Formación Integral 3.0

Living knowledge in Internet 3.0

Formación Integral 3.0 connects Totonacapan’s living heritage with research on learning, artificial intelligence and Internet governance. It starts with a concrete human capability: knowing how to do something, recognising when it works, correcting it and teaching it to another person.

Internet 3.0 brings a change in how knowledge circulates. AI assistants and agents find information, translate it, summarise it and turn it into instructions. Our work keeps the people who originate knowledge within that journey: their names, context, judgement and decisions about the use of what they teach.

The central question is direct: what must travel with knowledge so that it remains connected to the people and communities who sustain it? Formación Integral 3.0 makes this a research programme linking the Totonacapan Institute and Tegrity.AI within The Integral Management Society.

A trajectory that changes the direction of learning

1.0 · Learning with the master

Jewellery, silverwork, watchmaking and other Papantla crafts pass through observation, practice and correction. La Bola de Oro provides an entry into this history: the workshop brings together tools, materials, memory and relationships of trust. We use Formación Integral 1.0 to name this approach to learning between masters and apprentices.

2.0 · From the world to the region

Between 2005 and 2014, Formación Integral developed a digital-learning experience in Totonacapan supported by cyber-schools, teachers and local operators. Sharing computers and connectivity addressed part of the problem; personal support helped each learner find content, use the tools and turn access into useful learning.

Preparing teachers and facilitators was fundamental. They retained their position as trusted reference points and could guide students and families. The historical catalogue brought together more than seventy course lines, alongside resources for teachers, educational centres and digital libraries. That experience established a working principle: organise technology around the relationships that make learning possible.

3.0 · From the community to the world

The new phase reverses that journey. A master shares a practice; a younger person learns it; an AI system learns alongside them; and the community reviews the representation that may circulate. Heritage becomes a source of knowledge for technological development and for new generations of practitioners.

Three design principles

Portable context and locally controlled memory

Sources, instructions, decisions and learning progress should be maintainable in files and records controlled by people and communities. The architecture works with persistent memory, explicit processes and replaceable models. Learning can then continue as tools, providers or connectivity conditions change.

Distributed trust

Introducing AI begins with masters, teachers, elders and facilitators who know the setting. They prepare to interpret answers, compare them with practice and explain which information the system uses. Trust has recognisable people, clear responsibilities and ways to correct the work.

Demonstrated competence

Learning is organised through concrete challenges. The person performs, observes, decides, adapts and explains. Human and AI performance are assessed separately. Evidence includes what each participant can do, the corrections they need and what they can transfer to a new situation.

One master and two apprentices

The central exercise brings together a master, a human apprentice and an AI acting as a second apprentice. The master demonstrates a practice in its setting. The apprentice attempts it, asks questions and receives corrections. The AI organises what it observes and presents its interpretation. Errors and questions reveal details that an initial explanation may leave implicit.

Choose → agree → demonstrate → practise → test → transmit. Each step leaves a decision and evidence: who teaches, what is recorded, what was learned and which representation is authorised to circulate.

Working with chijol wood offers a particularly productive challenge. An experienced practitioner recognises grain, tool condition, pressure, angle and the moment to change a movement. Learning requires connecting material, tool, gesture and result. Correcting a failed attempt can reveal an essential part of the craft.

The same approach matters to jewellery and watchmaking: recognising wear, choosing a tool, assessing a repair and explaining why a particular object needs a particular treatment. The research output is a body of evidence that another practitioner can examine and use, with the master’s judgement as a reference.

Who does what in the exercise

The Formación Integral 3.0 model separates three tasks and brings the evidence together for review.

Master practitioner

Demonstrates a practice and explains corrections.

Human apprentice

Attempts the practice, asks questions and adjusts their work.

AI as second apprentice

Organises the authorised record and presents its interpretation.

Review of the evidence

Compare what the person could do with what the AI could explain. Evaluate each performance separately.

See the chijol case and evaluation questions →

Rose Luckin: making human capability visible

On 2 September 2026, Professor Rose Luckin read the discussion note and provided a written comment with permission to use it. She highlighted the reversal of direction—from world to region and from region to world—and encouraged a sharper account of how to demonstrate growth in a person’s capability.

