There’s a whole other side to learning.

The prevailing assumption is that work and learning are both distinct and separate. Work-Generated Learning (WGL) keeps the distinction but removes the unnecessary separation. Its premise is that a prevailing and unexamined paradigm that treats learning as acquisition is a major contributor to workforce development challenges faced today.

Participation-based models of learning already exist in practice, such as apprenticeship, work-based learning, and on-the-job training. In our view, however, these models suffer from an impoverishment of theory to bound and explain the practice across contexts, and organizational learning suffers as a consequence. This gap runs both ways. On one hand, there is insufficient theory. On the other, there is an abundance of learning theory that is descriptive, characterizing learning more than showing how to prompt it [16].

Hence, the purpose of WGL is to advance and institutionalize a participation paradigm of learning across industry, K-12 education, and higher education that is theoretically rich and practice-informed. The framework is available for public and organizational use in understanding why a practice may or may not work, catalyzing conditions for participatory learning, and appropriately measuring and analyzing the results.

We developed this framework to guide our own work at Intangled Learning and as a public resource for education leaders in participation-based models (apprenticeship, work-based learning, CTE, etc.). Intended cross sector audiences include (but not limited to):

• •
  Industry: Learning and Development, HR, and Operations teams at both the organizational and association levels.
• •
  K12 education: Teaching and Learning, Human Resources, and CTE departments.
• •
  Higher education: Industry and work-based learning teams creating degree programs grounded in employer partners' participation-based knowledge.

Apprenticeship has emerged as a national priority [25], at the same moment artificial intelligence has scaled. WGL reads the two trends as resting on sharply divergent epistemologies, and together they expose the limits of the acquisition metaphor. Large language models show that sophisticated linguistic performance can be produced by learning patterns from very large text corpora. Yet fluency is not evidence of the complex, situated judgment of a practitioner: formal linguistic competence and the functional competence to use language in the world are distinct [49]. Zahavy (2026) [75] argues that these models are strong at induction and increasingly at deduction, but "structurally incapable of the abductive 'Jump'" needed to formulate new premises. Codified, explicit knowledge is now widely accessible; what remains scarce, and therefore valuable, is the contextual, tacit knowing that develops through participation in practice. This brings renewed clarity to Polanyi’s claim that we know more than we can tell [53]. Fedorenko and colleagues (2024) conclude that "language does not appear to be a prerequisite for complex thought, including symbolic thought" [26]; language transmits cultural knowledge but reflects, rather than gives rise to, human reasoning. Their work suggests that some of what we know cannot be fully put into language.

At the same time, the prevailing participation model, apprenticeship, is theoretically impoverished. It is widely treated as a skills pipeline, a career pathway, or a workforce strategy. Those framings are not totally invalid, but they are insufficient: apprenticeship is fundamentally a theory of communal learning. Organizations may hold several communities of practice, and communities may cross organizational lines. Over forty years of relevant scholarship has accumulated, but much of it is descriptive; as Davis (2008) [16] notes of theories such as constructivism, their own explicators say these frames are not theories of education. Few frameworks translate it into organizational design that generates testable local hypotheses. Today, consortium-style best practices fill the gap, and they break down when carried into a new context. WGL is built for this moment: AI has made the limits of acquisition unmistakably obvious, and participation needs the theory and institutional form to meet the demand.

WGL is being developed through a transdisciplinary approach that combines applied research with theoretical inquiry. It draws on several disciplines, including cognitive science with an emphasis on 4E cognition (embodied, embedded, enacted, and extended), anthropology, complexity science, philosophy, and organizational learning theory. WGL does not aim at a universal model. It aims to be a framework within which context-sensitive models can be developed: each organization builds its own account of how learning and work connect, guided by shared principles.

WGL’s working hypothesis is that deliberately connecting learning and work helps organizations develop expertise and improve observable performance. Each organization’s history, culture, workflows, power dynamics, and ways of recognizing knowledge shape how that proposition should be investigated. Evaluation questions and relevant evidence are developed with participating organizations, examining organizational outcomes alongside individual and interpersonal learning. Inquiry must also weigh evidence that challenges the framework, alternative explanations, and unintended effects. The principles are orienting proposals whose usefulness must be examined in practice.

The framework is not intended to be linear but is laid out in the order in which, in our experience, challenges tend to be encountered when implementing participation-based learning. There are ten sections in three phases (start, learn, sustain). Each section leads with the practice and follows with further explanation:

1. 1
   The practice: what can be done.
2. 2
   The problem it solves: the situation that calls for it.
3. 3
   Further explanation: the proposed mechanism and practical considerations.
4. 4
   Connected theory: the research it draws on.
5. 5
   Ways to begin implementing: first steps, from low-tool to supported.
6. 6
   Intangled Learning's Shared Practice Lab: how Intangled Learning supports the practice.

Snowden (2024) [65] uses folded and entangled to illustrate two ways of thinking about systems. Complicated, ordered systems are folded (from the Latin complicare): they can be unfolded and analyzed part by part. Complex systems are entangled (complexus, "braided"): their parts cannot be understood apart from their interactions. Intangled, a term from Intangled Learning, the originator of WGL, names the intentional entanglement of learning and work that participation-based learning requires. This framework uses "entangled" for complex systems in general and "intangled" for WGL's deliberate design.

WGL is a public good. Work-Generated Learning originated and is maintained by Intangled Learning, but the framework belongs to the public. Intangled Learning stewards its development, incorporating external review and the experience of organizations that apply it, so that the framework continues to improve as a shared resource. It has no official organizational status. WGL is an open-source body of practice and theory that individuals and organizations can freely use, modify, adapt, and build upon for both commercial and non-commercial purposes, provided they credit the original authors and share their modifications under the same license terms.

Work-Generated Learning originated from a team of educators, entrepreneurs, and researchers at Intangled Learning, who continue to maintain it.

• 

  Dr. Kevin Berkopes

  Co-Founder and CEO

  An educator and entrepreneur responsible for the vision behind Intangled Learning. Grounded in both mathematics and education research, he focuses on developing the science behind how expertise is developed in shared practice with others.

• 

  Andrew Salmon

  Co-Founder and President

  An entrepreneur who brings industry expertise to the challenge of learning in practice. His focus lies in developing our framework for apprenticeship so that it bridges theory and implementation, ensuring programs maximize learning opportunities.

• 

  Dr. Christine Nemcik

  VP of Academic Affairs

  An educator with 30+ years in K–12 and higher education, specializing in teacher preparation and curriculum design. With 20+ years in academic research and instruction, she mentors future teachers and first-generation and English-learning students.

Acquisition and participation have always run in parallel. Acquisition treats knowledge as an object, made of concepts that can be held, transferred, and tested, and it has dominated our explicit attention. Apprenticeship has always privileged the other side: knowledge that is embodied, relational, and tacit. The history of Western ideas about learning is largely the story of the first metaphor gaining institutional form while the second persisted in practice without it.

The pursuit of abstract knowledge. Plato’s pursuit of perfect, abstract forms set an early direction: true knowledge lies beyond the particulars of experience. Sfard traces acquisition’s central puzzle, the learning paradox of how we can seek what we do not yet know, to Plato’s Meno [60].

Method orders knowledge. In the sixteenth century, Peter Ramus used the printing press to distribute charts that ordered all knowledge into a hierarchy of courses. Doll (2008) [22] describes "textbookizing" knowledge as a major part of Ramus' reform and of the legacy his followers left. Doll traces a lineage from Ramus through Comenius and Descartes, into Puritan thought on both sides of the Atlantic and New England schooling, then to nineteenth-century American schoolbooks and the efficiency movement epitomized by Frederick Taylor. It culminates in the Tyler Rationale [68]: choose goals, select experiences that express them, organize those experiences, and assess how well they fit the goals. As Doll notes, the goals themselves are never questioned. This sense of method, "bifurcating such knowledge into a hierarchal sequence of linear steps" [22], remains a foundation of mainstream pedagogy. WGL reads the Tyler Rationale as the most institutionalized form of defining learning objectives in advance.

The Enlightenment divides mind and body. Descartes separated mind from body, and with it knowledge from the knower’s embodied situation. Midgley offers a charitable reading: Descartes bracketed the subjective self to protect both the emerging sciences and the mind [50]. The effect was still a lasting split, in which knowledge could exist independent of the body. Locke’s empiricism ran alongside it. We are not born with innate ideas; knowledge comes through experience of the world. Yet Locke still placed that knowledge in the individual mind. Kant tried to reconcile the two traditions. His synthesis placed the active organizing of knowledge in the mind itself. That idea later shaped constructivist theories of learning, which, as Sfard notes, remained within the acquisition metaphor [60].

