Learning Beyond School

Learning does not respect the borders of a campus. People learn in workplaces, families, libraries, studios, online communities, volunteer organizations, museums, religious communities, open-source projects, and the practical demands of daily life. The argument of this chapter is that institutions should examine relevant evidence from these settings instead of excluding it merely because they did not supervise its production.

Recognition outside school must avoid two mistakes. The first is to treat informal learning as automatically authentic and formal learning as automatically artificial. Workplaces and communities can reproduce narrow routines, misinformation, and unequal access. Formal study can create protected time, conceptual structure, expert critique, and encounters unavailable in ordinary experience. The second mistake is to translate every human activity into a microcredential. Not all valuable learning needs to become a labor-market signal.

The useful middle is portable evidence. A learner should be able to present artifacts, demonstrations, verified responsibilities, explanations, and assessments produced across settings. An institution can then evaluate the evidence against a visible claim without pretending that every source is equivalent. Portability requires shared descriptors, provenance, context, and criteria, not a universal score.

Workplace learning reveals why transfer matters. A person must select knowledge under time, social, safety, and material constraints. Supervisors observe reliability, judgment, communication, and improvement over repeated performance. Yet workplace evidence can be uneven, biased, and difficult to carry to another employer. Structured observation and learner-owned records can make it more useful without turning every interaction into surveillance.

Remote learning expands access while exposing infrastructure. A U.S. Department of Education meta-analysis found that blended and online conditions sometimes outperformed face-to-face conditions, but much of the advantage was entangled with additional time, resources, and instructional differences. ¹ The lesson is not that online is superior. Modality is a bundle of design choices. Effective remote learning still requires active work, feedback, relationships, usable tools, and a place and schedule in which participation is possible.

Connectivity remains a global gate. The ITU estimated that almost three-quarters of the world’s population was online in 2025, while 2.2 billion people remained offline and affordability and quality gaps persisted. ² A mobile signal is not the same as a reliable device, private study space, accessible content, adequate bandwidth, digital skill, or time. Remote-first policy that ignores these conditions can widen the opportunity gap while reporting expanded reach.

Global learning also requires language and cultural plurality. Translating words is not the same as translating examples, prior knowledge, institutional expectations, or legal context. Open resources can lower cost and enable adaptation, but communities need authority and capacity to revise them. A global commons is not a single curriculum exported everywhere; it is shared infrastructure that supports local authorship and reciprocal exchange.

Leaders should build bridges among settings. Schools can validate workplace projects. Employers can state capabilities rather than degree filters. Libraries can host assessment and mentoring. Community organizations can contribute evidence without becoming credential factories. Learners can own a record that points back to work rather than requiring every new institution to trust the last institution’s brand.

Responsibility is shared, even outside school

Learning beyond school does not eliminate a learner’s responsibilities. It changes the conditions under which those responsibilities are exercised. My starting point is reciprocal: learners should engage honestly with the work, and institutions should make the work’s purpose, demands, and assessment intelligible. Neither independence nor accountability requires pretending that everyone begins with the same resources.

A learner should know what they are trying to learn, what evidence will count, what assistance is permitted, and how to ask for feedback. Those expectations should be stated in ordinary language. “Take ownership” is not enough when the learner does not know how to interpret an unfamiliar task. The institution owes more than an instruction to be motivated.

Self-regulation provides a useful framework for planning, monitoring, and reflecting on work. Zimmerman describes these processes as part of learning rather than an all-or-nothing personality trait. ³ I recommend teaching them in the context of a real task: estimate what is required, attempt the work, inspect the result, and adjust the plan. A digital planner can record that process; it cannot establish that the process is effective.

The responsibilities of an adult learner with a job, a young learner studying at home, and an apprentice on a worksite will differ. The standard should focus on the intended capability and on legitimate safety and integrity requirements. A schedule may be necessary for a collaborative task; it may be irrelevant to whether someone can independently analyze a document. Leaders should explain the reason for a demand rather than defend it simply because it is familiar.

Consider a hypothetical learner who repeatedly submits work late. The evidence establishes lateness, not its cause. A useful conversation would ask what the task required, what the learner understood, when time was available, and where the plan broke down. The next step might involve instruction, an adjusted arrangement, or a clearer boundary. The example does not assume a diagnosis or a successful intervention.

Academic honesty also needs an operational meaning. A learner should identify contributions from other people and tools, represent sources accurately, and distinguish their work from assistance. The assessor should provide tasks capable of revealing the relevant understanding. If an assignment can be completed by submitting unexamined generated text, the institution cannot repair the evidence merely by asking students to promise they learned.

A workable agreement might contain four items: the current learning aim, the next piece of work, the available support, and a date for reviewing what happened. Its purpose is not to generate an elaborate record of compliance. It is to make a next action possible and an eventual judgment fair. When the agreement fails, investigate before assigning a trait such as laziness or dependency.

