Updating in PTSD Write a theoretical paper in the style of Parr et Symbolic cognitive robotics , Most recent paper from 2023 Robotics & Embodied Explore the joint problem space of “symbolic ac “mortal computing”, with an emphasis on unsup Using symbolic processing, build a rudimentary whose behavior fulfills the requirements of Act Humanity’s Story of an Uncertain Self ~ Shagor Rahman: "Myth of objectivity and the origin of symbols" ActInf GuestStream 061.1 Producing an academic paper or blog that cont simulations, and ultimately a framework that ex humanity’s sociological-narrative framework. S of say, ancient epics, along with an set of econ us to model to simulate, predict, and give mast intractable world of humanity’s cultural niche.
Improving RxInfer.jl’s Model Visualization Capabilities Current work on visualization Our mission is to equip RxInfer.jl - and its releva model visualization modalities that prove usefu develop RxInfer.jl. To that end, we anticipate measuring the initial reception from RxInfer.jl’s core developers: TU/e therefore take the approval of the BIASlab as th CogNarr Ecosystem: Facilitating Group Cognition at Scale CogNarr (Cognitive Narrative) Ecosystem: Facilitating Group Cognition at Scale The initial mission is to advance the CogNarr p into a proof-of-concept demonstration, followe In concept, the CogNarr ecosystem of software component of a group’s cognitive architecture. Model-Centric cognition Wave Hypothesis Develop the central idea, raise awareness; asse brain-as-computer paradigm is needed. Some The project is a development of existing wave h The Einstein Model 2023 paper Bridging Psychoanalysis and Thermodynamics
Institute Programs The are the specific modes of active participation and engagement (beyond e.g. just watching ). Institute Programs Production For individuals: positions contribute to the Institute’s work in specific ways.Volunteer and provide structure for those looking to advance their learning and work. Internship Mentorship provides the opportunity for committed individuals to be recognized as a leader in research and education. Fellows Active Inference For organizations: The , , and programs all provide channels of support and bi-directional learning with the Institute. Partnership Philanthropy Grants • • • •
Volunteer Operationally, all participants of the Institute are volunteers. Volunteers join the Institute by emailing project facilitators, or communications from . Communications such as the website and also contain solicitations for signing up for general lists, and specific . Discord https://www.activeinference.institute/ Newsletter Volunteer Institute Projects As a community-driven open science organization, there are multiple opportunities for contribution. All backgrounds, time zones, time availability, and levels of familiarity with Active Inference are welcome and encouraged. Volunteers are active learners who want to contribute to ongoing projects at The Institute. Volunteers have the opportunity to engage in and lead a wide array of projects without any constraints on their type or quantity. These projects encompass a range of activities such as study groups, livestreams, marketing initiatives, publications, symposiums, and applied research. We look to develop and clarify how the position will work in 2025. Current thinking is exploring ideas around:Volunteer Official Recognition (via specific affordances and statuses such as: affiliation status for papers and communications, access to code repositories and digital resources, @activeinference.institute email address, letters of recognition, inclusion on , payment via , etc).Grants Philanthropy Role of in efforts and more broadly.Mentorship Volunteer Education Position and -specific Documentation, regular participation in prepare/measure cycles.Institute Projects Stewardship of specific paper sectionsDomains of Application Sampling among expertise areas to build Prediction Matter Expertise (PME), not just Subject Matter Expertise (SME). Facilitating discussions and answering basic questions. Contributing to projects.Open Source The volunteer program aims to balance structure with flexibility, providing clear value while maintaining active inference principles in learning and contribution. This framework allows volunteers to grow within the Institute while contributing meaningfully to its mission. to at the Institute. Complete this form Volunteer We keep Volunteers posted about affordances for Learning Groups, Projects, Internships, and more. Let us know in the final question response, or via email to if you have any questions. [email protected] • • • • • •
Hongju Pae 2025 Symposium Presentation 2025 Symposium Presentation I develop computational and theoretical frameworks for modeling how artificial agents … Personal webpage I develop computational and theoretical frameworks for modeling how artificial agents … Personal webpage Sheila Macrine As an Active Inference Institute Research Fellow, I am extending the theoretical framework … As an Active Inference Institute Research Fellow, I am extending the theoretical framework … Not synced yet Anna Pereira 0009-00089049-0707 5/2024 Cultivating a grass roots impact project (initially thro Principles. Active Inference is the key lens that then enabling humans to live more fulfilling lives, respond mutualistic opportunities for collaboration and seek Say hello, collaborate, or discuss at via anna@activ David Bloomin 10/2024 GuestStream 085.1 I am investigating how the principles of Active Infere foster cooperation and alignment in multi-agent env mechanism in gridworld simulations. The project aim minimize free energy. Through an open-source mod to aligned cooperative intelligence, informing the pa You can follow my progress at http://daveey.github. Hongju Pae 0000-00025174-8858 11/2025 2025 Symposium Presentation I develop computational and theoretical frameworks and achieve developmental alignment. My work inte identify computable markers of subjective experien simulation prototypes and mathematical models to without external reward shaping. My goal is to esta grounded machine minds. Personal webpage https://www.linkedin.com/in/hjpa Lab webpage https://hjpae.github.io/cear/ Jean-Francois Cloutier 0009-00011841-2279 5/2024 , Active Inference Symposium 2nd 3rd I seek to find out what it takes, at a minimum, for a nothing about. My research is the continuation of a and ground my understanding of cognition. Looking for answers has already taken me on an un drawn into Active Inference of course but also Kant collective intelligence, autopoiesis and constraint cl Name ORCID Starting date Image Livestreams Overview Research Fellows ~ Table
All information at: fellows.activeinference.institute John Boik 0000-00031289-7997 5/2024 Livestream #021 series: . , . , . , . , . , . 01 02 03 04 1 2 As an Active Inference Institute Research Fellow, th book and in two series of concept papers. That pro architectures that are, by design, fit for purpose. The first series describes how the approach can be governance systems), which are viewed as compon The second series describes how the approach can large-group setting. Robert Worden 0000-00017304-2752 10/2024 GuestStream #082 series: . , . , . , . 1 2 3 4 I have two main research interests: 3-D spatial cogn All animals need to understand the local 3-D space not by neural computing alone, but using a wave in a novel theory of consciousness – that it arises not using active inference. See Frontiers article on the t I also work on language – how it evolved, how we le a of language learning.demonstration Shagor Rahman 0009-00040460-0078 8/2025 GuestStream #061.1 I investigate how morality and symbolic thought coour capacity to model shared cultural expectations symbolic spaces. This framework offers a strong pe Employing multi-agent active inference simulations explicit moral beliefs form the foundation of our sym and impact psychological well-being through narrat This computational framework helps us understand religious prophets, cultural thought leaders of socia can reshape these symbolic systems. By formalizing offers insights into both human uniqueness and the Sheila Macrine 0000-00028600-0938 11/2025 As an Active Inference Institute Research Fellow, I a Embodied Intelligence: Multidisciplinary Perspect proposes a novel multi-dimensional taxonomy of ag systems. The first phase of this research investigate and Reflexivity—can be mathematically formalized a constructing formal "Agency Profiles," this work aim using precise metrics. For this model, Agency is the precision of intrinsic priors. Intelligence is the 'Engin efficiency that enables agency. This approach prov profiles characterized by high rationality but low au The second phase is dedicated to developing a sub Inference models across diverse scales—ranging fr understanding of how agency emerges, scales, and primary interest lies within the Theoretical Neurobio interdisciplinary collaboration and mathematical dev formalizing these concepts, aiming to operationalize
Research Fellow Application Research Fellow Application Package Requirements Confirmation that one has read & agreed to the . Research Fellows Terms Research proposal (1-6 pages, excluding citations), which can describe one or more research projects in detail or outline a research direction in more general terms. The proposal should include: Title Abstract (300 words or less) Research question(s) and objectives Significance and impact of the proposed research Approach and research methods Alignment with the Active Inference Institute’s mission Anticipated measurements (outcomes and deliverables) Timeline and milestones People and institutions involved Any dependencies or contingencies that might affect progress Cited references (not counted in page limit) In the research proposal, please clearly address: — Motivation for proposed work and for applying for a fellowship — Describe previous engagements/interactions with the Institute — Describe Institute programs or activities of particular interest and/or that you intend to participate in or facilitate. Curriculum vitae One to Three letters of recommendation (submitted in application packet, or sent separately to )[email protected] Up to five representative publications or products Application questions and completed packets should be sent to with [RESEARCH FELLOWS] in the Subject line. Please include all application components as separate PDF or document files attached to your email. [email protected] Research Fellows ~ Application Questions 1. 2. 3. 4. 5.