Positivism wins over pragmatism. In the nineteenth century, Comte’s positivism proposed extending scientific methods to all of human thought [50]. Peirce’s pragmatism offered another path. Knowing is tied to inquiry and consequence, and new ideas begin in abduction, the forming of explanatory hypotheses that are then tested. In WGL's reading, positivism prevailed, carried by the prestige of science and technology. The belief took hold that whatever matters can be reduced to parts and measured [50, 59].

Industrialization scales schooling. Industrialization needed a trainable workforce, and mass schooling scaled alongside civic and religious aims, from Massachusetts’ 1852 compulsory-attendance law to such laws in every state by 1918. Behaviorism became the learning theory of the industrial age. Following the positivist instinct, it treated mental processes as off-limits because they cannot be directly observed.

Cognitivism and constructivism turn inward. Cognitivism broke from behaviorism by restoring the mind, and constructivism followed, with learners actively building their own understanding. Both advanced learning science, and both intensified the focus on what happens inside the individual. Dewey, drawing on Peirce, ran in the background: appropriated in parts, but never integrated at the foundations.

Borrowed theories describe learning but do not prompt it. Davis (2008) [16] observes that education has long been a net importer of theories from other disciplines, often with dubious results. Mid-twentieth-century behaviorism reduced learning to observable changes in action and bracketed out thought, interpersonal dynamics, collective activity, and social context. Constructivism, rooted in Piaget, is the current import, and Davis argues it may be no better suited to the concerns of educators. He notes, drawing on Lave and Wenger (1991) [44], that the frames' own explicators have made clear that "these frames cannot be construed as theories of education" when teaching means prompting learners toward particular competencies [16]. Psychological theories are largely descriptive: they characterize a phenomenon more than they show how to affect it. A learner, a classroom, a discipline, and a culture are different phenomena, so deriving guidelines for teaching from knowledge of just one is, in Davis's words, a category error.

The computer becomes the metaphor. After World War II, the computer supplied a gripping metaphor. Combined with cognitivist accounts of mental activity, learning became information processing: inputs taken in, knowledge constructed and stored. Federal investment after Sputnik, through the National Defense Education Act of 1958, reinforced these models in science and mathematics education and in teacher preparation.

A counter-current forms. In the 1980s, prompted by A Nation at Risk, researchers at the Institute for Research on Learning, among them Brown, Lave, and Wenger, studied how people learn in practice, producing situated cognition and legitimate peripheral participation [8, 44]. Complexity science moved from the physical sciences and economics into the social sciences and education [17]. Enactivism [15, 70] and the extended mind [11] argued that cognition is more than internal mental activity.

Acquisition is institutionalized further. Assessment and funding arrangements predominantly favor individually measured outcomes, such as test scores, credentials, and completions. The No Child Left Behind Act of 2001, enacted in January 2002, cemented acquisition through high-stakes accountability. Venture investment scaled learning tools built around acquisition-based outcomes. Online learning separated higher education from the physical classroom, and coursework-based information transfer became both the means and the end. Participation was largely lost.

Apprenticeship returns as AI arrives. Now apprenticeship is re-entering the picture. It is driven not by an articulated theory of learning but by economics, and by the reality that most working adults are already situated with an employer. At the same moment, AI has made codified knowledge abundant while showing what it cannot do: exercise situated judgment. The result is an epistemic crisis, even for traditional education. WGL reads the dominance of the acquisition metaphor as this inheritance, and offers participation the theory and institutional form it has lacked.

The first six rows adapt Sfard's Table 1 [60]; the remaining rows are WGL extensions. The columns emphasize different ways of understanding learning; neither prescribes a single teaching method. Following Wenger's duality of participation and reification [72], competencies are reified as points of negotiation and to support engagement. Not every point of negotiation can be codified this way: expert judgment also draws on tacit knowledge that the criteria cannot fully capture.

Acquisition makes work and learning seem separate. The acquisition metaphor emphasizes the knowledge, concepts, and skills a person constructs and internalizes. This learning can happen within work, through social interaction, or through formal study. The participation metaphor emphasizes changes in how a person takes part in shared practice: the judgments they make, the responsibilities they assume, and their changing relationships with other participants. WGL privileges participation to help organizations make these developmental opportunities intentional, while recognizing the contribution of conceptual understanding and instruction.

Participation needs intentional design and appropriate evidence. Organizations commonly implement participation inspired learning models through apprenticeship programs and on-the-job training, but lack commensurable methods for analyzing how participation develops. Attainment scores alone cannot illustrate how a novice gains access to consequential work, develops judgment, or contributes to shared practice. WGL therefore delineates evidence of individual understanding, of changing participation, and of organizational outcomes, and places each measure within its proper context. These forms of evidence answer different questions. Together, they enable organizations to analyze correlations between learning, improvement of shared practice, and the organizational outcomes that are locally important.

Methods of measures don’t transfer. Measures are grounded in a paradigm, and paradigms are incommensurable with each other: each sets its own standards for what counts as a problem, a solution, and evidence [43]. There is no transcendent measure that can judge both, and a measure carried from one into the other risks distorting what it measures, because each paradigm shapes what counts as evidence. It is not due to an inherent flaw of participation that it has not had a form of comparable development as acquisition has, but rather a result of under-institutionalization.

Different communities weigh the metaphors differently. The two metaphors are two sides of one coin, but communities weigh each side differently. Organizations are epistemically heterogeneous in what they count as valid knowing [6]. Formal education privileges acquisition: students and schools are assessed and held accountable to it, and higher learning privileges abstract concepts, logical formations, and proofs [8, 9]. Acquisition tends toward individual ownership of knowledge, as in a research paper credited to one academic. Organizations that bring goods and services to market are judged on shared outcomes, so they tend to weigh participation more.

In our work, much of what is called a talent shortage turns out to be a learning and knowledge-continuity problem; recruitment, pay, and conditions matter too. WGL addresses the learning side.

The strain often shows between generations. In our work, we often see younger workers, recently out of formal education, arrive assuming learning is built on acquisition and teacher delivery. This is a practitioner observation, and it may reflect role and experience as much as age. Experienced practitioners have had more time to develop an implicit understanding of learning by taking part in practice socially over time. Because that participation-based experience was never expressed in explicit terms, they struggle to convey it to newcomers. School is largely teacher-directed. Work calls for what Knowles calls andragogy: adults bring experience as a resource, learn best around real problems, and grow more self-directed [38]. This is a second distinction alongside acquisition and participation: it concerns how learning is arranged for adults, while Sfard's concerns what learning is. When an organization gives participation no deliberate form, the two groups work side by side with no bridge between them. The experts' knowledge stays tacit, the newcomers wait to be taught, and each side reads the other as the problem.

Every training and development model blends acquisition and participation. The real question is which one receives the dominant investment. Large enterprises can afford teams that manage parallel training systems built on acquisition. These run beside production on a separate training budget. Participation can then grow over time as moments of convergence arise, when a trainee gets the chance to merge into authentic practice.

Small and mid-market manufacturers, trade companies, and school districts often cannot fund a parallel system. For them, participation has to be the dominant investment from day one. Learning must be built into production and contribute to it, with as little redundancy and distraction as possible. This is a serious challenge, and very hard with the wrong theory of learning. Senior experts must develop talent and produce at the same time. Newcomers must develop and produce at the same time. Both need shared expectations of how learning and production fit together.

The table contrasts two design options; many programs mix them, including how they are funded.

Backwards-mapping gives a false sense of certainty without a parallel system. Given these operational constraints, the instinct may be to reduce uncertainty by using the acquisition model to map clear learning objectives and assessments. Success in large-enterprise parallel systems makes this look like best practice: define a future state, map backwards to the present, and work toward it. It works there because those organizations can afford the energy to constrain it, insulated from the trade-offs of day-to-day operations. An organization without a parallel system gets only a false sense of certainty from the same approach. The learning trajectory is mapped backwards on paper, but operational realities often cannot deliver it. Provisional objectives, revised from practice, remain compatible with WGL.

These organizations need participation today. Understanding the theory of participation is essential for them to manage expectations and to build the social infrastructure that lets newcomers and senior experts learn productively together. That is the job WGL is built for.

The shift changes six things at once:

The columns mark a change in emphasis. WGL adds inquiry, coaching, and evidence of changing participation to instruction and assessment; it does not replace them.

This section gathers the theory behind the Introduction.

Polanyi (1966) [53] argues that we know more than we can tell: tacit knowing underlies the explicit knowledge we can state. Fedorenko, Piantadosi, and Gibson (2024) [26] review evidence that the brain's language network is distinct from the systems that support reasoning, and conclude that language is primarily a tool for communication rather than the medium of thought. Mahowald and colleagues (2024) [49] distinguish formal linguistic competence, knowledge of linguistic rules and patterns, from functional competence, using language in the world, and find large language models strong on the first and uneven on the second. Zahavy (2026) [75] draws on Peirce's distinction among deduction, induction, and abduction to argue that these models have mastered induction and are advancing in deduction but lack a mechanism for abduction. WGL reads these findings together: what remains scarce is the situated, tacit knowing that develops through participation.