The institution remains accountable for promises made to learners. If it offers feedback, someone must provide it. If it permits flexible pacing, assessment arrangements must make that flexibility real. If it accepts outside evidence, it must explain how that evidence is judged. Responsibility becomes credible when it runs in both directions.

Independence is a practice, not a forecast

I do not assume that every educational system is moving toward the same technological future. The practical question is what learners need when information, assistance, and assessment travel across settings. My answer is the ability to judge a source, plan work, seek help, explain a decision, and demonstrate what remains when assistance is removed.

Technology can offer material, feedback, and possible next steps. It cannot relieve the learner or institution of deciding whether the resulting evidence supports the intended claim. Chapter 11 examines AI interventions that produced promising results alongside evidence that unrestricted assistance can damage later unaided performance. The implication here is procedural: distinguish practice with support from a demonstration of independent capability.

An independent learner is not someone who refuses help. Independence includes recognizing when expertise is needed and understanding what the help contributes. A person may ask for an explanation, compare two worked approaches, or request feedback on a draft without surrendering authorship. The test is whether they can account for the important decisions and undertake the relevant work under the conditions their claim requires.

Competency-based progression is one proposed way to make capability more visible. In this book, it means requiring evidence against published criteria instead of treating elapsed time as sufficient. That is a definition and a policy proposal, not proof that every competency-based program is better. Weak criteria, narrow tasks, or unreliable judgments can undermine the proposal just as they can undermine a time-based course.

Interdisciplinary work raises the same issue. A project connecting computing and public health should make clear which disciplinary claims the learner is qualified to make. An attractive prototype does not validate a medical prediction. Where safety or professional competence is involved, qualified supervision and appropriate evaluation remain necessary. “Interdisciplinary” cannot become permission to skip the standards of either discipline.

Learners also need to understand the social terms of collaboration. Who owns a shared artifact? What may be disclosed? Which parts can be reused? How are contributions acknowledged? I would teach those questions through actual agreements and review of actual work. A group project does not automatically create intercultural understanding or ethical judgment simply because members live in different places.

The future-facing responsibility of institutions is to preserve intelligible evidence as tools change. That may require collecting an explanation, a supervised demonstration, a revision history, or a second task under different conditions. It does not require collecting every keystroke or monitoring a learner’s face. Choose the least intrusive evidence that can responsibly address the claim.

A learner should leave an educational experience with more than a record of completion: an understanding of what they can do, the limits of the evidence, and a route for further learning. That is the version of adaptability I defend. It is a practical commitment to revision, not a prediction that every new platform will make learners self-directed.

Public places as learning infrastructure

A library, museum, or archive offers a different encounter with knowledge from a conventional course. A collection can make an original object, document, interpretation, or specialist resource available for inquiry. The educational opportunity lies in what the learner can investigate. It should not be assumed that access alone produces critical thinking, higher grades, or a lifelong love of learning.

There is a useful named experiment to replace generic success stories. Erickson, Watson, and Greene studied repeated culturally enriching field trips with randomly assigned fourth- and fifth-grade students. The published report describes benefits on several academic, attendance, behavioral, and tolerance outcomes. It did not find effects on empathy, social perspective taking, or desire to participate in arts. ⁴ That mixed pattern supports taking cultural opportunities seriously while refusing to promise improvement in every domain.

The study is not evidence that every museum visit has the same effects, and it says nothing by itself about every library or archive program. A single field trip, a sequence of visits, a guided inquiry, and an unstructured outing are different experiences. Leaders should describe the actual design before borrowing an outcome from research.

For a library partnership, I recommend beginning with a specific inquiry need. Perhaps learners need to compare sources unavailable in their usual materials, practice a search strategy, or obtain help interpreting a catalog. Ask the library what support is feasible and whether learners can access the resource after the assigned visit. A partnership exists in the working arrangement, not just in the announcement.

For a museum project, ask what the object or exhibit permits learners to notice that a summary does not. An assignment might require distinguishing a label’s interpretation from an observation, comparing two representations, or explaining what evidence is absent. Accessibility should be built into the task: the institution should consider sensory access, physical routes, language, cost, and alternatives when a visit is not possible.

For an archive, the intellectual challenge includes provenance and incompleteness. A surviving letter is a document produced for a purpose by a particular person, not a transparent window onto an entire event. A hypothetical exercise could ask students to compare the document with another account and identify what neither source establishes. The aim is a defensible interpretation, not the appearance of authenticity conferred by handling an old object.

Students should know what will count as evidence of learning before they go. The task might culminate in an explanation, a source comparison, a catalog entry, or a design response. It need not end in a generic reflection claiming that the experience “opened their minds.” A learner should be permitted to find an exhibit unpersuasive and explain why.

Public institutions also have purposes beyond serving school assessment. Leaders should respect staff expertise, collection rules, community ownership, and capacity. A school should not assume that its curriculum takes priority over all other users. The proposed partnership should identify responsibilities, costs, permissions, and what happens when plans change.