“Formación Integral 3.0 addresses something that much of the current discussion about AI in education avoids: the difference between fluent output and human competence. AI can generate polished results while the person using it learns very little. The proposed pilot is interesting because it makes human learning the thing that must be demonstrated, and because it prepares trusted educators and practitioners before wider adoption rather than after. The research question I would encourage the team to sharpen is how to evidence that a practitioner’s capability has increased: what they can do, judge and adapt without AI support, and whether they can explain why. The parallel learning of practitioner and AI agent also raises a useful evaluative question about what an AI system fails to capture when tacit and culturally situated knowledge is involved, and communities are well placed to be the judges of that.”

Rose Luckin · Comment for Formación Integral 3.0, 2 September 2026.

Her observations lead to three strands: an independent demonstration by the apprentice, reflection on learning and on AI errors, and a clear definition of who holds the portable context and how it can be withdrawn. These strands guide evaluation design and the involvement of knowledge custodians.

Europeana: four questions that develop the method

On 3 September, Lorena Aldana of Europeana Foundation shared observations after consulting training and data-governance specialists. Her questions addressed hallucinations, topic relevance, learning pace, assessment of craftsmanship and data governance. On 16 September, we replied with four developed notes.

Sources, errors and recovery

The first note separates original sources from generated interpretations. It proposes preserving records, reconstructing a task’s path, comparing outputs and repeating a step with another model. Evaluation examines source fidelity, omissions, introduced information, retained context and recoverability. Project memory remains available to revisit and correct an interpretation.

The judgement of the craft

The question of how AI might judge woodworking leads to a practice architecture: materials, tools, sequence, environmental conditions, tolerances, quality signals and exceptions. The master and other competent practitioners establish the criteria used to examine the result. AI supports observation, documentation, comparison and retrieval.

The chijol challenge

The third note develops learning from the difference between expert performance and the apprentice’s attempt. Changes of grip, material selection, unexpected questions and corrections help identify the information needed to transmit a practice. The transfer test asks another competent person to use the preserved knowledge to perform a bounded part of the work.

Data, attribution and future value

The fourth note treats data governance, authority over knowledge, provenance and the value of future reuse separately. It develops distinct decisions for teaching, recording, AI processing, representation, publication and reuse. The design includes visible recording states, local operation, restricted materials, attribution rules and review of benefits that may emerge over time.

A dialogue connecting territories

Xanath and Totonacapan

The Xanath framing brings together vanilla cultivation, habitat stewardship, material selection and everyday management of the setting. It provides a way to study knowledge combining ecological observation, experience and practical decisions under limited resources. Alongside jewellery and other crafts identified through the Census work, it places research challenges in concrete practices and places.

Jokkolabs Banjul · The Gambia

On 8 September, Poncelet Ileleji confirmed his agreement with the proposals and the inclusion of Jokkolabs Banjul in the framing. On 9 September, he contributed the perspective of his work with the Gambia National Internet Governance Initiative, the Gambia NRI. His intervention placed the comparison within a learning, participation and governance ecosystem connecting local work with the IGF.

This contribution strengthens a practical question: how to adapt AI learning to the resources, trust networks and challenges of each setting. Correspondence continued in October, when Poncelet acknowledged the proposal update.

United Diplomats of Africa · Senegal

Amath Ndiaye contributed four priorities for African adaptation: local trust structures, orality and local languages, low-connectivity operation and challenges connected with everyday life. He proposed developing the approach with communities and identified masters, elders, women’s and youth associations among the relevant actors.

To make this concrete, he presented a case study in Diakhao, in the Fatick Region, focused on transmitting Sérère knowledge of plants and traditional care practices. The framing proposes oral interaction in Sérère-Siin and Wolof, a French interface and local support on Android phones. Comparison with Xanath provides a way to study the same architecture across different knowledge domains, languages and conditions of use.

In September, Amath reviewed and expressed agreement with the material concerning UDA and Diakhao/Fatick. In October, he again offered help to continue the work. His contributions develop custodians’ authority, local memory and cultural adaptation as central elements of the programme.

Research ICT Africa

Liz Orembo and Kola Ijasan received the concept with interest and proposed reviewing the note and discussing it. The exchange opened a route to consider the architecture alongside questions of agency, justice, contextual knowledge and digital governance in Africa. Correspondence records the preparation of that dialogue and both researchers’ willingness to examine the approach.