Executive Order 14278 (2025) [25] directs federal agencies to plan the expansion of Registered Apprenticeships into new industries and occupations, placing apprenticeship among national workforce priorities. Davis (2008) [16] argues that the learning theories education has imported are largely descriptive, and that their own explicators do not offer them as theories of education; WGL responds with a framework meant to be both descriptive and pragmatic. Doll (2008) [22] traces the method-driven curriculum from Ramus to the Tyler Rationale (Tyler, 1949) [68]. Brown and Duguid (1991) [9] describe organizations as communities of communities, which is why WGL treats communities of practice as able to sit within or cross organizational lines.

WGL's transdisciplinary approach draws on 4E cognition. Varela, Thompson, and Rosch (1991) [70] describe cognition as embodied and enacted, Clark and Chalmers (1998) [11] argue that it can extend into tools and the environment, and Hutchins (1995) [30] shows it distributed across people and artifacts in a working setting. Davis and Sumara (2006) [17] bring complexity science into education. Snowden and Boone (2007) [66] distinguish complicated systems, where cause and effect can be worked out through analysis, from complex systems, where they can be seen only in retrospect. Snowden (2024) [65] describes the same contrast as folded and entangled, which WGL's term "intangled" builds on.

Sfard (1998) [60] identifies acquisition and participation as two different ways of understanding learning. She warns that relinquishing either "may have grave consequences, whereas metaphorical pluralism embraces a promise of a better research and a more satisfactory practice" [60]. The distinction is not individual versus social: acquisition can include socially mediated learning and the internalization of shared concepts. Sfard (2009) [61] describes moving between these discourses according to the questions being investigated.

Kuhn (2012) [43] argues that a scientific paradigm supplies a community's shared problems, methods, and standards of evidence. Competing paradigms are incommensurable: there is no transcendent measure outside them by which to judge one against the other. Sfard (1998) [60] applies Kuhn's term to the two learning metaphors but calls them incommensurable rather than incompatible. Each needs its own standards, yet both can be held at once, much as Bohr's complementarity holds two descriptions of light. She cautions that the persuasive power of data may not carry beyond the paradigm in which it was produced, so findings can be lost in translation between frameworks.

Lave and Wenger (1991) [44] describe learning as legitimate peripheral participation, through which newcomers develop their participation in a community's practice. Wenger (1998) [72] treats participation and reification (the tools, documents, and procedures that give experience a form) as a duality. Brown, Collins, and Duguid (1989) [8] emphasize that activity and context help constitute what is learned. Midgley (2001) [50] critiques extending scientific methods to all of human thought, and Schatzberg (2012) [59] traces how the language of applied science subordinated artisanal knowledge; the link to learning metaphors is WGL's reading.

Before designing a program, build the business case that ties learning to work. Start with the knowledge risk case: what critical knowledge the organization holds, where it lives (in which people, tools, routines, and working relationships), what it would cost to lose it, and by when the cost would arise. This may be obvious enough to make the case on its own, or it may require investigation. Add the opportunity case: the performance that could improve and/or the upside that could be captured. Estimate the full cost of learning, including mentor and apprentice time, slower production, and rework. A case grounded in the organization's own operations is what lets learning be funded from operations instead of from a grant or a training budget. Most importantly, it creates a plausible hypothesis to begin with.

Why does learning conflict with work demands every time budgets get tight? Learning programs are usually justified by headcount targets, compliance, or goodwill, and those reasons lose when production pressure rises. Business cases also usually inherit the acquisition metaphor, which places knowledge in individual heads. Risk becomes a list of people who might leave, and learning becomes transferring what they know to someone else. Knowledge held in tools, routines, and interactions stays out of view, and so do the ways to develop it. Leaders cannot see their exposure if a key expert retires or leaves, so the cost of doing nothing stays invisible. Reliance on time-limited funding creates a sustainability risk, and a subsidy may never catalyze the activity needed to address the knowledge risk.

A program that records only training hours and individual completions leaves much of learning's operational value unexplained. WGL examines how developing expertise changes an organization's capacity to perform and sustain its work. That capacity depends on people, their relationships, and the tools and practices they use. This makes learning relevant to operational risk and opportunity, alongside individual development. Adapting Snowden's ASHEN example [62], WGL describes a dependency through the capabilities and resources required for the work, alongside the people who currently provide them. "Only Linda can do X" becomes "X requires this combination of artifacts, skills, heuristics, experience, and talent, and right now only Linda has it." The second description helps the organization investigate how that capability could be developed more broadly.

1. 1
   Map critical decision points. List the decisions the work depends on. For each, name the knowledge it requires, sorted with ASHEN, and the roles or combinations of roles that currently make it. Flag decisions that rest on a single person, and those that need several roles together.
2. 2
   Estimate the exposure. For each critical decision, ask what it would cost if the knowledge behind it left, and how soon that could happen.
3. 3
   Name the opportunity. Identify the backlog, capacity, or new work that more capable decision-makers would unlock.
4. 4
   Decide what kind of case you are making. Risk, opportunity, or both.
5. 5
   Tie learning to production measures operations already tracks. That way development shows up in the same reports as the work.
6. 6
   Put the case in front of the finance lead, and revisit it. Update it as the narratives and the knowledge map accumulate (Section 9).

A spreadsheet of decisions and roles and a one-page memo with two numbers, the cost of lost knowledge and the size of the backlog, can get you started.

Boisot (1998) [5] treats knowledge as an organizational asset whose value depends on how it is created, codified, and shared, and organizations hold many kinds of it. Snowden's (2000/2004) [62] ASHEN framework sorts those kinds: artifacts, skills, heuristics, experience, and natural talent. Research on distributed and extended cognition (Hutchins, 1995; Clark & Chalmers, 1998) [11, 30] holds that, under certain conditions, tools and other people form part of a cognitive process, which lets an organization look beyond knowledge in individual heads. Vygotsky's (1978) [71] zone of proximal development makes a related point about interaction: what a learner can do with guidance or with more capable peers exceeds what they can do alone, so capability lives partly in the relationship. DeLong (2004) [20] examines the organizational risks of losing experienced employees; seen through participation, that risk also attaches to the relationships and routines those employees sustain. Brown and Duguid (1991) [9] argue that separating knowledge from the practices in which it is developed and applied is unsound. WGL uses these as lenses to frame knowledge as an organizational asset, not a commodity held by individuals.

The participation metaphor lets the organization's day-to-day practices become the curriculum. Learning comes out of the work instead of competing with it for time, and competence means making sound calls in context, not finishing courses. This practice is metaphor-based. Instead of looking externally for knowledge about what your organization does in practice, orient inward to the work done every day. The work that is happening every day, while not codified into text, is the most real curriculum available.

How do we schedule and organize training around work? Where experts are approaching retirement and expertise is concentrated in a few people, shortages can mount quickly. The organization has tried training programs, opened its talent funnel, and launched learning initiatives. People finish courses but still cannot be relied on to make the decisions that advance the work.

How learning is organized affects whether it supports or competes with production. Training scheduled separately from work creates difficulties when staffing and time are constrained, while participation in production requires deliberate access, guidance, and opportunities to reflect. WGL uses real work as the main setting for development and connects instruction to the demands practitioners encounter. Acquisition and participation offer complementary ways to examine this process: what people come to understand, and how their judgment, responsibility, and contribution develop within practice.

1. 1
   Start from the business case and decision map in Section 1. They show where learning most needs to advance the work.
2. 2
   Decide your dominant investment. Prioritize learning through shared practice. Connect instruction and further study to the needs revealed by the work.
3. 3
   Center competence on decision-making. Alongside reliable performance, relevant knowledge, and safe practice, look for "makes sound calls in our context," not only "completed the task."
4. 4
   Show leadership the two metaphors. Show what each helps explain, and use both to examine how understanding develops and how participation changes.

The decision map from Section 1 and a one-page summary of the two metaphors can get you started.

Sfard's (1998) [60] acquisition metaphor treats knowledge as a possession with a clear end point. Her participation metaphor replaces the noun "knowledge" with "knowing," an activity with no end point. Brown, Collins, and Duguid (1989) [8] argue that schooling often keeps learners from the culture of a domain because that culture is not present in the classroom. Klein's (1998) [35] recognition-primed decision model describes how experienced practitioners recognize a plausible response and, when needed, mentally simulate its consequences. WGL draws on Deacon's (1997) [18] distinction between indexical and symbolic understanding to treat competencies as pointers to experience, not containers of mastery. Snowden's (2000/2004) [62] ASHEN framing (artifacts, skills, heuristics, experience, natural talent) shows that only some of what an organization knows can be taught in a course.