I have X educational degree / have no PhD — can I apply? Yes, you can apply. There is no requirement for a PhD, or any specific degree I am unemployed / doing Active Inference research on my own time / am employed doing other work / would be a part-time Fellow — can I apply? Yes, all employment statuses are acceptable in principle, as long as the appl is clear about what current/future obligations are and how one plans to proc How many Fellows will the Institute accept? We do not have a set fixed limit as of 2024. What are the Term limits? If 2 years is default, can we allow a term of less than 2 years? Yes, we will consider terms of less than 2 years. Fellowships are eligible for how many renewals? We do not have a set fixed policy on this as of 2024. I have some other questions Application questions and completed packets should be sent to with [RESEARCH FELLOWS] in the [email protected] Question Answer
Research Fellows Terms Terms v1 (April 2024) Terms for Active Inference Institute Nonemployee Research Fellows 1. Code of Conduct The Research Fellow (Fellow) agrees to adhere to the highest standards of scientific integrity, professional ethics, and responsible conduct of research. This includes honesty, objectivity, fairness, respect, accountability, and transparency in all research activities and interactions. 2. Fellowship Agreement The Active Inference Institute (Institute) reserves the right to terminate the fellowship agreement (agreement) at any time. The specific terms, conditions, and duration of the agreement will be outlined in the initial offer letter, which must be signed by both parties. The Fellow is responsible for ensuring that this agreement is compatible with their other institutional affiliations, contracts, and policies. 3. Intellectual Property (IP) Nothing about this agreement - changes the status of IP arising during the program (e.g. the Institute does not claim any access to Fellow’s private work, nor does the Fellows participation in Institute products/processes affect the Fellow’s or Institute’s IP rights), except as noted next. The Fellow may be required to sign additional IP agreements or disclosures as outlined in the initial offer letter (depending on the applicant’s situation and status). 4. Affiliation and Acknowledgment The Fellow is encouraged to list the Institute as a professional affiliation on research outputs, presentations, and professional communications related to their fellowship activities. The official title for this affiliation is "Research Fellow, Active Inference Institute". 5. Open Science and Research Productivity The Institute encourages and supports Fellows to disseminate their research outputs through open access channels, such as preprint servers, open data repositories, and open source software platforms. The Fellow should strive to meet the research objectives outlined in their approved proposal . 6. Reporting and Evaluation • • • • • • • • • • •
The Fellow shall provide regular progress reports (e.g. quarterly) to the Institute, describing their research activities, achievements, challenges, and plans. The Fellow shall participate in an annual evaluation process, which may include a written self-assessment, an oral presentation, and feedback from Institute mentors and collaborators. Satisfactory performance and progress, as determined by the Institute, are required for continuation and possible renewal of the fellowship. By signing the initial offer letter, the Fellow agrees to abide by the terms of this document. The Institute reserves the right to modify these terms as needed, with written notice to the Fellow. • • •
Philanthropy We are thrilled to share that in 2024 the Active Inference Institute has officially been recognized as a 501(c)(3) nonprofit organization by the United States IRS. This significant milestone is the result of several years of dedicated effort. Your generous support will help sustain and extend the work of the Institute and ecosystem, supporting the accessibility, rigor, and applicability of Active Inference. With our 501(c)(3) organizational status, your donation may be tax-exempt. If you value our mission and wish to contribute, please consider making a donation at: donate.activeinference.institute/ At this time, for any other comments or questions on philanthropy and donations to the Institute, please communicate with . [email protected] For the organizational Partnership affordance (which may include structured financial or in-kind donations), see .Partnership Thank you for your continued attention and consideration. We look forward to all the next moves we’ll take together.
Grants is primarily a volunteer organization. We seek sustainable approaches to scaffold our work through and . The Active Inference Institute Grants Philanthropy As part of our commitment to , submitted grants are made public whenever possible, by uploading to a preprint server (such as Zenodo) as a publication. In this way, we leave a stigmergic trace on the ecosystem reflecting our plans, and history of assembling teams to tackle areas of research and . Open Source Education Current Grants In June 2025 a team of researchers in the were awarded $270,000 through the Institute. AICACP Previous we have applied for:Grants In 2022 we applied for, and did not receive, “Systems Modeling and Cognitive Audits for Hypercert Ecosystems”. The application was published on . Zenodo In 2023, we applied for, and did not receive, an NSF Pathways to Enable Open-Source Ecosystems ( ) grant. Along the way, we collaboratively wrote the “The Active Inference Institute and Active Inference Ecosystem” — the structure and text of which, was the initial conditions (prior) for document. POSE 2023 paper Institute & Ecosystem In 2024, the applied for, and did not receive, an grant “FarmWorks: Decentralized AI Agents for Personalized Solutions” ( ). RxInfer.jl Learning Group FLI Zenodo link In 2024, the applied for, and did not receive, a grant “VILLAGE (Validating Inference for Large-scale Agent Governance Ecosystems)” ( ). RxInfer.jl Learning Group Foresight Institute document link In 2025, a team applied for, and did not receive, a grant “Increasing the Accessibility and Applicability of Active Inference: Generative Playbooks and Open-Source Summer School Curriculum Development” ( ). Dana Frontiers Zenodo link In 2025, a team applied for, and did not get accepted into, . Our team’s collaboration did lead to product development and two papers: and . a Google.org accelerator for generative AI ResNei: Solution Design Document The Discovery Engine: A Framework for AI-Driven Synthesis and Navigation of Scientific Knowledge Landscapes We have written multiple letters of support, collaboration, and for others in their applications and . Partnership Grants • • • • • • •
Partnership Organizational Partnership Program at the Active Inference Institute About the Partnership Program Partnership Program fosters collaboration with organizations aligned with its mission to advance the understanding, application, and accessibility of Active Inference. The Active Inference Institute By creating mutually beneficial relationships, the program enables partners to contribute to and benefit from the Institute’s research ecosystem, which spans diverse such as cognitive science, artificial intelligence, education, and organizational dynamics. Domains of Application Through partnerships, the Institute supports the development of , educational materials, frameworks, and tools that leverage Active Inference principles to address real-world challenges. Partners gain access to a vibrant network of researchers, developers, and thought leaders while contributing to the growth of the global Active Inference community. Projects Learn more: See and details of our engagementsCurrent Partners and Partnership Application Partnership Terms What the Institute can Provide Recognition and Access: Public acknowledgment on the Institute’s website and materials, along with access to its network of researchers, contributors, and interns. Collaboration Opportunities: Regular meetings with Institute personnel to co-develop programs, projects, and initiatives in areas of shared interest. Support for Specific Programs: The Institute facilitates collaborations on targeted initiatives such as educational , livestream s, , , themes, and region-specific work. Courses Production Internship Fellows Applied Active Inference Symposium Guidance and Expertise: Strategic insights into applying Active Inference principles in organizational or scientific contexts through workshops, training sessions, and mentorship. Flexibility in Engagement: Tailored levels of involvement based on partner preferences, ranging from casual participation to formalized collaboration. What the Partner may Contribute Financial or In-Kind Contributions: Provide monetary support and/or resources (e.g., compute power, datasets, development expertise) to sustain and expand the Institute’s work. Alignment with Goals: Submit an application demonstrating alignment with the Institute’s mission of advancing Active Inference and broadening scientific participation. • • • • • • • • •
Time and Attention: Dedicate an agreed-upon level of involvement, from casual engagement in projects to formal facilitation of programs. This can include direct participation on , service on , mentors for , or other avenues of engagement. Projects Scientific Advisory Board Internship Reliable Communication: Designate a point of contact for regular communication and coordination with the Institute. Programs and Projects That Can Be Supported The Active Inference Institute offers opportunities for partners to support a range of operational programs and specific projects. Benefits of supporting specific programs include: Partners can directly contribute to advancing science, education, or applied research in their areas of interest. Financial or in-kind contributions enable impactful initiatives that align with both partner goals and the Institute’s mission. Collaboration fosters mutual growth while expanding the reach of Active Inference principles across domains. By supporting these or projects, partners play a pivotal role in shaping the future of Active Inference research, education, and applications globally. See the table below for some ideas, or reach out if you have other ideas. Programs for Partnership support Institute Programs 4 Institute Projects 8 program Internship Mentorship-focused development opportunities for emerging talent. Cultivates future leaders in Active Inference research and applications. program Research Fellows Supports self-directed research projects aligned with Active Inference principles. Advances cutting-edge studies while fostering professional growth for researchers. Grants Builds administrative capacities for funding proposals. Enhances the Institute’s ability to secure resources for long-term sustainability. and Open Science Open Source Promotes transparency and accessibility through open-source tools and frameworks. Strengthens the global ecosystem of Active Inference research and applications. Open Source Code Development (e.g., RxInfer.jl) Develops generative modeling frameworks and tools for Bayesian agents. Provides foundational resources for researchers and practitioners worldwide. Active Inference Journal Publishes translations, transcripts, papers, and other scholarly outputs. Increases accessibility to key insights across languages and disciplines. Active Inference Ontology Formalizes Active Inference structures across languages and domains. Facilitates interdisciplinary collaboration by creating standardized conceptual frameworks. Applied Active Inference Symposium Organizes events focused on practical applications of Active Inference theories. Encourages knowledge exchange between academia, industry, and broader communities. Program/Project Description Impact Enabled by Support Category Programs for Partnership support • • 1. 2. 3.