Masakhane

Lydia Kila Taban asked for a concrete account of what Masakhane’s participation would entail and took the question to her team. The dialogue helped sharpen the linguistic dimension: preserving meaning and context when working with local languages, and involving their speakers in design and evaluation.

GCSAYN and communities of practice

Divine Ntiokam reported that the GCSAYN team had reviewed the proposal and requested a conversation to explore it further. The exchange connects research with priorities in youth, learning and community practices. Etienne and Bev Wenger-Trayner, in turn, found the question of AI’s contribution to cultural continuity interesting and explained that their publishing work occupied their available time.

IGF 2026: proposal, guidance and review

The IGF conversation began around the MAG Chair Office Hours. Jennifer Chung highlighted the value of learning across regions and connected the exchange with Anja Gengo, the Secretariat’s IG Labs focal point. On 1 September, Anja emphasised the specific value of regional comparison and offered an opportunity for discussion at the call’s webinar.

On 15 September, we submitted Formación Integral 3.0 to IG Labs. The proposal appears as number 136, under AI-facilitated information discovery and the Web ecosystem. It brings together the Totonacapan trajectory, African dialogue and a governance question: retaining custodianship, permissions, attribution and correction authority as knowledge circulates through AI systems.

On 30 October, the Secretariat communicated assessments by three reviewers in a process covering 133 proposals. Formación Integral 3.0 was not selected for the IG Labs 2026 onsite programme. The review left three concrete contributions for the research to take forward.

Reviewer 1 · Defining the problem and implementation

The reviewer requested a clearer problem definition, a concrete technical implementation and differentiation from earlier initiatives. This observation focuses the work on a bounded question and a specification that can be tested.

Reviewer 2 · A question the laboratory can test

The reviewer highlighted what must travel with community knowledge to keep it connected to its custodians. They described it as a difficult question that the laboratory can examine concretely.

Reviewer 3 · Permissions, provenance and collective authority

The reviewer identified granular permissions and community capacity to correct, withdraw or refuse a representation as strengths. They requested development of the portable record’s technical form and interoperability, its relationship with provenance, identity and content-use standards, and the authority to decide over collectively held, intergenerational or partly public knowledge.

The Secretariat encouraged continued development of the ideas and dialogue with the IGF community. The review centres the next phase on the portable record and the demonstration of human competence.

The record that accompanies knowledge

The design object is a portable record connecting five dimensions: source and custodianship; permitted uses; attribution and transformations; review and withdrawal; evidence of learning.

The record identifies what was learned, where it comes from, who may review the representation and which conditions accompany each use. It retains versions and corrections, distinguishes the source from AI interpretation and supports tracing the journey across tools. Alongside it, the learning record shows what the apprentice can do and explain independently.

The research agenda connects these elements with tests across models and languages, local operation, collective authority and craft continuity. The Institute brings work on living heritage and communities of practice; Tegrity.AI develops questions of architecture, context, provenance and systems. Formación Integral 3.0 is the bridge between them.

Documents and continuity

This page brings together the Formación Integral narrative, the discussion note, the four replies to Europeana and exchanges held between August and October 2026. Rose Luckin’s comment is reproduced with her permission. IGF observations are presented by reviewer as communicated to the project.

To contribute a practice, a research question or a comparable experience: contact the Institute.

AI & Ancestral Wisdom is part of Formación Integral 3.0: it is the methodological development of the same programme of learning, knowledge transmission and research with artificial intelligence. Its complete content is brought together here with the history and exchanges that shape the programme.

Practice notebooks · Totonacapan Institute

AI & Ancestral Wisdom: from human apprentice to artificial apprentice

The original approach begins with human learning. A community identifies a practice it wishes to transmit, agrees what may be observed and documented, and recognises the people able to evaluate learning. Artificial intelligence enters afterwards, as a subject of research within that knowledge environment.

Why Totonacapan is part of the research

Totonacapan offers an especially rich context for this question. Its crafts, languages, ecological practices, artisanal techniques, oral memory and transmission between masters and apprentices contain knowledge that is not fully represented in manuals or large digital corpora. The research interest is not to turn the territory into a source of data, but to understand how a judgement that depends on people, materials, relationships and specific circumstances is recognised, exercised and transmitted.