Once leadership has recognized the theory of learning at stake, hold a group session with the organization's senior experts. Share the learning challenges the organization faces and how it is taking a more intentional approach, grounded in sound research and practice. Follow the session with voluntary one-to-one interviews in which experts share their expertise. The interviews surface some of the organization's knowledge assets. They also show where mentoring is already happening, who people go to when they are stuck, and which experts are willing to mentor. Discuss their reasons, availability, support needs, and relevant capabilities, then revisit their commitment through experience. Build on the relationships that already exist instead of assigning new pairs. These people become the core of the mentor council, whose job is to support and learn from mentoring relationships, not to manufacture them.

Who will carry this, and how do we find them? Programs often appoint mentors by title or seniority and announce the program to them as a finished plan. Senior experts hear another initiative that adds to their workload. The people who would mentor well are never asked what they know or whether they want to take part. Programs also overcodify the relationship itself, with required pairings, meeting schedules, forms, and ratings. Formalizing a relationship that was working risks crowding out the reasons people mentored in the first place, and once mentoring is tracked and scored, people begin to manage the record instead of the relationship (see Section 9).

Intrinsic motivation and informal relationships go hand in hand. Experts mentor because they care about the work and the people who will carry it on, and trust develops naturally in relationships people enter freely. Formal assignment tries to supply the trust and motivation up front, and it often cannot. Informal mentoring also distributes the work of mentorship across the organization. It grows out of the way work already happens, so it does not depend on a program office or a small set of appointed mentors. Formal oversight still has a role, and the role changes. Because leadership can see where learning relationships are forming, it can act strategically: where the picture shows a knowledge gap, such as a critical skill with no mentor or a team no one learns from, leadership can take formal action there, through targeted recruitment, protected time, or a defined role. Formality is applied where the informal network is thin and left out where it is working. Planning can still establish useful expectations, learning goals, and safeguards, but premature codification can create misaligned expectations that end in disappointment. WGL uses an initial design that participants can negotiate and co-create together, so the pathway develops through planning, participation, and review.

1. 1
   Hold the group session after leadership names the theory. The business case and leadership conversation in Sections 1 and 2 come first.
2. 2
   Explain the why. Share the learning challenges the organization faces and the more intentional, research-grounded approach it is taking.
3. 3
   Name the trade-off openly. Operations teams gain more natural learning opportunities and more say over how learning happens. In return, they accept more ambiguity, because practices are given room to emerge rather than being specified up front.
4. 4
   Invite, do not assign. Make the follow-up interviews voluntary.
5. 5
   Interview about decisions. Ask each expert to recall a specific, difficult case and walk through it step by step. Probe the cues they noticed, the options they weighed, and what a less experienced person would have missed.
6. 6
   Record two things. Note the knowledge assets that surface, sorted by type. Note who expresses willingness to mentor, and talk through their reasons, availability, and support needs.
7. 7
   Map the mentoring that already exists. Ask in the interviews who people learn from and who they go to when stuck.
8. 8
   Invite the willing volunteers to form the mentor council.
9. 9
   Keep the council light. Its work is to support relationships, remove obstacles, and notice gaps, with no required pairings, forms, or ratings.
10. 10
    Act formally on gaps. Where the map shows a skill or team with no learning relationships, decide what formal support to add. Section 5 picks up from here.

One group session, a one-page interview guide, and a notebook can get you started.

Cognitive task analysis (Crandall et al., 2006) [13] and Klein's Critical Decision Method (Klein et al., 1989) [37] elicit expertise by walking an expert back through a specific, challenging incident and probing the cues, goals, and options at each decision point. The Knowledge Audit (Militello & Hutton, 1998) [51] adds probes such as what a less experienced person would have missed. Polanyi (1966) [53] explains why direct questions fall short: we know more than we can tell. Snowden's (2000/2004) [62] ASHEN framing helps sort what the interviews surface. de la Croix, Doepke, and Mokyr (2018) [19] treat moral hazard between master and apprentice as the central problem apprenticeship institutions had to solve, because the training contract is incomplete. They describe informal institutions based on reputation and trust as one way the problem was managed, alongside guilds and the state, and they show the limit of closed arrangements: where training stayed within families, knowledge spread slowly beyond them. For WGL, this supports building on trust in informal relationships while leadership watches for gaps.

Research on workplace mentoring suggests that relationship quality matters more than formal assignment. Underhill (2006) [69] found larger associations with career outcomes for informal mentoring than for formal programs, and Ragins, Cotton, and Miller (2000) [54] found that a dissatisfying mentor can be no better than none for job and career attitudes. Allen, Eby, and Lentz (2006) [2] found that perceived input into the mentoring process and training perceived as high in quality were consistently related to mentoring behaviors and relationship quality, and Eby and colleagues found small favorable associations between mentoring and protégé outcomes (2008) [23], and that instrumental support, psychosocial support, and relationship quality relate to different outcomes (2013) [24]. In a qualitative case study of newly hired engineers at one manufacturer, newcomers learned largely through their immediate work group, not only an assigned mentor (Korte, 2009; Korte & Lin, 2013) [39, 40], which supports seeing mentoring as a developmental network rather than a single pair (Dobrow et al., 2012) [21]. WGL's preference for informal relationships, volunteer mentors, and a mentor council draws on this research; the specific recruitment approach remains a design choice to evaluate.

Conventional thinking emphasizes designing the apprenticeship experience first. However, this risks premature codification of practices that are not plausible or probable in your environment. A design that is both emergent and distributed is needed: it takes shape through practice, and the practitioners themselves interpret the experience and develop the solutions. Hire apprentices before the design is finished, for two reasons: it will never be perfect, and apprentices are a key part of the design process. Bring newcomers in with a legitimate role at the edge of real work, give them a research method and a provisional plan instead of a fixed roadmap, and let the program take shape from what they and the experts work out together. Codify early only what the organization agrees on and can reliably deliver, such as pay or a review rhythm, because these reduce an apprentice's uncertainty. Set the rest of the expectations co-creatively, with logic and theory that can be explained to the apprentice. WGL works with the adjacent possible: the next steps the organization can actually take from where it is. Treating apprentices as researchers of the practice draws on their experience and self-direction.

Why does it never feel like everything is designed enough to launch? It never will be. The organization delays because competencies are not defined, mentors are not trained, and the system is not designed. It tries to backwards-map from competencies or enrollment targets, as if it knew the destination. Or it reaches for design thinking and builds personas, journey maps, and prototypes of a program before any apprentice is in the building. Meanwhile, newcomers arrive expecting a teacher and a clear sequence. Codifying everything early is not the only route to certainty, and it carries its own risk. Organizations bring a history and path dependence, and apprentices bring their own points of view. A pathway fixed in advance risks conflicting with the organization's constraints and misreading what the work actually affords.

Backwards-mapping is the acquisition instinct described in the premise. It reduces uncertainty by fixing learning objectives and assessments in advance, but without a parallel system to protect it, operations often cannot deliver a plan fixed in advance. Its lineage runs back through the Tyler Rationale [68], whose flow Doll (2008) [22] describes as top-down and deductive, leaving no room to move abductively. While design thinking and instructional design methods both offer structured ways to explore and specify a program, their approach centers on a designer interpreting the problems and proposing solutions. A program designed in advance also fixes a hypothesis before the experience exists. WGL calls for a design that is both emergent and distributed. Emergent means it takes shape through practice instead of being fixed in advance. Distributed means practitioners, not an outside designer, interpret experience and develop the solutions.

WGL's recommended design framework is Distributed Practice Research, which builds on the action research and participatory action research traditions [34, 47]:

• •
  Action research: practitioners are the researchers of their own practice, not subjects studied by outsiders.
• •
  Practice-oriented: the inquiry is about the work itself: its decisions, cues, and judgments.
• •
  Distributed: the practice community does the research together. Apprentices, mentors, and peers share the interpretation.
• •
  Applied: the aim is solutions that evolve the practice, not findings for their own sake.

Through Distributed Practice Research, apprentices and experts make sense of their own experience, form plausible hypotheses, and test and refine the pathway through practice. In Peirce's terms, abduction proposes an explanatory hypothesis, deduction derives its consequences, and induction tests them against experience [52]. Narratives and practice supply the experience from which hypotheses are formed. Participation is not designed into practice. It is negotiated and co-created between newcomers and experts, so the newcomers have to be present for the design to happen. Under participation, the directional aim is fuller membership in a practice that keeps changing, and the path cannot be planned in full. Work from the present forward instead: start where the organization already learns, try safe things, and reinforce what works.

The table below outlines an analysis of different design methods commonly used.

Placements are WGL's reading of each method's typical use; practitioners adapt all of them.