Partnership Terms Terms v1 (April 2024) Active Inference Institute Partnership Program Terms and Conditions These Terms and Conditions ("Terms") govern the partnership between the Active Inference Institute ("Institute") and the partnering organization ("Partner"). By submitting an application to the Partnership Program ("Program"), the Partner agrees to be bound by these Terms. 1. Partnership Scope and Duration 1.1 The scope and duration of the partnership shall be as mutually agreed upon by the Institute and Partner in writing, based on the Partner's application and any subsequent discussions. 1.2 The partnership shall commence on the date the Partner is notified of acceptance into the Program and shall continue until the agreed upon end date, unless terminated earlier in accordance with these Terms. 2. Partner Obligations 2.1 The Partner shall make the financial or in-kind contributions specified in their application and agreed upon with the Institute. Contributions are non-refundable. 2.2 The Partner shall participate in and help promote Institute activities relevant to the partnership, as mutually agreed upon. 2.3 The Partner shall provide a designated point of contact who is responsive and reliable in communicating with the Institute. 2.4 The Partner shall respond to periodic Institute requests for information and feedback in a timely manner. 2.5 The Partner shall inform the Institute of any changes to their organization that may impact the partnership. 3. Institute Obligations 3.1 The Institute shall provide public recognition of the partnership, subject to the Partner's approval of any use of their name, logo, or other identifying information. 3.2 The Institute shall facilitate connections between the Partner and relevant Institute stakeholders, as appropriate for the agreed upon scope of the partnership. 3.3 The Institute shall coordinate joint activities with the Partner around areas of shared interest, as mutually agreed upon. 3.4 The Institute shall consider Partner feedback in its planning and decision-making related to the Program. 4. Intellectual Property
4.1 Each party shall retain ownership of any intellectual property they create or possess prior to or independently of the partnership. 4.2 Any intellectual property created jointly by the Institute and Partner in the course of the partnership shall be owned jointly, unless otherwise agreed in writing. 4.3 Neither party shall use the other party's intellectual property, including trademarks and logos, without prior written consent. 5. Confidentiality 5.1 The Institute and Partner may exchange confidential information in the course of the partnership. Each party shall maintain the confidentiality of such information and not disclose it to third parties without prior written consent, except as required by law. 5.2 Confidential information shall not include information that is publicly available, independently developed, or obtained from a third party without breach of any obligation of confidentiality. 6. Termination 6.1 Either party may terminate the partnership at any time upon written notice to the other party. 6.2 Upon termination, the Partner shall cease use of any Institute intellectual property and shall return or destroy any confidential information of the Institute in their possession. 6.3 Termination shall not affect any rights or obligations accrued prior to the effective date of termination. 7. Limitation of Liability 7.1 Neither party shall be liable to the other for any indirect, incidental, special, or consequential damages arising out of or related to the partnership. 7.2 The Institute's total liability under these Terms shall not exceed the amount of financial contributions made by the Partner in the 12 months preceding the event giving rise to liability. 8. Governing Law and Dispute Resolution 8.1 These Terms shall be governed by and construed in accordance with the laws of the jurisdiction in which the Institute is incorporated (Delaware, USA). 8.2 Any disputes arising out of or related to these Terms shall be resolved through good faith negotiation between the parties. If negotiation fails, the parties shall submit to mediation before proceeding to arbitration or litigation. 9. Miscellaneous 9.1 These Terms constitute the entire agreement between the parties with respect to the Program and supersede any prior agreements or understandings.
9.2 These Terms may be amended only by a written document signed by both parties. 9.3 Neither party may assign these Terms without the prior written consent of the other party, except that the Institute may assign these Terms to an affiliated entity. 9.4 If any provision of these Terms is held to be invalid or unenforceable, the remaining provisions shall continue in full force and effect. By submitting an application to the Active Inference Institute Partnership Program, the Partner acknowledges that they have read, understood, and agree to be bound by these Terms. Contact: [email protected]
Open Source is key to the of . Open Source Mission, Vision, Values, and Principles The Active Inference Institute The default license information for all Institute materials is CC BY-NC-SA 4.0 and is described below. Check with specific products and collaborators for more information. Open Source CC BY-NC-SA 4.0 Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International This license requires that reusers give credit to the creator. It allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, for noncommercial purposes only. If others modify or adapt the material, they must license the modified material under identical terms. BY: Credit must be given to the creator(s) of the work, the specific people where know & as hosting or publishing entity. The Active Inference Institute NC: Only noncommercial use of your work is permitted. Noncommercial means not primarily intended for or directed towards commercial advantage or monetary compensation. SA: Adaptations must be shared under the same terms (unless otherwise specified and agreed upon by creators). Some of our main are listed in the table below, nested within the . Open Source Repositories Institute Github Active_Inference_Ontology Snapshots of Active Inference Ontology ActiveBlockference repository for integrating with and moreActive Blockference Active Inference cadCAD ActiveInferAnts Active Inference Ant simulations, and much much more ActiveInferenceCategoryThe ory curriculum and materialsCategory Theory ActiveInferenceJournal Primary repository for the , with transcripts from Active Inference Journal Production AEOS Snapshots of Active Entity Ontology for Science (AEOS) Biofirm Active Inference agents for PyMDP Bioregional Modeling CEREBRUM Case-Enabled Reasoning Engine with Bayesian Representations for Unified Modeling GEN24 Generative AI experiments and deployments as part of ( )Active Blockference here GeneralizedNotationNotation Information on Generalized Notation Notation Journal-Utilties Utilities for Active Inference Journal PyDMB Dysfunctional Markov Blanket package to accompany research paper Symposium Synthetic intelligence methods for Applied Active Inference Symposium Textbook Repository for Textbook Group Name URL Description Open Source Repositories • • •
Ecosystem Support Activities at the Institute offer resources and participation opportunities for individuals and organizations. These epistemic and pragmatic services include: Informational Commons & Production Active Inference Journal Largest corpus of Active Inference education materials available to date. Open Source Common Forum. Providing , hosting online forums, , discussion groups, and channels where learners, researchers, and practitioners can connect, ask questions, and share insights. Fostering a community that helps individuals overcome challenges, exchange ideas, connect on collaborations, and receive support from peers and experts. Institute Projects Applied Active Inference Symposium Communications Opportunities to Share and Present Work. Provide myriad opportunities to share and present relevant work on Active Inference, offering opportunities for unique collaborations and new knowledge discovery catalyzed by Active Inference and the consequent amplified leveraging of expertise and practices across disciplines, domains, and paradigms. Infrastructural and Administration directions Infrastructure. Maintaining and developing information systems to support The Institute’s activities, iteratively improving usability and efficacy. Pending funding, working with requirements engineering and user experience professionals to overhaul existing systems. Improve provisioning and access to e.g. , , & materials, Livestreams from , , etc. Implementations of Active Inference Active Inference Ontology Textbook Group Education Production Educational Standards & Qualifications Partnership Managing and growing relationships with , research, application, and service partners.Education Philanthropy Development relationships with potential donors and sponsors, and, pending funding, developing the necessary infrastructure (e.g., accounting, legal, digital affordances, materials) to request and receive donor and sponsor support, and to offer and dispense micro-grants and financial support to researchers. Funding Discovery & Support. Providing a variety of support mechanisms for participants to search for and submit to grant and funding opportunities, as well as assist them in forming partnerships (e.g., with other researchers, companies, and universities). Grants Professionalization. Developing a curriculum of training programs for Officers and Directors of commercial entities and officials of governmental and civil society organizations to enhance their understanding of sentient behavior (as described by Active Inference) and its implications for organizational interactions in the areas of Business, Operations, Legal, Technical, and Social. • ◦ • • • ◦ • ◦ • ◦ • •