The project does not seek to replace masters or bearers of knowledge, nor to extract knowledge without its custodians. Documentation and any AI trial must begin with consent, attribution, cultural context, benefit to participants and the possibility of restricting or withdrawing particular uses. Technology is an instrument subject to evaluation; this does not give it authority to define or represent the practice.

Two kinds of ambiguity to distinguish

Responsibility and authority: who contributes a record, who may correct it and who has legitimacy to decide about a practice or a use.
Version: which document is current, what changed and whether two descriptions contradict each other or express legitimate variants.

Organising these aspects makes a more demanding question possible: when documentation is clear, what still cannot be transmitted through a record? The project distinguishes loss caused by documentary disorder from difficulty associated with gesture, context, material perception or expert judgement. Tacit knowledge does not automatically become a documentation error.

AI as a parallel apprentice

The design proposes observing what a system reconstructs from authorised procedures and comparing it with a person’s learning. The evaluation should identify omitted steps, sequence errors, ignored conditions and apparently coherent responses that alter the practice. Those who can recognise such differences in the craft are responsible for assessing them.

AI supporting a new learner

A second stage asks whether the representation learned by the system can help teach. The criterion would be what another person can understand and perform with its support, under human review. Repeating a persuasive explanation and exercising a capability are different outcomes that must be measured separately.

Comparing without erasing context

The research proposes comparing AI architectures under controlled documentary conditions and agreed tasks. Variables include record quality, clarity of responsibilities, version control and the amount of context required. Each practice’s limits should shape the experiment.

These are hypotheses and a proposed method, rather than findings already obtained. Any trial involving a community’s materials requires prior agreement on participation, access, review, AI uses and the return of findings. The purpose is to learn what supports transmission and to recognise where human relationships remain indispensable.

Complete proposed model · 21 sections

Proposed model: from living knowledge to artificial intelligence evaluation

AI & Ancestral Wisdom proposes a progressive, auditable method for studying how human knowledge is preserved, formalised, transmitted and tested against contemporary artificial intelligence systems. It does not begin with AI. It begins with living communities of practice, master–apprentice learning, documented procedures, traceability and research into the ambiguities that arise when we try to represent a practice through information.

1. The research premise

A significant part of human knowledge is not fully expressed in rules, documents or datasets. It lives in gesture, timing, relationships with materials, territory, shared memory, ritual and judgement acquired through experience. Michael Polanyi drew attention to this dimension of tacit knowledge: we can often do more than we can explicitly explain. The research asks which parts of a capability can be described, taught or modelled, and which continue to depend on practice and living human communities.

2. Why this is more than cultural digitisation

Digitising documents, recording languages, describing crafts, organising expert knowledge and establishing ethical rules for data are valuable but distinct tasks. This proposal connects them through one question: what happens to a capability when it moves from a living practice to a formal representation? The purpose is not to turn a culture into digital content. It is to study what is retained, transformed and lost in that transition, and how to recognise those differences with the people who know the practice.

3. The methodological principle

Before introducing AI systems, the work identifies practices, activities, procedures, data, transformations, responsibilities, versions, evidence, decision rules, exceptions and validation criteria. This foundation makes it possible to distinguish a written instruction from the operation actually performed and, in turn, from what an expert resolves through experience without being able to reduce it entirely to a rule. Documentation is not assumed to be a perfect reflection of practice: that difference is also observed.

4. First phase: census and priorities

The first step is to identify capabilities, techniques, crafts and people who wish to participate in their documentation. The census seeks to recognise which practices exist, which depend on very few masters, where generational succession is weakening, which materials or spaces are indispensable and which forms of transmission need support. Its usefulness goes beyond a list of names: it provides an initial map of living capabilities and the conditions for their continuity. Priorities must be agreed with the people who practise that knowledge.

5. Second phase: communities of practice

Knowledge does not reside exclusively in individuals. It is also sustained in workshops, shared routines, tools, materials, learning, corrections, relationships of trust and collective quality criteria. The research identifies existing communities or the conditions in which they could be strengthened, without imposing an external structure on them. Comparing variants and sharing criteria can reduce dependence on a single master, provided each practice’s legitimate diversity is respected and its bearers retain their authority.