1. 1
   Bring apprentices in early. Do it once supervision, suitable work, pay, and a review rhythm are in place. Be transparent that the rest is an evolving model and explain its logic. You don't need to sell a finished program; sell the coherence of a process that is emergent in nature.
2. 2
   Codify what you can deliver. Set pay and a review rhythm at entry, and explain which parts of the pathway remain provisional and why.
3. 3
   Frame the early cohort as co-creators. They help set the way for later apprentices, which is a leadership opportunity.
4. 4
   Reframe the apprentice as a researcher. Their job is to investigate what the organization does, what decisions experts make, and why.
5. 5
   Give them a journal alongside a provisional learning plan. They keep a log of what they see people do, problems being solved, and decisions witnessed, then bring it to senior practitioners. The plan is revised as the journal shows what the work affords.
6. 6
   Decide what a newcomer can break. Mentors and apprentices each keep a list of things to try, and ask whether each is something the apprentice can break.
7. 7
   Tune expectations. Tell apprentices early that learning here is not delivered the way school delivered it. This will need to be repeated over time.

A shared notebook and a weekly conversation can get you started. Stories about real moments carry what a summary or exit interview loses.

Lave and Wenger's (1991) [44] legitimate peripheral participation gives newcomers real but limited access to the work from the start. The periphery is a position in the practice, not a waiting room outside it, and newcomers bring their own perspectives into the community that admits them. Wenger-Trayner (2015) [73] note that the term "community of practice" was coined for the community that acts as a living curriculum for the apprentice. Snowden's safe-to-fail probes (Snowden & Boone, 2007) [66] and Juarrero's (2023) [32] catalyzing conditions describe how to act when outcomes cannot be predicted: set conditions, watch what emerges, add energy to what you like, and take energy away from what you do not.

Wenger (1998) [72] makes the design point directly: learning "cannot be designed: it can only be designed for" [72]. He locates a community's coherence in its practice, which does not reside in a structure that precedes it [72]. WGL's gloss: the practice is the structure. Gibson (1979) [28] and Billett (2001) [3] describe affordances, what a setting offers a learner, and Kauffman (2000) [33] describes the adjacent possible, the set of next steps available from a given state.

Peirce (1903/1998) [52] describes abduction as the process of forming an explanatory hypothesis, "the only logical operation which introduces any new idea" [52]. Snowden (2020) [64] treats narrative as an abductive act of knowing, and Snowden and Boone's (2007) [66] safe-to-fail probes give a way to test emerging hypotheses in complex settings.

Lewin (1946) [47] framed action research as cycles of planning, action, and fact-finding about the results of action, and Kemmis and McTaggart (2005) [34] describe participatory action research as practitioners investigating their own practice together. Wiggins and McTighe (2005) [74] set out backward design, and Tyler (1949) [68] the rationale WGL reads as its ancestor (Doll, 2008) [22].

Don’t frame mentorship as the act of teaching. Rather, position mentors as partners in the apprentice's investigation of practice and guides to developing capability. Use real decisions to surface expert reasoning, then help apprentices develop their own judgment through observation, demonstration, explanation, supported practice, and feedback. A mentor council can be formed to share this responsibility across senior practitioners. WGL proposes that it support a broader developmental network rather than a set of pairs [21]; apprentices also learn from peers and relationships outside the council. The council helps the group examine how its practice is sustained and improved.

Why do our seniors struggle to teach what they do? Senior practitioners may default to the instructional model, or feel apprentice pressure to. Without a curriculum to deliver, they are unsure what to teach. Assigning mentors by title often produces relationships of poor fit, and research suggests a dissatisfying mentor can be no better than none for job and career attitudes [54]. Informal mentoring has shown larger associations with career outcomes than formal programs [69], and newcomers learn largely through their immediate work group rather than a single assigned mentor [40]. WGL therefore starts with willing mentors and the existing informal network. Compensation and workload adjustments should recognize mentoring work, but leading with payment risks framing mentoring as a burden; that framing is a WGL design judgment to evaluate.

Experienced practitioners often find it difficult to articulate the judgments that guide their work. Elicitation helps make aspects of that reasoning available for examination: what cues they noticed, which possibilities they considered, and why they acted. Apprentices also need opportunities to develop and test their own understanding. Mentors support that development by arranging access to meaningful tasks, demonstrating their approach, coaching attempts, and adjusting responsibility as capability grows. WGL brings these activities into shared practice, with apprentice inquiry helping focus the guidance and instruction provided.

1. 1
   Define the mentor's role. Mentors investigate practice with apprentices, review their interpretations, and provide access, demonstration, coaching, explanation, and feedback.
2. 2
   Map the informal network first. Before assigning anyone, ask who apprentices and experienced staff already turn to for advice and whom they trust.
3. 3
   Ask about decisions, not knowledge. "What were you thinking?" and "What cues did you notice?" draw out far more than "Explain how you do this."
4. 4
   Form a mentor council. Start with the volunteers identified in Section 3. A small group of senior practitioners shapes the apprenticeship together, informal first, formal later.
5. 5
   Build on motivation that already exists. Many experienced people already want to develop others. Find them first, then recognize the work through compensation and workload adjustments.
6. 6
   Give conversations a starting point. Apprentice narratives, indexed to the competencies they relate to, keep mentor meetings from drifting into status updates.

A monthly meeting, a shared folder of apprentice stories, and a short list of decision questions can get you started.

Sfard (1998) [60] describes the teacher within the participation metaphor as an expert participant and preserver of practice. Lave and Wenger (1991) [44] locate apprenticeship learning within a wider community of relationships, and Wenger (1998) [72] emphasizes the explicit and tacit dimensions of practice. Klein's Critical Decision Method (Klein et al., 1989) [37] and the Knowledge Audit (Militello & Hutton, 1998) [51] help investigate expert judgment through specific decisions. Collins, Brown, and Newman's (1989) [12] cognitive apprenticeship complements this elicitation with modeling, coaching, scaffolding, articulation, reflection, and exploration.

Krackhardt and Hanson (1993) [41] and Cross, Borgatti, and Parker (2002) [14] show that the informal networks of advice and trust often differ from the formal chart. Knowledge moves through those networks when ties carry trust (Levin & Cross, 2004; Abrams et al., 2003) [1, 45], and people seek information from those they know, value, and can access (Borgatti & Cross, 2003) [7]. Dobrow and colleagues (2012) [21] describe mentoring as a developmental network of several relationships.

Vygotsky (1978) [71] locates development in the zone between independent performance and what a learner can do through "problem solving under adult guidance or in collaboration with more capable peers" [71]. Bruner (1986) [10] describes the tutor as lending the learner awareness and control until the learner can take over the task. Lave and Wenger (1991) [44] review readings of the zone, from scaffolding to its broader social interpretations, and argue that learning is better understood as changing participation in a community's practice. WGL uses these as lenses on mentoring to describe a community and distributed activity.

Lead with participation and phase in acquisition purposefully. Mentors move from the idea of "teacher" to partner in the apprentice's research. Sometimes that means teaching something directly, but there are many other ways to learn. Part of the mentor's role is to pay attention to how a piece of knowledge can best be learned: through stories, rules of thumb, repetition, shadowing, or training. ASHEN gives mentor and apprentice a shared vocabulary to sort the knowledge and generate options, and that negotiation is part of the learning.

Why do apprentices say they aren't getting enough training? Apprentices who expect instruction ask to be trained on everything, and attribute gaps in their knowledge to not getting enough training. The organization assumes that learning comes through training, and that training requires spending on curriculum. Mentors know that some things are hard to train and could be learned faster and better another way.

Many arrive expecting instruction; treat a training request as a learning need to investigate. Knowles's andragogical model [38] emphasizes adults' experience and the problems in front of them; WGL applies it where it fits. Organizations hold different kinds of knowledge, and no single mode fits it all. ASHEN offers a practical vocabulary for discussing knowledge dependencies [62]. Typical routes: artifacts, such as documents and tools, can be read and referenced; skills can be trained, though they take time to internalize; heuristics travel through stories and worked examples; experience comes from time in the work, often through repetition; natural talent can be noticed and put to use. WGL uses these categories to generate learning options, rather than prescribe one method per category. Combine instruction, guided practice, simulation, stories, and work experience according to the task and learner, and evaluate the results. Treat judgments of natural talent cautiously. The way the work is organized also teaches, or blocks, learning without anyone planning it.