Institute Projects See for updated information on active projects. Activities are the primary means of participation with .Institute Projects The Active Inference Institute To date, The Institute has hosted or facilitated the development of hundreds of licensed products which serve various functions in including Awareness, Education, Commons, Support, and Governance. Open Source The Active Inference Ecosystem Project Rhythm Through and Prepare Measure The Institute implements a unique "Prepare and Measure" system that structures project work through alternating phases of preparation and measurement. To complete a , participants propose a phase of activity — their “packed backpack” and intention for developing artifacts, research, or create educational materials while receiving ongoing feedback. Project ~ Preparation This is followed by making a , where the participant documents their reports and reflections. Following the measurement, next steps are explored. This rhythmic approach creates natural checkpoints for reflection while maintaining momentum Project ~ Measurement Benefits and Implementation The prepare-measure cycle embodies active inference principles by balancing exploration with evaluation (see ). Rather than following rigid schedules or purely passive learning, participants actively sample their environment through concrete project work, while regular measurements provide the feedback needed for learning and course correction. Physics course This system helps cultivate a culture of active sensemaking, where putting work out for feedback is encouraged over passive consumption. The flexibility of this approach allows it to scale from individual contributors to large collaborative projects, while maintaining rigor through consistent documentation and assessment. Towards a Project FrameworkSystems Approach We make ongoing incremental updates to the approach taken across . Current thinking on this is considering updates in the area of: Institute Projects Project Structure Stronger connection with or .EduActive (Education) ReInference (Research) Organizational Units Each project will have a clearer standardized public profile featuring: Clear mission statement and objectives Timeline with key milestones and deadlines Contribution pathways and skill requirements Active measurement cycles and preparation phases Project Management Approach Time Management Dedicated work blocks outside of meetings Regular preparation and measurement cycles • • ◦ ◦ ◦ ◦ • •
Minimal reliance on email/Discord for core work Task Organization Public task tracker with clear ownership Regular progress updates and milestone reviews Documentation of both successes and learning opportunities Integration with prepare/measure cycles The Institute aims to implement these projects using active inference principles, ensuring each initiative contributes to our mission of making active inference more accessible, rigorous, and applicable while serving our growing global community. https://projects.activeinference.institute/ • • • • •
AICACP AI Capabilities & Alignment Consensus Project We are happy to share that in June 2025, the Survival and Flourishing Fund has awarded a $270,000 grant to the to support work on the AI Capabilities & Alignment Consensus Project ( ).Active Inference Institute AICACP AICACP is a multi-year initiative designed to reshape the conversation around AI capabilities, alignment, and regulation. By combining high-impact journal collections, in-person discussion-oriented workshops, and academic media content for public outreach, the project aims to bridge the divide between AI “doomers” and “accelerationists” through deeply exploring the meanings of “world models” and “agency,” and what these concepts mean for AI development. (at Allen Discovery Center at Tufts University) is the creator, principal investigator, and organizing editor for this project. This grant supports an expanded team to assist with efforts to manage the special issues and workshops. Adam Safron For those interested in learning more, please check out this document: AI Capabilities and Alignment Consensus Project - Google Docs Please reach out to with any thoughts or [email protected]
Active Blockference Much information is contained in the . Active Blockference project documentation Active Blockference is an open-source project developed by the that aims to create a comprehensive cognitive modeling framework for complex systems. The project combines two powerful technologies: (Complex Adaptive Dynamics Computer-Aided Design) and . The Active Inference Institute cadCAD Implementations of Active Inference Project Overview The primary goal of Active Blockference is to develop a simulation environment that can model the cognitive processes and goal-directed behavior of agents within various complex systems across . This framework is designed to: Domains of Application Facilitate rigorous analysis of multi-agent systems and their emergent behaviors Serve as a sandbox for exploring cognitive, micro-economic, behavioral, and decision-making processes Enable cognitive audits of protocols and systems across multiple domains Key Components Open-source package: Integrates cadCAD and Active Inference implementations for theoretical and applied studies Multi-agent simulation: Expands from single-agent to multi-agent models to explore cognition and behavior in various settings Educational resources: Develops materials to onboard new users to the Active Blockference community Rationale for Combining cadCAD and Active Inference The integration of cadCAD with an Active Inference kernel provides a powerful synergy for modeling complex systems: cadCAD: Offers a robust framework for simulating complex adaptive systems, allowing for the modeling of multi-agent interactions and system-level dynamics. Active Inference: Provides a principled approach to modeling goal-directed behavior and decision-making processes of individual agents. By combining these technologies, Active Blockference can model both the macro-level system dynamics and the micro-level cognitive processes of agents within those systems.Project Status and Development As of October 2024, Active Blockference is in active development. The project welcomes participants from various backgrounds to contribute to the growing codebase on .GitHub Get Involved Interested individuals can participate in Active Blockference through: Joining ongoing discussions on the Active Inference Institute's Discord server Contributing to the for asynchronous collaborationCoda document Exploring and contributing to the GitHub repository By developing this cognitive layer for complex systems modeling, Active Blockference aims to enhance our understanding of multi-agent dynamics and decision-making processes across a wide range of applications. 1. 2. 3. • • • 1. 2. 1. 2. 3.
AII ~ Active Inference Ontology https://coda.io/d/_djD38E5fJk_
Active Entity Ontology for Science (AEOS) represents a key framework developed by a 2022 , created to bridge centralized and decentralized approaches to scientific organization. Active Entity Ontology for Science (AEOS) Institute Projects The work was published as “An Active Inference Ontology for Decentralized Science: from Situated Sensemaking to the Epistemic Commons” ( ), and is also available on and in an . link Github interactive Coda format Here are the key aspects of AEOS: Core Components Framework Structure Uses Active Inference principles to model different forms of scientific activity as a collective cognitive process occurring in a niche. Provides a composable and versionable system for modeling various scientific systems Integrates BOLTS perspective (Business, Operations, Legal, Technical, Social) for comprehensive analysis Key Functions Maps relationships between different scientific entities and processes using Active Inference entity partitioning Enables modeling of both traditional institutional science (CeSci) and decentralized science (DeSci) approaches Facilitates bottom-up sensemaking while maintaining systematic organization Implementation Goals Scientific Organization Supports emergence of epistemic communities through organic collaboration Enables transparent resource allocation and knowledge sharing Provides tools and analytic methods for decentralized scientific governance Practical Applications • • • • • • • • •
Guides development of tools for scientific collaboration, Web3 or otherwise. Helps structure new kinds of organizations for research purposes Supports integration of blockchain and other technologies into scientific workflows The AEOS serves as a bridge or blanket between: In theory and in terms of generalities: Active Inference Ontology In practice: Existing and emerging decentralized approaches, providing a structured way to understand and implement new forms of scientific organization while maintaining rigor and effectiveness in and traditional scientific settings.DeSci • • • • •
Applied Active Inference Symposium The Applied Active Inference Symposium highlights ongoing work related to Active Inference across domains. See the live 2025 Symposium Program Online, November 12-14, 2025 ( ).See the program The proceedings of past are available: Applied Active Inference Symposium 1st in 2021 — Karl Friston ( , video part , , )transcript 1 2 3 2nd in 2022 — Robotics 3rd in 2023 — Enacting Ecosystems of Shared Intelligence — , video , , , of videos4th in 2024 Program part 1 part 2 part 3 playlist — “Industry” November 12-14: 5th in 2025 See the program More See Sponsorship at the Symposium Any other ideas? Please .email us https://symposium.activeinference.institute/ • • • • • • •
Call for Presenters Information on being a presenter for the Applied Active Inference Symposium We are excited to invite researchers and practitioners to submit presentations for the upcoming 5th Applied Active Inference Symposium 2025, to be held on 12-14 November, 2025. This online Symposium will focus on exploring the frontiers of Active Inference with a special focus on its applications in industry. As with last year’s Symposium, the keynote address and panel will feature Karl Friston. All information: symposium.activeinference.institute/ to submit your presentation proposal, and/or read on for more details. Please feel free to share this invite with anyone else who you think would be a great addition to the Symposium. Complete this form Symposium Themes and Topic We are welcoming presentations and workshop submissions that align with the Symposium themes and tracks: Overall Theme Applying Active Inference in Industry: In 2025 we are seeing increased application of Active Inference across industry domains. What are the success stories? What was learned along the way? What challenges remain? What kind of applied research would accelerate applications? Topics Open Science & Tool Development: Increasing the function and accessibility of Active Inference software across languages. Focused tracks and workshops on practical aspects of Active Inference. Learning & Education: Share initiatives and experiences in teaching and spreading knowledge about Active Inference. Connecting the global Active Inference community with each other, resources, and research/education opportunities. Interactive Aspects & Specific Affordances Submissions of the following types are welcomed: Presentations: Pre-recorded video presentations of 15-60 minutes, or livestream slots of 30-90 minutes for panels or group discussions. Workshops, Tutorials, and Hackathons: Hands-on sessions of 60-180 minutes, that engage participants with Active Inference methods and provide open source materials/outputs. Accessibility & Participation We are committed to making this online symposium accessible to all. Participant registration: The Symposium will be online, with no financial barrier to participation. We will provide pathways for participants to foster deeper involvement with in the ecosystem. ongoing research and projects • ◦ • ◦ ◦ • • •