6. Third phase: observing master–apprentice transmission

The master–apprentice relationship is the principal field of observation. Attention is paid to what is explained verbally, what is shown with the hands, what is learned through successive corrections, what depends on timing or material and what requires expert judgement. Acceptable changes are distinguished from errors, as are the conditions in which two masters may act differently without either being wrong. The aim is to make learning observable without prematurely reducing its complexity to data.

7. Fourth phase: procedures and formalisation

Once the real practice has been understood, procedures can be developed to describe activity sequences, tools, materials, functions, responsible people, versions, transformations, quality criteria, exceptions, tests and relationships between documents. Photographs and other records are included only under the authorised conditions. Formalising does not mean imposing a single version: it means distinguishing what can be standardised, which variants must be retained and which aspects continue to depend on circumstances and human judgement.

8. Fifth phase: traceability and auditability

The next step is to establish whether a description can be traced to its sources and whether a competent person can review how it was obtained. Document control, version management, quality criteria and process auditing make it possible to move from an observation to a verifiable and partially transmissible record. The question is not whether everything can become a procedure, but which parts are consistent, which variations are legitimate and which aspects would lose their meaning if artificially stabilised.

9. Sixth phase: a hypothesis about informational entropy

The model proposes, as a hypothesis to be tested, that certain informational ambiguities could help identify areas where unformalised knowledge persists. It pays particular attention to ambiguity of responsibility or authority—who contributes, safeguards, validates or can decide about a practice—and ambiguity of version—which description is used, how it changed and why different variants coexist. These ambiguities do not in themselves demonstrate tacit knowledge: they can indicate either a documentary defect or real contextual complexity. Making that distinction is itself part of the research.

10. Correctable disorder and structural tacit knowledge

When ambiguity comes from incomplete documents, unclear responsibilities, accidental duplication or lack of version control, the response can be documentary and organisational: clarify authorities, consolidate records, establish provenance and verify compliance with agreements. By contrast, when two variants legitimately respond to different materials, seasons, territories, masters, rituals or circumstances, eliminating the difference can destroy knowledge. The task is to understand it, document its conditions and validate it without forcing a simplification.

11. Four kinds of ambiguity to distinguish

The working model separates four situations. Documentary disorder arises when authorship, version or control is missing and can be corrected through better records. Operational duplication occurs when two areas maintain information without coordination and requires clear responsibilities and master sources. Contextual ambiguity occurs when several versions are valid under different conditions and requires those conditions to be described. Finally, structural tacit knowledge arises when a practice depends on expert judgement that has not yet been formalised; it requires observation, comparison and preservation of context. Classifying these cases avoids treating every difference as an error or, conversely, invoking mystery to justify a correctable defect.

12. Seventh phase: AI as apprentice

AI enters only after a sufficient human and documentary foundation has been built. Its first role is as an apprentice alongside the human learner: it receives procedures, observations, authorised images, variants, rules, exceptions, contextual language and corrections recorded during transmission. What a human apprentice understands and performs is compared with what the system appears to understand from those materials. The coherence of a textual response is not treated as evidence of understanding the practice.

13. Eighth phase: AI supporting teaching

The second role is more demanding. The study examines whether the system can explain a practice in a way that helps another person learn it, such as a new apprentice, an artisan from another context or a technical observer. Remembering information is no longer sufficient: sequences, conditions, variants and limits must be retained. Evaluation distinguishes between producing a persuasive explanation and actually contributing to a capability that another person can exercise under appropriate supervision.

14. Human review of AI-assisted transmission

Explanations generated by AI systems must be reviewed by people who are competent and have the legitimacy to assess the practice. Review examines whether the working sequence is retained, context respected, legitimate variants recognised, exceptions identified and provenance and limits of use maintained. It also looks for omissions, invented assertions and a false sense of coherence. A system can write fluently while altering the technical, cultural or symbolic meaning of what it purports to teach.