1. 1
   Map what the environment already affords. Entry-level work can create learning opportunities, but their value depends on access, guidance, feedback, task variety, and individual engagement. Account for the time and resources these require, and note which kinds of access are easy and which are hard.
2. 2
   Sort the knowledge with ASHEN, then generate learning options. For each competency, ask which parts are artifacts, skills, heuristics, experience, or natural talent, and which combination of routes fits the task and learner. Reserve formal training for what truly needs it.
3. 3
   Make the mentor a partner in the apprentice's research, not only a teacher. The mentor may sometimes teach something directly, but the larger job is to notice how a piece of knowledge can be learned and to help the apprentice get access to that route: a story, a rule of thumb, a repeated task, or a chance to shadow.
4. 4
   Co-create the training you do need. When the apprentice asks for training and the mentor agrees it is warranted but hard to replicate, build the experience together so the apprentice can learn the principles inside it.
5. 5
   Do not import another organization's model whole. What worked elsewhere depends on variables that may not match yours.

An afternoon sorting a list of competencies by ASHEN type, then into "train," "shadow," "story," "repeat," and "self-study," is a good first exercise.

Wenger (1998) [72] separates the explicit side of practice (language, tools, documents, codified procedures) from the implicit side (tacit conventions, subtle cues, untold rules of thumb, embodied understandings). Training reaches the first far better than the second. Snowden (2000/2004) [62] makes a related point through ASHEN: skills are the most readily codified knowledge an organization holds, while experience is the most valuable and the hardest to pass on. Sfard's (1998) [60] argument that each metaphor covers what the other cannot is the case for mixing modes. Fuller and colleagues (2005) [27] show that how work and production are structured shapes the forms of participation open to a learner, and so shapes learning. Gibson's (1979) [28] affordances and Billett's (2001) [3] work on workplace affordances ask what a setting offers a learner. Collins, Brown, and Newman's (1989) [12] cognitive apprenticeship describes modeling, coaching, and fading as ways to learn without a lecture.

Apprentices may experience uncertainty, frustration, or confidence that is poorly calibrated to performance. Investigate learning needs, task difficulty, feedback, and working conditions. Compare self-assessment with observed performance and review progress over time. Give mentors the learning science behind these experiences, and let apprentices read their own record.

Why do apprentices feel stuck, or overconfident, and what should mentors do? An apprentice says "I'm not learning anything" while growing quickly. Another grows confident faster than competent and misses gaps. Mentors take the frustration personally or lose patience. "How did it go?" reviews draw only "good."

When progress is judged mainly by content covered or tasks completed, changes in judgment and participation can remain difficult to recognize. Revising an earlier understanding may feel like losing ground even when it supports development. WGL helps apprentices examine what they now notice, how they explain decisions, and which responsibilities they can undertake. Confidence and performance can also develop unevenly. Comparing self-assessment with observations, work outcomes, and specific feedback helps apprentices and mentors recognize both growth and continuing learning needs.

1. 1
   Teach mentors the science of learning. Frustration and miscalibrated confidence are common, and neither is a verdict on the apprentice or the mentor.
2. 2
   Depersonalize it. A mentor is positioned to treat the struggle as a human process, not a conflict.
3. 3
   Give specific feedback. Tie feedback to observable decisions and work outcomes. Ask apprentices to compare predictions with results and identify what to try next.
4. 4
   Let apprentices review their own record. Reading their own narratives often shows them how much they have learned.
5. 5
   Ask a noticing question in every review. "What would a less experienced person have missed?" shows the apprentice what an expert attends to.

A weekly short entry from the apprentice and a monthly read-back together can get you started.

Klein's cognitive transformation theory (Klein & Baxter, 2006) [36] holds that expertise requires repairing flawed mental models, not just filling gaps. Kruger and Dunning (1999) [42] found substantial overestimation among low performers on several specific tests, a finding that has since been debated. This supports checking self-assessment against independent evidence; it does not establish a universal pattern for apprentices. Research in mathematics education and nursing simulation informs WGL's attention to the cues learners notice and the interpretations they make (Jacobs et al., 2010; Lobato et al., 2013; Rooney & Boud, 2019) [31, 48, 57]. WGL adapts these ideas through journals and mentor conversations. Whether these records reliably indicate workplace growth must be evaluated alongside observed performance and work outcomes. Sfard's (1998) [60] discussion of transfer describes competence as both continuity and adaptation across situations.

Senior practitioners, front-line managers, and performance teams own the competencies and the apprentice trajectory. Codify at entry what the organization agrees on and can deliver, describe levels in broad strokes, let the rest of the path emerge, and codify further only what has proved stable.

How do we keep this from becoming another corporate program that lives in HR? A program administrator designs a set of competencies and hands them down. The executive level owns the program, and it never reaches daily work. Apprentices then ask who they are becoming and by when, and the organization is tempted to answer with fixed levels and timelines it cannot yet back.

Competencies developed away from the work may fail to capture what practitioners need to recognize and accomplish. Their meaning develops through use and negotiation among the people involved. Agreed goals and competence thresholds can support participation without implying that learning has a final endpoint. WGL establishes clear expectations and revises them as experience reveals what the work requires. Structure needs to provide enough clarity for apprentices to understand their responsibilities, progression, and compensation while allowing the pathway and the practice to develop.

1. 1
   Put competency-setting with senior practitioners. The people on the floor, in a senior role, define competencies and use the program to improve their work and manage workloads.
2. 2
   Describe levels in broad strokes. Say what each level means to the organization in plain words, not as a checklist.
3. 3
   Let the trajectory emerge. The mentor council weighs in on where to aim, and apprentices are told which parts will be codified in time.
4. 4
   Codify at entry what you can deliver. Codify what the organization agrees on and can reliably deliver, such as compensation and a review rhythm, and explain which parts of the pathway remain provisional and why.
5. 5
   Codify on time. Move from enabling conditions, to reinforcing what works, to prescribing only what has proved stable. Use Boisot's social learning cycle as the test: knowledge that has not yet been worked through in practice and shared is not ready to codify, and codifying it early risks cutting the cycle short. Provisional codification that keeps its context and can be revised is fine. Decide when levels and recognition get formalized.
6. 6
   Settle who carries the risk and who gets the benefit. Discuss how effort, risk, and benefit are shared among mentors, apprentices, and the company, and review whether the arrangement keeps people engaged. Holmström (1979) [29] offers a lens on incentives when effort is hard to observe. de la Croix, Doepke, and Mokyr (2018) [19] make moral hazard between master and apprentice the central problem that apprenticeship institutions had to solve, because the training contract is incomplete: what is taught, how well, and the apprentice's diligence and loyalty cannot be fully specified.
7. 7
   Judge competence through intersubjective consensus. Senior practitioners judge competence, informed by explicit criteria and documented evidence. Investigate disagreements and provide a transparent route for review.

A one-page description of levels and a list of what is fixed and what is provisional can get you started.

Wenger (1998) [72] argues that external forces have no direct power over a practice; it is the community that negotiates its enterprise [72]. Practitioner involvement is essential because communities interpret organizational requirements through practice. Competence decisions should also incorporate explicit criteria, organizational accountability, and relevant external expertise. Wenger also warns that reification cuts both ways: a good tool can amplify an activity, but it can also ossify activity around its own inertness.

Snowden's later discussion of enabling and governing constraints (2017) [63] and Juarrero's (2023) [32] account of coherence inform WGL's approach to revisable structure. Their application to apprenticeship design is WGL's theoretical interpretation. Boisot's (1998) [5] I-Space and social learning cycle describe how knowledge moves from tacit to codified and abstract, diffuses, is absorbed, and returns to practice. Holmström (1979) [29] shows how arrangements that share risk and reward depend on what can be observed.

Give leaders an aggregate picture of where knowledge is strong and where it is thin. Keep individual reflection out of evaluation, and spread interpretation across the group instead of handing it down.

How do we manage development strategically without distorting it? As competencies mature, the organization wants to plan time, allocate people, and see where expertise is thin. But when leaders can see individual reflections and use them to rank people, practitioners may shape their accounts to demonstrate success. Once the desired outcome is known anywhere in the system, accounts drift toward it and the data no longer shows what happened (Goodhart's law: a measure that becomes a target stops being a good measure) [67]. Leaders and programs also tend to interpret apprentices' experiences for them, so apprentices never build the language to do it themselves.

Measures influence how people act, especially when they become targets. Using private reflection to rank people can encourage accounts that demonstrate success rather than support inquiry. WGL gives leadership an aggregate picture of organizational learning. Understanding that learning requires evidence at several connected levels: how individuals develop judgment, how mentors and apprentices interact, and how communities sustain and improve practice. Individual feedback and judgments of competence support development and progression. Organizational evaluation examines how these developments contribute to shared capacity and outcomes.

Protecting exploratory reflection is a WGL design commitment. Aggregated accounts indicate participants' perceptions and reported experiences; they are not direct measurements of organizational knowledge. Interpret them alongside observed performance, work records, and other evidence. Review whether small groups or distinctive stories could identify individuals.