Collaboration with Active Inference Journal: We will publish the Abstracts of presentations ( ) to give visibility to proceedings and enhance availability of the materialssimilar to last year Translation: We will translate materials into different languages where possible, to provide content that reaches a broader audience (the organizing team will support presenters on this). Multimodal Content: audio, visual, interactive/real-time simulations or other media that help communicate complex ideas. The content will remain posted on our YouTube channel for asynchronous viewing following the symposium. Important Dates Presenter Submission Deadline: October 31, 2025 Symposium Dates: November 12-14, 2025 How to Submit Please submit your abstracts and proposals via . this form Symposium co-organizers: Alexander Ororbia, Alexandra Mikhailova, Andrew Pashea, Bradly Alicea, Christian Martens, Cory Slater, Daniel Friedman, Marc Broberg, Maria Garcia, Maria Luiza Iennaco, PabloFM, Rorik Smith, Sylvia Zhang, Bleu Knight, Zohreh Rahmannejad, Alex Vyatkin Contact Information: For any inquiries, please reach out to . We are also at this time. Information on participant registration will be closer to the date. [email protected] seeking sponsors available here • • • • •
Sponsorship at the Symposium Support the Applied Active Inference Symposium with financial or in-kind donations More information on in 2025 coming soon. Sponsorship at the Symposium Why Sponsor the Applied Active Inference Symposium? Active Inference represents a cutting-edge approach to artificial intelligence and cognitive science, especially in the modern and open science setting. By supporting the , sponsors are contributing to: Applied Active Inference Symposium Scientific Innovation: Active Inference offers a rigorous, first-principles approach to AI and cognitive science, promising significant computational benefits and innovative system designs. Open Science & Collaboration: The Active Inference Institute is a 501(c)(3) non-profit organization that promotes open-source practices, global participation, and collaborative learning, accelerating progress through shared knowledge. This Symposium is one of the few inclusive opportunities available at this time to participate in hands-on workshops and connect with other people applying Active Inference. Broad Impact: Supporting Active Inference research and applications, can contribute to addressing global challenges, fostering interdisciplinary applications, and promoting ethical transparency in AI development. Strategic Advantage: Sponsorships provide visibility to talent, aligns with funding requirements, and facilitates valuable partnerships, positioning sponsors at the forefront of AI innovation. How Organizations Can Contribute at desired sponsorship level (or suggest something else?)Provide financial support Offer in-kind support (e.g., services, products) 1 Listed in Symposium Program and materials. Acknowledged in Livestreams • • 2 As above plus: Logo included in the Adventure space Video Presentation to be played on livestream (1-5 minutes pre-recorded video) We can make introductions with you and: Symposium Presenters, Co-Organizers, Participants Institute Research Fellows, Officers, Board of Directors, Scientific Advisory Board, Interns. • • • ◦ ◦ 3 As above plus: Consideration of a session/speaker, or theme, to be included the program. Arrangement before or after the Symposium, of a private session on theme of your choice. Optional: Custom commissioned hand-drawn Active Inference art • • • Sponsorship level Benefits Sponsorship tiers • • • • • •
Optional: Contribute additional resources (e.g., presenters, materials) Impact of Your Support Your sponsorship directly contributes to: Advancing open-source Active Inference research and applications The sustainability of the Active Inference Institute, and success of the . Institute & Ecosystem Promoting innovation in AI and cognitive science Supporting the coming generations of Active Inference researchers and practitioners Join us in shaping the future of intelligent systems through Active Inference! Email with any inquiries or questions. Provide financial support . [email protected] directly at this link • • • • •
CogNarr (Cognitive Narrative) Ecosystem: Facilitating Group Cognition at Scale An ongoing Active Inference Institute project facilitated by John Boik, PhD Introduction Human groups of all sizes and kinds engage in deliberation, problem solving, strategizing, decision making, and more generally, cognition. Cognition in the group setting serves a similar purpose to cognition in individual humans. From an active inference perspective, that purpose is to achieve and maintain, with high certainty, those preferred conditions that promote health and wellbeing. As we know from common experience, the quality of group cognition can range from functional to dysfunctional, productive to unproductive, and thoughtful to superficial. As such, the quality of a group’s cognitive process can either lead the group toward or away from health, wellbeing, security, and goal achievement. Achieving functional, productive, and thoughtful cognition is especially difficult in the large group setting. The small-group setting often involves face-to-face dialogue, which can support rich and dynamic interactions that allow all voices to be heard. But such interactions are more difficult to achieve in the large-group setting, which typically requires some form of online communication. New approaches are needed to facilitate the kind of rich communication and information processing that are required for effective, functional, productive cognition in the online setting, especially for groups characterized by hundreds, thousands, or millions of participants who wish to share potentially complex, nuanced, and dynamic perspectives. The incipient CogNarr (Cognitive Narrative) Ecosystem is intended to facilitate functional cognition in the large-group setting. A key perspective is to view a group as an organism that uses some form of cognitive architecture to sense the world, process information, remember, learn, predict, make decisions, and adapt to changing conditions. The CogNarr ecosystem is designed to serve as a component of that architecture. The CogNarr project at the Active Inference Institute is intended to bring CogNarr to life. CogNarr is potentially a massive project, involving many topics, tasks, needs, researchers, staff, volunteers, and so on. If you have questions or suggestions, want to learn more, or want to help, please write to me via “[email protected]” with “[COGNARR]” in the subject line and I will respond. You can also join our biweekly meetings, every other Wednesday in the morning US Mountain Time. Meeting events are listed on the . Active Inference Institute calendar John Boik, Research (May 2024 - )Fellow : 0000-00031289-7997 ORCID
Social Science course In 2023, hosted a course titled “Constructing cultural landscapes: Active Inference for the Social Sciences”, organized by Avel Guenin-Carlut, Ben White, Mahault Albarracin, Lorena Sganzerla and Daniel Friedman. The twelve-week course introduced participants to conceptual tools to understand the relation between social and cognitive sciences. Recordings of the talks, and more information are available at the . The Active Inference Institute public link https://coda.io/@active-inference-institute/active-inference-social-science-aii-2023
Active Inference for the Social Sciences ~ AII 2023 https://coda.io/d/_dVVFg3pdihg
Physics course In 2023, the hosted a course titled “Physics as Information Processing”, taught by Chris Fields. The six-week course introduces participants to formal Quantum Information Theory as a concept and tool for understanding physical interaction as communication. Recorded lectures and course materials are available at the , and of all videos. The Active Inference Institute public link here is the YouTube playlist https://coda.io/@active-inference-institute/fields-physics-2023
Physics as Information Processing ~ Chris Fields ~ AII 2023 https://coda.io/d/_dhwI_xbGzuD
Educational Standards & Qualifications Engagement Pathways at the Active Inference Institute Learning Paths, Modes, and Seasons Browser: Discovers active inference through key word searches, algorithmic recommendations, bibliographic tracing, or word-of-mouth, engaging with occasional content such as Production Regular Consumer: Follows dedicated channels and educational content about active inference and related topics Active Learner: Independently seeks out technical materials, research papers, and in-depth resources. Taking notes, making personal synthesis artifacts, engaging in solo or group . Ecosystem Projects Institute : Participates in with a defined role (e.g. facilitator, .0 video preparation collaborator). Volunteer Institute Projects Textbook Group Production : Engages in focused project work while receiving mentored guidance and education.Internship : Dedicated, possibly funded, focus on larger scale initiatives. Fellows , , Scientific Advisory Board Board of Directors Officers The Active Inference Institute aims to make these learning pathways accessible to a global audience, meeting learners wherever they are in their journey. Through a multi-tiered approach, we aim to create entry points and paths for everyone from casual browsers to researchers and practitioners. 2025 Learning Initiatives The Institute is enthusiastically preparing for expanded offerings in 2025. We recognize the growing interest in active inference across disciplines and are look to develop new s and to support learning needs and our . Education Partnership Institute Programs Mission, Vision, Values, and Principles We will focus on building collaborative learning environments that bridge theoretical foundations with practical applications, while fostering a meaningful and productive community of practice that spans academic, industry, and independent researchers. • • • • • • •
FarmWorks FarmWorks: Decentralized AI Agents for Personalized Solutions. https://zenodo.org/records/13754586 FarmWorks is the name of a project to develop a platform for human-AI interaction in agriculture, enabling personalized, farmscale solutions that resist power concentrations associated with centralized AI systems. A submitted in September 2024. Grants Work continues in the page at .RxInfer.jl Learning Group this link
Fundamentals of Active Inference We worked with during 2023-2024 to support development of a textbook (expected public release in 2025). We look to share more information about future as we can. Sanjeev Namjoshi Fundamentals of Active Inference For more on the book & Sanjeev’s project, see: Sanjeev Namjoshi ~ Active InfeSanjeev Namjoshi ~ Active Infe……Sanjeev Namjoshi ~ Active InfeSanjeev Namjoshi ~ Active Infe…… Sanjeev Namjoshi ~ ~ Education, Expectation-Maximisation, Evolution Active Inference Insights 018 https://www.youtube.com/watch?v=sAwPXw-WNg4 The Hidden Math Behind All LiviThe Hidden Math Behind All Livi……The Hidden Math Behind All LiviThe Hidden Math Behind All Livi…… The Hidden Math Behind All Living Systems (on )Machine Learning Street Talk https://youtu.be/hf18w6CuY8o?