15. Experimental hypothesis

The central hypothesis proposes that the greater a practice’s contextual complexity and density of legitimate ambiguities, the harder it may be for an AI system to represent, stabilise and transmit it faithfully. This relationship still needs to be defined and tested empirically; it is not a demonstrated result. To study it, profiles can describe the number of variants, authority over records, version control, dependence on context, expert judgement, implicit transmission, territorial differences, ritual or symbolic significance and resistance to uniform formalisation.

16. Comparative research

The method is designed to compare different practices without imposing a single cultural measure. Possible cases include embroidery from Totonacapan and other regions, jewellery, vanilla cultivation and curing, weaving, watchmaking, restoration, ecological knowledge and complex workshop techniques. Each case requires its own agreement with the bearers of the knowledge and its own assessment criteria. Comparison seeks to understand which aspects can be made explicit and which require situated experience, rather than rank cultures by how easily they can be digitised.

17. What the model could reveal

A predominantly procedural practice may be easier to record, review and transmit. Another practice, highly dependent on changing circumstances, will need contextual rules and human validation. Other capabilities depend on material perception, gesture, timing, community authority or ritual knowledge and may resist deeper formalisation. These are working profiles, not a definitive classification of people or cultures. One of the most valuable outcomes could be to identify precisely where AI systems fail and which human conditions cannot be replaced.

18. Proposed contribution to AI evaluation

Much current AI evaluation focuses on language, reasoning, search, mathematics, programming or task execution. This proposal offers a complementary kind of trial: testing models against situated tacit knowledge, master–apprentice transmission, legitimate variations and situations in which documentation does not exhaust the context. The aim is to observe which architectures or strategies can recognise their limits and work with ambiguity without producing falsely confident explanations or reducing cultural differences to an undifferentiated average.

19. Proposed contribution to preservation

The same method can help identify capabilities at risk, strengthen communities of practice, document variants, support teaching across generations, organise traceable archives and improve cultural attribution. It can also prepare useful information for studying formal protection mechanisms where appropriate and with consent. Preservation does not simply mean storing a recording or a description: it means maintaining the conditions that allow a capability to continue being practised, corrected and transmitted.

20. Institutional status and implementation conditions

This text describes a proposed research methodology, not a completed experimental programme or recognition granted by an external institution. Its possible implementation requires voluntary communities, clear agreements on rights and access, competent researchers and evaluators, field resources, documentation and review criteria. The Integral Management Society and the Institute bring experience of working with communities of practice, documentation, process management and complex systems, but each case must specify its own participants, scope, permissions and evidence before being presented as an activity carried out.

21. The question that remains open

AI & Ancestral Wisdom does not seek to demonstrate that artificial intelligence can replace human wisdom. It seeks to establish, practically and verifiably, which knowledge can be documented, which can be partially taught, which can be modelled and which must remain connected to people, communities, materials and territories. The central question is not whether AI can store ancestral knowledge. It is whether it can learn, explain and support its transmission without losing the coherence of the context or erasing what makes a practice legitimate and alive.

See the programme sequence →Research at Tegrity.AI ↗

Conceptual area of work

Technology can help organise an archive. Understanding knowledge also requires the people who live it.

AI & Ancestral Wisdom explores documenting living knowledge and preparing records for educational and research uses while retaining context and provenance. It is an area of work within the Institute; this page does not present a deployed system or pilot findings.

Before digitisation

The first step is to agree what should be preserved and why. Records must respect language, their bearers’ interpretation and access conditions. Some knowledge can be described without disclosing its content; some requires restricted access.

Four questions for each record

Provenance: where does it come from, and who contributes the information?
Context: which language, practice and situation give it meaning?
Permission: which uses have been agreed, and within what limits?
Review: who can correct a translation or interpretation?

The role of artificial intelligence

Digital tools may support classification, search and draft transcription. Such assistance requires human review and traceability. An automatic translation does not become a community’s authentic voice, and an inference is not testimony.

Access and reuse

Publishing a description does not grant rights to train models or reproduce all related material. Each use needs suitable permission, attribution and responsibility. The purpose is to strengthen memory and transmission while preserving participants’ ability to decide.

A connection with IMSV’s research

Dialogue with IMSV’s technology work helps frame questions about documentation and integrity. It does not imply validation of a tool or an agreement with third parties.

Discover Tegrity.AI →

Propose a conversation →