1. 1
   Write the data-use rules down first. Say who sees what, and commit to never ranking individuals. Narratives are shared only with the consent of the person who wrote them. Set a minimum group size for any aggregate view, and review distinctive stories and roles that could identify someone. Suppress or combine results when needed, and tell participants the limits of anonymity.
2. 2
   Build a knowledge map from two perspectives. Apprentices mark the knowledge they think they need to develop. Mentors mark the knowledge each person needs for the competency.
3. 3
   Read the aggregate. Together the two views show how the organization perceives expertise development at both levels. Check the picture against observed performance and work records.
4. 4
   Measure the shop, never rank the people. Leadership sees aggregate data only.
5. 5
   Distribute interpretation through narrative sessions. Open regular collective reflection with members' own narratives, using anecdote-circle prompts [4, 64]. Draw rules of thumb and practices from the stories, keep each linked to the stories it came from, and feed patterns, not individual stories, into the knowledge map.

Two spreadsheet columns per competency, one for apprentice and one for mentor, will produce a first map.

Sfard (1998) [60] warns that the acquisition metaphor invites sorting people by measured "potential," while participation describes actions rather than labeling people. Evaluation and reporting arrangements may influence what people do and disclose. Strathern (1997) [67] highlights risks associated with turning indicators into targets. Rosen and Tesser (1970) [58] examine reluctance to communicate undesirable information. Historical Hawthorne findings (Roethlisberger & Dickson, 1939) [55] require caution because later analysis challenges common interpretations (Levitt & List, 2011) [46]. WGL should investigate these risks in its own setting rather than assume their presence or magnitude.

Boisot's (1998) [5] I-Space gives a way to map where knowledge sits and how codified it is. Snowden (2020) [64] treats narrative as an abductive act of knowing, and Blignaut (2007) [4] describes anecdote circles for eliciting stories from groups with shared experience.

Fund the program from operations and build its case on organizational knowledge, as risk, opportunity, or both, so it holds up today. Spread mentoring so no single person holds the program up. WGL makes organizational capability a primary strategic outcome while examining individual development, relationships, and community practice as connected levels of explanation.

Why do apprenticeship programs fail to sustain, and what keeps one alive? A program that reports only participation, completion, or credentials may leave its contribution to organizational outcomes unclear. Without that connection, its business case remains vulnerable to changes in funding or leadership. WGL connects individual development to the organization's capacity to perform and sustain its work. It also addresses the practical concerns of mentors, apprentices, and leaders about the effort they contribute and the benefits they receive.

WGL makes organizational knowledge and capacity the primary focus of its strategic evaluation. These develop through people's participation, their relationships, and the tools and practices they use. Explaining that development requires attention to individual learning, interpersonal interaction, and community practice together. The boundaries of a community of practice may also cross departments or organizations. WGL therefore connects evidence of developing expertise and changing practice to organizational outcomes, including reduced knowledge risk and greater capacity, while recognizing that learning occurs across several connected settings and levels.

1. 1
   Make the organization the primary unit of analysis from the start. The business case and decision map in Section 1 set the baseline.
2. 2
   Ask about knowledge, not people. "What knowledge does this decision need?" lets you grow it in more places than "who holds it?"
3. 3
   Distribute mentoring across a council. Reliance on one veteran is a risk the program should reduce.
4. 4
   Explore degree pathways. Work with higher education's work-based learning teams on pathways that recognize demonstrated learning under the partner institution's assessment and credit requirements. Evaluate completion, employee mobility, and retention rather than assuming a retention benefit.
5. 5
   Revisit the case as data accumulates. Update it with what the narratives and the knowledge map now show.

A short memo to the finance lead, with a number for the cost of lost knowledge and a number for the backlog, can get you started.

Rogoff (1995) [56] examines sociocultural activity through inseparable personal, interpersonal, and community planes. This provides a way to connect apprentice development, mentoring relationships, and collective practice without treating any one as a complete explanation. Brown and Duguid (1991) [9] describe organizations as communities of communities, while Fuller and colleagues (2005) [27] show how the organization of work shapes opportunities for participation and learning. Boisot (1998) [5] and DeLong (2004) [20] help frame the organizational consequences of developing or losing expertise.

75 references. Numbers match the bracketed citations throughout the framework.