Generalized Notation Notation is a text-based language designed to standardize the representation and communication of generative models. It aims to enhance clarity, reproducibility, and interoperability in the field of Active Inference and cognitive modeling. Generalized Notation Notation Active Inference code link: Open Source https://github.com/ActiveInferenceInstitute/GeneralizedNotationNotation Original publication: Smékal, J., & Friedman, D. A. (2023). Generalized Notation Notation for Active Inference Models. Active Inference Journal. https://doi.org/10.5281/zenodo.7803328 GNN provides a structured and standardized way to describe complex cognitive models. It is designed to be: Human-readable: Easy to understand and use for researchers from diverse backgrounds 🤖 Machine-parsable: Can be processed by software tools for analysis, visualization, and code generation 🔄 Interoperable: Facilitates the exchange and reuse of models across different platforms and research groups 🔬 Reproducible: Enables precise replication of model specifications GNN addresses the challenge of communicating Active Inference models, which are often described using a mix of natural language, mathematical equations, diagrams, and code. By offering a unified notation, GNN aims to streamline collaboration, improve model understanding, and accelerate research. Generalized Notation Notation: From Plaintext to Triple Play Active InferAnt Stream #014.1 Active InferAnt Active InferAnt ……Active InferAnt Active InferAnt …… GNN for Generative Model Supply Chains: A Golden Spike Moment for Multiagent Trajectory Planning with RxInfer.jl Active InferAnt Stream #014.2 Active InferAnt Active InferAnt ……Active InferAnt Active InferAnt …… The Sound of Uncertainty: Auditory Rendering of Generative Models in the Field of Streams Active InferAnt Stream #014.3 Active InferAnt Active InferAnt ……Active InferAnt Active InferAnt …… • • • •
Knowledge Engineering As of the end of 2022, is a ongoing project ( ) at the that analyzes the literature related to Active Inference and Free Energy Principle, published as: The Free Energy Principle & Active Inference: a Systematic Literature Analysis Knowledge Engineering code repository The Active Inference Institute https://zenodo.org/record/7449368 We performed a literature analysis of publications in scientific literature using the term “Free Energy Principle” or “Active Inference”, with an emphasis on works written by Karl J Friston. For a subset of papers with accessible full texts, we performed manual annotation (related to structural, visual, and mathematical features) and automated analyses (related to the terms in the Active Inference Institute’s Active Inference Ontology). The initial analysis here, at the scale of thousands of citations and hundreds of annotated papers, is presented as a first step towards the development of systems which could: Encompass increased scope of relevant works, including non-textual Integrate multiple forms of annotation and participation Facilitate integration of manual and artificial contributions Feature richer interfaces for use in learning & research Address field-specific local questions and provide transferable approaches Speak to broader questions in the history and philosophy of science The paper is pre-printed at: https://zenodo.org/record/7449368 This project has an and a .interactive Coda site Github repository The initial work was done in 2022 and we look forward to revisiting and improving this work in the years to come. • • • • • •
Knowledge Engineering Frontend Want to print your doc? This is not the way. Want to print your doc? This is not the way.
political/institutional changes. Below are examples of important epochs but important is the ability to identify lower level shifts. Empirical Predictions
Developmental Trajectory: Children should show increasing α values (cultural precision) with age, corresponding to greater moral conformity and symbolic capacity Neural Correlates: fMRI studies should reveal distinct activation patterns during transcendental inference versus ordinary perspective-taking Cross-Cultural Variations: Cultures should vary in hierarchical depth and precision coupling strength between levels Digital Communication: Anonymous platforms without dyadic buffering will naturally produce more extreme moral typing and polarization Shaggy’s own Mythic Journey/ Road Map Phase 1 (Months 1-6): Low-dependency prototypes, simulation development, narrative vignettes Phase 2 (Months 7-18): Lightweight collaborations, mid-scale simulations, behavioral pilots Phase 3 (Months 19-36): Resource-intensive studies, large-scale simulations, organizational pilots Implications for Artificial General Intelligence Hierarchical Organization AGI systems must develop explicit mechanisms for traversing hierarchical scales, integrating multiple Markov blankets, and selecting appropriate levels of abstraction for different contexts. Cultural Alignment The precision parameter α that enables cultural-level constraints to guide individual behavior may be crucial for AGI systems operating across personal, social, and institutional scales. Transcendental Model Selection Current AI architectures lack the fundamental capacity for transcendental model selection, representing a core barrier to achieving human-level general intelligence. Epistemic Depth True alignment requires moving beyond black box architectures to systems with explicit epistemic depth - a capacity that is feasible to implement computationally. About Shaggy 1. 2. 3. 4.
Email: [email protected] Professional Background: Senior Product Manager, focus in Financial Services and Data/ML based systems. Writing: Please follow/subscribe here: https://shaggy.substack.com/ Interested in: Collaborations on empirical validation, category theory formalization, and AI/AGI applications
Production We produce educational content in the form of on , , and replication across other platforms. Production YouTube Podcasts on Podbean We have multiple kinds of formats for the content, and we add new ones as availability/capacity arises. Some current formats and area of focus for are:Production (focused on specific papers, 58 papers discussed from 2020 through 2025). Livestream 100+ s, highlighting a wide range of work in and related fields. GuestStream Active Inference (computational models), (social and organizational topics), (formalisms and math), streams in 2024, (art and aesthetics), (coding, modeling, synthetic intelligence), (organizational updates), Courses ( , ), s, and more ModelStream OrgStream MathStream MathArt ArtStreams InferAnt Streams Roundtables Social Science course Physics course Textbook Group One of the highlights of 2024 was Active Inference Insights ( , ), hosted by Darius Parvizi-Wayne.Podbean YouTube Active Inference Insights is a podcast which introduces listeners to the wondrous land of Active Inference. Guided by our diverse array of guests, from physicists and mathematicians to cognitive scientists and philosophers, you will not only learn about cutting-edge theory, but also come to see the world in a whole new way, in which all things can be tied together by a single imperative: the minimisation of free energy. To date, we have released over 500 videos, all available .Open Source xWe aim for all videos are productions, to be transcribed, analyzed, translated, and published by the .Active Inference Journal https://video.activeinference.institute/ • • • • ◦
Videos and Podcasts Want to print your doc? This is not the way. Want to print your doc? This is not the way.
Research Research activities and resources The page hosts information on research projects such as , , and other . Research CogNarr (Cognitive Narrative) Ecosystem: Facilitating Group Cognition at Scale Wave Hypothesis Research Resources Some research and products of Active Inference Institute and participants are below: 2024 Karl J. Friston, Thomas Parr, Conor Heins, Axel Constant, Daniel Friedman, Takuya Isomura, Chris Fields, Tim Verbelen, Maxwell Ramstead, John Clippinger, Christopher D. Frith, Federated inference and belief sharing, Neuroscience & Biobehavioral Reviews, Volume 156, 2024, 105500, ISSN 0149-7634, https://doi.org/10.1016/j.neubiorev.2023.105500. https://www.sciencedirect.com/science/article/pii/S0149763423004694 From the project: Broken link “ ”. June 3, 2024 (Collaboration) Aligning Active Inference Ontology to SUMO https://zenodo.org/records/11463326 “ ”. April 24, 2024. Aligning Spatial Web Terms to SUMO https://zenodo.org/records/11062810 Albarracin, M.; Pitliya, R.J.; St. Clere Smithe, T.; Friedman, D.A.; Friston, K.; Ramstead, M.J.D. Shared Protentions in Multi-Agent Active Inference. Entropy 2024, 26, 303. https://doi.org/10.3390/e26040303 Friedman, D. A., & Tickles, D. (2024). Four-fold Fields of Quantum Dreams (Version v1). https://doi.org/10.5281/zenodo.10798145 Dean Tickles, Daniel Ari Friedman, Why Paleolithic Rockstars were both enigmatic and sporadic: A comment on: ‘Snakes and Ladders’ in paleoanthropology: From cognitive surprise to skillfulness a million years ago, Physics of Life Reviews, Volume 50, 2024, Pages 4-6, ISSN 1571-0645, https://doi.org/10.1016/j.plrev.2024.04.010. https://www.sciencedirect.com/science/article/pii/S1571064524000447 2023: August 2023 publication from the Institute: " ".The Active Inference Institute and Active Inference Ecosystem " ", Francesco Balzan, John Campbell, Karl Friston, Maxwell James Ramstead, Daniel Friedman, Axel Constant, 2023. Distributed Science - The Scientific Process as Multi-Scale Active Inference " ", Karl Friston, Daniel Ari Friedman, Axel Constant, V. Bleu Knight, Thomas Parr, John O. Campbell, Entropy, 2023. A variational synthesis of evolutionary and developmental dynamics " ", Daniel Friedman & Jakub Smékal, 2023. Generative Research Teams: Active Inference Compositions For Research and Meta-Science " ", Jakub Smékal & Daniel Friedman, 2023.Generalized Notation Notation for Active Inference Models • • ◦ ◦ • • • • • • • •
" ", Shohei Wakayama and Nisar Ahmed, 2023. Active Inference for Autonomous Decision-Making with Contextual Multi-Armed Bandits " ", Eric Saund and Daniel Friedman, Cognitive Systems Research, 2023. A single-pheromone model accounts for empirical patterns of ant colony foraging previously modeled using two pheromones 2022: " ", Virginia Bleu Knight; RJ Cordes, Daniel Friedman. 2022. The Free Energy Principle & Active Inference: a Systematic Literature Analysis , Jakub Smékal, Daniel Friedman. 2022. Catechism for: "Towards Active Diffusion: A Tale of Multiple (den)Cities" " ", Sean O'Connor and Daniel Friedman. 2022. Predictive Processing Interpretation of the Mirror Test and Implications of a Reflection Prediction for Human Cognition " ", Jakub Smékal, Arhan Choudhury, Amit Kumar Singh, Shady El Damaty & Daniel Ari Friedman, from IWAI 2022 (International Workshop on Active Inference). Active Blockference: cadCAD with Active Inference for Cognitive Systems Modeling " " & the An Active Inference Ontology for Decentralized Science: from Situated Sensemaking to the Epistemic Commons Active Entity Ontology for Science 2021: Transcript of: Karl Friston, 1st Applied Active Inference Symposium https://zenodo.org/record/5797041 “ ”Active Inference in Modeling Conflict “ ”.Narrative Information Management Cognitio 2021 https://www.cognitio2021.com/ "Thinking like a State: Active inference and the deep roots of complex societies", Bleu Knight, https://osf.io/dxnzt/ “Intelligence without creativity: can Active Inference ground our understanding of life, cognition and society", Avel Guénin-Carlut "Evolution of Latent Model for Collective Cognition", Amit Singh "Active InferAnts: The basis for an active inference framework for ant colony behavior", paper www.frontiersin.org/articles/10.3389/fnbeh.2021.647732/ 2nd International Workshop on Active Inference "Active Inference & Behavior Engineering for Teams", poster on the 2020 paper “Active Inference & Behavior Engineering for Teams” https://zenodo.org/record/4021163 • • • • • • • • • • • ◦ ◦ ◦ • •
"ColIective Intelligence as Latent Imagination", Amit Singh, International Conference on Cognitive Modeling (ICCM'21), 73 "Context Switching in Machine Minds", Amit Singh, Society of Mathematical Psychology 2021, https://youtu.be/4647USeygmg 2020: " " Alex Vyatkin, Ivan Metelkin, Alexandra Mikhailova, RJ Cordes, Daniel Friedman Active Inference & Behavior Engineering for Teams • • •
Research Resources Research Projects & Resources See for resources on: “ " “ " Control Flow Control flow in active inference systems Part I: Classical and quantum formulations of active inference Control flow in active inference systems Part II: Tensor networks as general models of control flow See for resources on: " " (2023) Variational Evolution A Variational Synthesis of Evolutionary and Developmental Dynamics
Control Flow Resources for: Control flow in active inference systems Supplementary Materials for: Chris Fields et al., "Control flow in active inference systems Part I: Classical and quantum formulations of active inference," in IEEE Transactions on Molecular, Biological and Multi-Scale Communications, doi: 10.1109/TMBMC.2023.3272150. (2023)https://ieeexplore.ieee.org/document/10113698 Chris Fields et al., "Control flow in active inference systems Part II: Tensor networks as general models of control flow," in IEEE Transactions on Molecular, Biological and Multi-Scale Communications, doi: 10.1109/TMBMC.2023.3272158. (2023)https://ieeexplore.ieee.org/document/10113744 Control_Flow_Supplementary-Information-Table-1. pdf
Start Onboarding Curriculum and Learning Paths for Active Inference, across languages and backgrounds is project, started in December 2024, introduced in video Active “Symbol’s Greetings: Onboarding to Active Inference across backgrounds & languages”, and continued in Active “START/HERE: A Map & What Might Happen Next” https://github.com/ActiveInferenceInstitute/Start Open Source @Software Development Production InferAnt Stream 008.1 InferAnt Stream 015.1 Start here: https://github.com/ActiveInferenceInstitute/Start/blob/main/here.md START (Scalable, Tailored Active‑inference Research & Training) (Scalable, Tailored Active‑inference Research & Training) is a modular pipeline that generates high‑quality educational materials on Active Inference and the Free Energy Principle, tailored to professional domains and individual learners. It integrates live web research via Perplexity and advanced LLMs via OpenRouter to produce evidence‑based, professionally structured content, with visual analytics and multilingual localization. The system emphasizes real data, reproducibility, and quality assurance through tests and linting, and it ships with comprehensive documentation and an interactive CLI experience. Start Live docs site: https://activeinferenceinstitute.github.io/Start/ Repository: https://github.com/ActiveInferenceInstitute/Start What it produces Domain research: 3,000–5,000 word analyses of professional fields (e.g., neuroscience, AI, healthcare), grounded in current sources (Perplexity). Audience/entity research: 5,000–8,000 word learner profiles with actionable learning strategies. Curricula: 40–60 hour, module‑based programs with structured sections, objectives, and assessments. Visualizations: PNG charts of curriculum metrics and Mermaid flow diagrams for structure and learning pathways. Translations: Native‑quality, culturally adapted outputs in 11+ languages (Chinese, Spanish, Arabic, Hindi, French, Japanese, Russian, Swahili, Tagalog, and more). How it works (pipeline) Inputs: YAML configurations define domains, entities/audiences, and target languages. Research → Curriculum Generation → Visualization → Translation → Outputs under data/. Prompts are curated for domain analysis, curriculum generation, section authoring, and translation, ensuring consistent structure and completeness. Key technologies and quality model • ◦ ◦ ◦ ◦ ◦ • ◦ ◦ ◦ •
Research: Perplexity API for current, multi‑perspective domain insights. Content: OpenRouter LLMs for robust, structured curricula and translations. Quality: pytest coverage, ruff linting, black formatting, type hints, and CI‑ready workflows. Architecture: Clear separation across src/ (core logic), learning/ (scripts), data/ (artifacts and configs), docs/ (user/developer guides), and tests/. Running the system Interactive terminal: run.sh provides an end‑to‑end guided experience from research through translation. Documentation workflows: run_docs.sh supports serve, build, and deploy to GitHub Pages. Python environment uses uv for reproducible setup; keys for Perplexity and OpenRouter are required. Outputs accumulate incrementally in data/domain_research/, data/audience_research/, data/written_curriculums/, data/visualizations/, and data/translated_curriculums/. Who it is for Educators and program designers building university or professional development courses. Researchers seeking personalized, evidence‑based learning roadmaps and current domain syntheses. Organizations adopting Active Inference frameworks for training, strategy, and decision support. ◦ ◦ ◦ ◦ • ◦ ◦ ◦ ◦ • ◦ ◦ ◦
Systems Approach A modern third-generation systems approach is essential for managing today’s complex adaptive systems. This approach transcends traditional linear models, embracing continuous evolution, interactivity across scales, and the dynamic, constructivist perspective where systems actively reshape themselves in response to changing conditions. Combining active inference principles with a systems approach provides a pathway to designing resilient, self-organizing systems that are responsive to diverse environments. Evolving systems approach Classical systems approach focused on structured relationships within defined boundaries, effective for static and predictable systems. However, with the growing complexity of modern systems like cyber-physical networks and adaptive ecosystems — a more dynamic approach is needed. Third-generation systems approach builds on these foundations to handle layered, openended development and real-time adaptability. A systems approach today emphasizes flexible frameworks that enable continuous learning and adaptation across varied contexts, positioning systems to better navigate and anticipate change. provides a foundation for adaptive systems by defining systems as (nested, interacting) agents. This aligns with a systems approach by enabling systems to organize themselves dynamically in response to environmental changes. Key characteristics of this model include: Active Inference Continuous adaptation: Systems evolve iteratively, continuously integrating new information rather than following a rigid lifecycle. Anticipatory action: Systems use predictive models to take preemptive actions, reducing disruptions before they occur. Interactions within and across scales: Systems function cohesively across micro and macro levels, preserving stability and