1. [1]Abrams, L. C., Cross, R., Lesser, E., & Levin, D. Z. (2003). Nurturing interpersonal trust in knowledge-sharing networks. Academy of Management Executive, 17(4), 64–77. https://doi.org/10.5465/AME.2003.11851845
2. [2]Allen, T. D., Eby, L. T., & Lentz, E. (2006). Mentorship behaviors and mentorship quality associated with formal mentoring programs: Closing the gap between research and practice. Journal of Applied Psychology, 91(3), 567–578. https://doi.org/10.1037/0021-9010.91.3.567
3. [3]Billett, S. (2001). Learning through work: Workplace affordances and individual engagement. Journal of Workplace Learning, 13(5), 209–214. https://doi.org/10.1108/EUM0000000005548
4. [4]Blignaut, S. (2007). Using Cognitive Edge methods for knowledge creation and collective sense-making. In A. S. Kazi, L. Wohlfart, & P. Wolf (Eds.), Hands-on knowledge co-creation and sharing: Practical methods and techniques (pp. 17–41). KnowledgeBoard.
5. [5]Boisot, M. H. (1998). Knowledge assets: Securing competitive advantage in the information economy. Oxford University Press.
6. [6]Boisot, M. H., MacMillan, I. C., & Han, K. S. (2007). Explorations in information space: Knowledge, agents, and organization. Oxford University Press.
7. [7]Borgatti, S. P., & Cross, R. (2003). A relational view of information seeking and learning in social networks. Management Science, 49(4), 432–445. https://doi.org/10.1287/mnsc.49.4.432.14428
8. [8]Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educational Researcher, 18(1), 32–42. https://doi.org/10.2307/1176008
9. [9]Brown, J. S., & Duguid, P. (1991). Organizational learning and communities-of-practice: Toward a unified view of working, learning, and innovation. Organization Science, 2(1), 40–57.
10. [10]Bruner, J. (1986). Actual minds, possible worlds. Harvard University Press.
11. [11]Clark, A., & Chalmers, D. J. (1998). The extended mind. Analysis, 58(1), 7–19. https://doi.org/10.1093/analys/58.1.7
12. [12]Collins, A., Brown, J. S., & Newman, S. E. (1989). Cognitive apprenticeship: Teaching the craft of reading, writing, and mathematics. In L. B. Resnick (Ed.), Knowing, learning, and instruction: Essays in honor of Robert Glaser (pp. 453–494). Lawrence Erlbaum Associates.
13. [13]Crandall, B., Klein, G., & Hoffman, R. R. (2006). Working minds: A practitioner's guide to cognitive task analysis. MIT Press.
14. [14]Cross, R., Borgatti, S. P., & Parker, A. (2002). Making invisible work visible: Using social network analysis to support strategic collaboration. California Management Review, 44(2), 25–46.
15. [15]Davis, B. (1995). Why teach mathematics? Mathematics education and enactivist theory. For the Learning of Mathematics, 15(2), 2–9.
16. [16]Davis, B. (2008). Complexity and education: Vital simultaneities. In M. Mason (Ed.), Complexity theory and the philosophy of education (pp. 46–61). Wiley-Blackwell.
17. [17]Davis, B., & Sumara, D. (2006). Complexity and education: Inquiries into learning, teaching, and research. Lawrence Erlbaum Associates.
18. [18]Deacon, T. W. (1997). The symbolic species: The co-evolution of language and the brain. W. W. Norton.
19. [19]de la Croix, D., Doepke, M., & Mokyr, J. (2018). Clans, guilds, and markets: Apprenticeship institutions and growth in the pre-industrial economy. The Quarterly Journal of Economics, 133(1), 1–70. https://doi.org/10.1093/qje/qjx026
20. [20]DeLong, D. W. (2004). Lost knowledge: Confronting the threat of an aging workforce. Oxford University Press.
21. [21]Dobrow, S. R., Chandler, D. E., Murphy, W. M., & Kram, K. E. (2012). A review of developmental networks: Incorporating a mutuality perspective. Journal of Management, 38(1), 210–242.
22. [22]Doll, W. E. (2008). Complexity and the culture of curriculum. In M. Mason (Ed.), Complexity theory and the philosophy of education (pp. 181–203). Wiley-Blackwell.
23. [23]Eby, L. T., Allen, T. D., Evans, S. C., Ng, T., & DuBois, D. L. (2008). Does mentoring matter? A multidisciplinary meta-analysis comparing mentored and non-mentored individuals. Journal of Vocational Behavior, 72(2), 254–267.
24. [24]Eby, L. T., Allen, T. D., Hoffman, B. J., Baranik, L. E., Sauer, J. B., Baldwin, S., Morrison, M. A., Kinkade, K. M., Maher, C. P., Curtis, S., & Evans, S. C. (2013). An interdisciplinary meta-analysis of the potential antecedents, correlates, and consequences of protégé perceptions of mentoring. Psychological Bulletin, 139(2), 441–476.
25. [25]Executive Order No. 14278. (2025). Preparing Americans for high-paying skilled trade jobs of the future. Federal Register, 90(80), 17525–17527.
26. [26]Fedorenko, E., Piantadosi, S. T., & Gibson, E. A. F. (2024). Language is primarily a tool for communication rather than thought. Nature, 630(8017), 575–586. https://doi.org/10.1038/s41586-024-07522-w
27. [27]Fuller, A., Hodkinson, H., Hodkinson, P., & Unwin, L. (2005). Learning as peripheral participation in communities of practice: A reassessment of key concepts in workplace learning. British Educational Research Journal, 31(1), 49–68.
28. [28]Gibson, J. J. (1979). The ecological approach to visual perception. Houghton Mifflin.
29. [29]Holmström, B. (1979). Moral hazard and observability. Bell Journal of Economics, 10(1), 74–91.
30. [30]Hutchins, E. (1995). Cognition in the wild. MIT Press.
31. [31]Jacobs, V. R., Lamb, L. L. C., & Philipp, R. A. (2010). Professional noticing of children's mathematical thinking. Journal for Research in Mathematics Education, 41(2), 169–202. https://doi.org/10.5951/jresematheduc.41.2.0169
32. [32]Juarrero, A. (2023). Context changes everything: How constraints create coherence. MIT Press. https://doi.org/10.7551/mitpress/14630.001.0001
33. [33]Kauffman, S. A. (2000). Investigations. Oxford University Press.
34. [34]Kemmis, S., & McTaggart, R. (2005). Participatory action research: Communicative action and the public sphere. In N. K. Denzin & Y. S. Lincoln (Eds.), The Sage handbook of qualitative research (3rd ed., pp. 559–603). Sage.
35. [35]Klein, G. (1998). Sources of power: How people make decisions. MIT Press.
36. [36]Klein, G., & Baxter, H. C. (2006, December). Cognitive transformation theory: Contrasting cognitive and behavioral learning [Paper presentation]. Interservice/Industry Training, Simulation and Education Conference, Orlando, FL.
37. [37]Klein, G. A., Calderwood, R., & MacGregor, D. (1989). Critical decision method for eliciting knowledge. IEEE Transactions on Systems, Man, and Cybernetics, 19(3), 462–472.
38. [38]Knowles, M. S. (1980). The modern practice of adult education: From pedagogy to andragogy (Rev. ed.). Cambridge Adult Education.
39. [39]Korte, R. F. (2009). How newcomers learn the social norms of an organization: A case study of the socialization of newly hired engineers. Human Resource Development Quarterly, 20(3), 285–306. https://doi.org/10.1002/hrdq.20016
40. [40]Korte, R., & Lin, S. (2013). Getting on board: Organizational socialization and the contribution of social capital. Human Relations, 66(3), 407–428. https://doi.org/10.1177/0018726712461927
41. [41]Krackhardt, D., & Hanson, J. R. (1993). Informal networks: The company behind the chart. Harvard Business Review, 71(4), 104–111.
42. [42]Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77(6), 1121–1134.
43. [43]Kuhn, T. S. (2012). The structure of scientific revolutions (50th anniversary ed.). University of Chicago Press.
44. [44]Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge University Press.
45. [45]Levin, D. Z., & Cross, R. (2004). The strength of weak ties you can trust: The mediating role of trust in effective knowledge transfer. Management Science, 50(11), 1477–1490. https://doi.org/10.1287/mnsc.1030.0136
46. [46]Levitt, S. D., & List, J. A. (2011). Was there really a Hawthorne effect at the Hawthorne plant? An analysis of the original illumination experiments. American Economic Journal: Applied Economics, 3(1), 224–238.
47. [47]Lewin, K. (1946). Action research and minority problems. Journal of Social Issues, 2(4), 34–46. https://doi.org/10.1111/j.1540-4560.1946.tb02295.x
48. [48]Lobato, J., Hohensee, C., & Rhodehamel, B. (2013). Students' mathematical noticing. Journal for Research in Mathematics Education, 44(5), 809–850.
49. [49]Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2024). Dissociating language and thought in large language models. Trends in Cognitive Sciences, 28(6), 517–540. https://doi.org/10.1016/j.tics.2024.01.011
50. [50]Midgley, M. (2001). Science and poetry. Routledge.
51. [51]Militello, L. G., & Hutton, R. J. B. (1998). Applied cognitive task analysis (ACTA): A practitioner's toolkit for understanding cognitive task demands. Ergonomics, 41(11), 1618–1641.
52. [52]Peirce, C. S. (1998). The nature of meaning. In Peirce Edition Project (Ed.), The essential Peirce: Selected philosophical writings, Vol. 2 (1893–1913) (pp. 208–225). Indiana University Press. (Original lecture delivered 1903.)
53. [53]Polanyi, M. (1966). The tacit dimension. Doubleday.
54. [54]Ragins, B. R., Cotton, J. L., & Miller, J. S. (2000). Marginal mentoring: The effects of type of mentor, quality of relationship, and program design on work and career attitudes. Academy of Management Journal, 43(6), 1177–1194.
55. [55]Roethlisberger, F. J., & Dickson, W. J. (1939). Management and the worker. Harvard University Press.
56. [56]Rogoff, B. (1995). Observing sociocultural activity on three planes: Participatory appropriation, guided participation, and apprenticeship. In J. V. Wertsch, P. del Río, & A. Alvarez (Eds.), Sociocultural studies of mind (pp. 139–164). Cambridge University Press.
57. [57]Rooney, D., & Boud, D. (2019). Toward a pedagogy for professional noticing: Learning through observation. Vocations and Learning, 12(3), 441–457. https://doi.org/10.1007/s12186-019-09222-3
58. [58]Rosen, S., & Tesser, A. (1970). On reluctance to communicate undesirable information: The MUM effect. Sociometry, 33(3), 253–263.
59. [59]Schatzberg, E. (2012). From art to applied science. Isis, 103(3), 555–563. https://doi.org/10.1086/667979
60. [60]Sfard, A. (1998). On two metaphors for learning and the dangers of choosing just one. Educational Researcher, 27(2), 4–13. https://doi.org/10.3102/0013189X027002004
61. [61]Sfard, A. (2009). Moving between discourses: From learning-as-acquisition to learning-as-participation. AIP Conference Proceedings, 1179, 55–58. https://doi.org/10.1063/1.3266753
62. [62]Snowden, D. J. (2004). Organic knowledge management: Part I — The ASHEN model: An enabler of action [Revised version]. https://cdn.cognitive-edge.com/wp-content/uploads/sites/2/2020/11/16123749/7-Organic-KM-1-of-3-ASHEN.pdf (Original work published 2000 in Knowledge Management, 3[7].)
63. [63]Snowden, D. (2017, August 27). The knotty issue of constraints [Blog post]. Cognitive Edge (now The Cynefin Company).
64. [64]Snowden, D. J. (2020). Narrative as abductive acts of knowing [Blog post]. The Cynefin Co. https://thecynefin.co/narrative-as-abductive-acts-of-knowing/
65. [65]Snowden, D. (2024). As through a glass darkly: A complex systems approach to futures. In R. Poli (Ed.), Handbook of futures studies (pp. 48–65). Edward Elgar.
66. [66]Snowden, D. J., & Boone, M. E. (2007). A leader's framework for decision making. Harvard Business Review, 85(11), 68–76.
67. [67]Strathern, M. (1997). 'Improving ratings': Audit in the British university system. European Review, 5(3), 305–321.
68. [68]Tyler, R. W. (1949). Basic principles of curriculum and instruction. University of Chicago Press.
69. [69]Underhill, C. M. (2006). The effectiveness of mentoring programs in corporate settings: A meta-analytical review of the literature. Journal of Vocational Behavior, 68(2), 292–307.
70. [70]Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. MIT Press.
71. [71]Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes (M. Cole, V. John-Steiner, S. Scribner, & E. Souberman, Eds.). Harvard University Press.
72. [72]Wenger, E. (1998). Communities of practice: Learning, meaning, and identity. Cambridge University Press.
73. [73]Wenger-Trayner, E., & Wenger-Trayner, B. (2015). Introduction to communities of practice: A brief overview of the concept and its uses.
74. [74]Wiggins, G., & McTighe, J. (2005). Understanding by design (Expanded 2nd ed.). ASCD.
75. [75]Zahavy, T. (2026). Position: LLMs can't jump. In Proceedings of the 43rd International Conference on Machine Learning (PMLR 306, pp. 172863–172872). https://proceedings.mlr.press/v306/zahavy26a.html