coherence regardless of scale. These qualities position active inference as a crucial tool for developing systems that are resilient, responsive, and able to selfcorrect in changing environments. Systems as Constructors A modern systems approach treats systems as constructors, entities that not only adapt but also actively build and modify their environments. Systems continuously refine their models based on feedback, supporting informed decision-making and efficient resource allocation. This constructivist perspective emphasizes: Dynamic modeling: Systems adjust internal models based on ongoing sensory input, which helps them make real-time, informed decisions. Open-ended development: Systems remain open to generating novel solutions and can reorganize to meet emerging challenges, enhancing robustness and resilience. This approach is especially applicable in fields requiring systems to maintain functionality amid complex, changing conditions, like AI and distributed cybersecurity. Systems built on this constructivist foundation are inherently flexible, robust, and capable of evolving independently. Collaborative ecosystems and community-driven development The active inference framework is built within a collaborative, development model ( ) that aligns with third-generation systems approach. This community-driven ecosystem encourages knowledge sharing and real-time updates, ensuring that systems evolve alongside technological and societal needs. Collaborative development fosters rapid adaptation and inclusivity, allowing systems to better meet diverse user requirements. An open-source model also supports common standards, providing a strong foundation for sustainable and accessible system design across interconnected fields. Open Source The Active Inference Ecosystem • • • • •
Adaptive systems approach Integrating active inference within a modern systems approach offers a robust, adaptive model for managing complexity. This combination encourages resilient, coherent, and evolving systems that can operate autonomously and flexibly across scales. By embracing dynamic modeling, constructivist principles, and active inference, this approach provides a foundation for systems that not only withstand change but actively respond to it, supporting a broad range of applications in both technical and social domains. Reference “ ” by Anatoly LevenchukToward an Ontology for Third Generation Systems Thinking
Textbook Group registers you to participate in the Active Inference on the book “Active Inference: The Free Energy Principle in Mind, Brain, and Behavior” By Thomas Parr, Giovanni Pezzulo and Karl J. Friston ( ). This form Textbook Group 2022 Enter the , with pages for questions, resources, past meetings, and more. Textbook Group’s interactive document here The Textbook Group is about learning Active Inference in an open science setting. All backgrounds and level/type of familiarity with Active Inference are welcomed and encouraged! See the completed playlists of , , , , , , . Cohort 1 Cohort 2 Cohort 3 Cohort 4 Cohort 5 Cohort 6 Cohort 7 The main focus of the Textbook Group is to help you learn Active Inference. We’re expecting lots of different backgrounds, but our goal is to meet you where you’re at to help you understand the textbook. There will be no wrong answers or incorrect . You’ll be encouraged to make connections with what’s familiar and authentic to you. The textbook includes connections to biology, psychology, physics, mathematics, computer programming, etc. questions Group facilitators and participants will be actively maintaining and updating , which we use as a . Facilitators will be available to answer questions and connect you with other participants to compliment and reinforce learning. If you’re interested in facilitating please indicate in the form above. the Coda shared epistemic niche The Institute is exploring exciting research and applications of Active Inference. The Textbook Group is a great place to learn more about this. If you’re interested to learn more, please complete the form below or .email us To register, !complete the form below
What is your full name, or what do you prefer to be called? * What is the best email address for sending you emails & calendar events? * I understand that all work done by everyone participating in this Textbook Group will be licensed under the Creative Commons CC BY unless specifically otherwise stated. * More information on Creative Commons: https://creativecommons.org/licenses/ Yes Not sure, I still have questions about this What is your full name, or what do you prefer to be called? * What is the best email address for sending you emails & calendar events? * I understand that all work done by everyone participating in this Textbook Group will be licensed under the Creative Commons CC BY unless specifically otherwise stated. * More information on Creative Commons: https://creativecommons.org/licenses/ Yes Not sure, I still have questions about this This Onboarding will be an email from , containing a link to that will be the single source of truth for this Textbook Group cohort. [email protected] the Coda document will have supporting material and learning practices to understand each chapter, information about how to contribute, as well as information on the calendar of the Textbook Group. Everyone will have an individual learning space, so you can easily share your work, collaborate with others, and get help. The Coda • •
Tech Tree A for us is creating an Active Inference (a “ ”) to guide development. current area of interest Tech Tree tool to map science and tech Open Source For now, work on this can be found within the documentation document , and in the Github repository here , where we are processing public participant information for the and applying LLM methods to this. RxInfer.jl Learning Group here https://github.com/ActiveInferenceInstitute/Symposium/tree/main/output Applied Active Inference Symposium
Theoretical Neurobiology (TNB) Group Theoretical Neurobiology (TNB) Group Objective The TNB Group has been fostering interdisciplinary research and collaboration for decades. Our mission is to advance the understanding and application of active inference, a theoretical framework developed by Prof. Karl Friston. This is achieved through regular online meetings featuring presentations and discussions, which may include empirical data and its analysis, simulations, and mathematical development. We welcome contributions and perspectives from diverse fields, including neuroscience, mathematics, machine learning, psychology, philosophy, medicine, and biology. Recordings of past meetings, organised by research area, can be found on the . Note that we are gradually adding more videos from our archive alongside recently recorded sessions. TNB YouTube Channel Meeting Details Schedule: Mondays and Tuesdays, 2:30 pm (UK time) Duration: ~2 hours Structure: ~40-minute presentation ~40 minutes of Q&A ~40 minutes of discussion and feedback with Prof. Friston Frequency: Weekly Platform: Zoom - the meeting link is sent via our mailing list (email us at to join the mailing list) [email protected] How to Participate Our meetings are open to researchers, students, and professionals worldwide. With no membership fees, we provide a relaxed, no-pressure environment for engagement, whether through active participation or as an observer. You are welcome to join any session that interests you. To join our mailing list for updates on upcoming presentations, active inference events, or job opportunities — or to request to present your work — email us at . Please note that presentation slots typically book two to three months in advance. [email protected] Chairs Riddhi J. Pitliya, PhD. • • • ◦ ◦ ◦ • •
My research focuses on active inference, human cognition, and multi-agent systems. I completed my PhD at the University of Oxford, where I investigated how individuals infer causal structures and agency in their environments, particularly across the depression spectrum. We found that depressive symptoms are linked to reduced sensitivity to inhibitory causal relations, reduced perceptions of agency, and a tendency to engage in frequent but less goal-directed actions when learning about causal structures. Currently, I work at in the Intelligent Systems Lab, where I develop computational models of theory of mind, leveraging active inference to facilitate collaboration and competition among multiple agents. VERSES Miguel De Llanza Varona I’m currently a PhD student at the University of Sussex under the supervision of Christopher L. Buckley and Anil Seth. My research lies at the intersection of AI and theoretical neuroscience where I explore the theoretical underpinnings of representation learning in bounded rational agents. My main research interests are twofold: first, how cognitive constraints (e.g., metabolic costs or limited memory) interfere with optimal Bayesian inference; and second, what are the challenges of learning representations in service of reconstructing the data in misspecified generative models (e.g., VAEs). Peter Thestrup Waade My research focuses on computational cognitive modelling of multi-scale social interaction, particularly from the perspective of active inference and predictive processing. I did my PhD with Chris Mathys at the Interacting
Minds Centre at Aarhus University, and am starting a postdoc position at the Translational Neuromodelling Unit at ETH Zürich with Klaas Stephan. I develop Julia software for in general, the and - I also do some work in consciousness research, on joint action in partner dancing and on Chinese philosophy and predictive processing. cognitive modelling Hierarchical Gaussian Filter active inference with POMDP’s Robert Chis-Ciure, PhD. I’m an ERC postdoctoral research fellow in Anil Seth’s lab at the and . Our research focuses on formalised notions of emergence and computational neurophenomenology. We’re using hybrid predictive coding and active inference formalisms to model various phenomenal properties of experience and validate them experimentally. In doing this, we’re building toward a new methodological paradigm, Phenomenomics, to comprehensively characterise the “inner worlds” of human and, eventually, all other observers—their phenomenome—by also leveraging AI/ML strategies on . Before Sussex, I was a Fulbright postdoc at NYU under David Chalmers, a Tatiana Foundation postdoc in Georg Northoff’s lab, and a Fulbright Ph.D. student in Giulio Tononi’s lab, working on consciousness at the intersection of philosophy, neuroscience, and computational modelling. In my free time, I do various projects as an affiliated researcher at the . University of Sussex Sussex Centre for Consciousness Science large scale datasets Wolfram Institute Will Yun-Farmbrough I am a PhD student at the Sussex Centre for Consciousness Science, supervised by Anil Seth and Chris Buckley. My research investigates how perceptual phenomenology in human subjects can constrain and inform predictive coding models of cortical processing — what are the algorithmic underpinnings for how our world appears to us? I am also interested in predictive processing approaches to the meta-problem of consciousness, seeking to understand how intuitions of conscious experience and qualia might arise naturally in certain generative model hierarchies. I enjoy surfing, houseplants, and zen. Work in Progress