My recommendation is to treat public cultural resources as part of a broad opportunity infrastructure. Their artistic, historical, intellectual, and communal value does not need to be converted entirely into academic scores. Where an educational outcome is claimed, evaluate it. Where the purpose is access to a worthwhile human practice, state that purpose without inventing an additional effect.

Support without taking over

Family involvement is not one behavior. Encouragement, pressure, help with a difficult task, attendance at a meeting, provision of resources, and respect for a learner’s initiative are not interchangeable. A recommendation to increase involvement is incomplete until it identifies the kind of involvement, the need it addresses, and whose judgment it respects.

Wang and Wei’s 2024 meta-analysis found a positive association between parental involvement and mathematics performance, with differences across involvement types and contexts. The evidence is correlational; the review also notes inconsistent classification and limits in the populations and literature represented. ⁵ It does not establish that any additional parental action raises achievement.

Xu and colleagues’ homework-focused synthesis offers an important qualification. Across its included studies, the overall association between involvement and achievement was weakly negative, while autonomy-supportive involvement was positively associated with achievement. Direct content support, control, and frequency did not show the same positive pattern. These are associations, not proof that help causes harm: existing difficulty can prompt adults to become more involved. ⁶

The apparent contrast is useful because the reviews ask different questions. General involvement in mathematics and specific involvement in homework are not identical exposures. Leaders should not select whichever pooled result supports a preferred message to families. The defensible conclusion is that the form and context of support matter, and that simple quantity is a poor guide.

My recommendation is to give families concrete, optional ways to support a learner without taking authorship away. Ask what the learner is trying to do, invite them to explain a choice, help locate an appropriate resource, and contact the educator when the task or expected help is unclear. A parent need not become an unpaid subject specialist to demonstrate commitment.

A hypothetical homework conversation can make the difference visible. An adult might ask, “Which part have you tried, and where did you get stuck?” rather than immediately completing the exercise. But the question is not a universal script. A learner may need direct instruction, an accessible version of the material, or help from someone with the relevant expertise. Autonomy support is not abandonment disguised as independence.

Schools must be explicit about the help an assignment assumes. If a project requires equipment, transport, money, or specialized knowledge, the institution should provide a fair route to those resources or redesign the task. Otherwise the resulting artifact partly assesses household provision while the grade is presented as the child’s capability.

Family goals also deserve discussion rather than simple endorsement. An adult’s aspiration may be meaningful and generous while differing from the learner’s own aims. Educators should explain options, prerequisites, costs, and uncertainties. They should avoid promising a particular career, diagnosing a fixed aptitude from an early score, or treating disagreement as a lack of ambition.

Communication should be usable. Families need understandable information, appropriate language access, realistic meeting arrangements, and a clear route for raising concerns. An online portal is one possible channel, not proof that communication is effective. Ask whether the message was received, understood, and actionable before concluding that an absent response means indifference.

The purpose of parental involvement, in this account, is not to maximize pressure or produce a polished proxy for the child’s work. It is to support the learner’s growing capability while keeping responsibility for instruction and assessment visible. Families matter; so do the limits of what schools may fairly ask of them.

Workplace transfer needs its own evidence

Workplace training and academic retention are related questions, not interchangeable outcomes. A worker can become proficient in a local procedure without retaining an unrelated school subject. A person can remember a principle yet fail to recognize when it applies under workplace constraints. The evidence must follow the specific capability across the change in setting.

Blume and colleagues’ meta-analysis of training transfer found associations with motivation, learner characteristics, and a supportive work environment. The relationships varied by the kind of skill and by how transfer was measured. The authors identified limited evidence on maintenance over time and noted that using the same source to rate predictors and outcomes could inflate associations. ⁷

This is a reason to examine both training and opportunity to use it. It is not evidence that participation in work automatically preserves academic knowledge, or that employees from a particular culture retain more. Those claims would require direct evidence about the task, the learner, the practice, and the follow-up measure.

I recommend beginning a training agreement with the work itself. What should the person be able to do? What instruction and supervised practice will be provided? Under what conditions may they perform independently? What would count as a meaningful mistake, and how will it be reviewed? A completion certificate should not stand in for answers to these questions.

Consider a hypothetical technician learning a diagnostic procedure. A training exercise may establish that the technician can follow the steps when the fault is already identified. A later task might ask them to select the procedure from several alternatives and explain why it is appropriate. A further observation could examine use under ordinary working conditions. These are distinct samples of performance, not assumed stages of improvement.

Retention requires a time interval and a relevant check. If a leader wants to claim that training preserves a capability, the evaluation should include later evidence of that capability. It should state whether the person had opportunities to practice in between and whether assistance was available at assessment. Repeating a familiar exercise immediately after instruction answers a narrower question.

Workplace records also need context. A supervisor can describe a responsibility and an observed performance, but their judgment may not generalize beyond the work they saw. A learner-owned portfolio should therefore state the task, conditions, criteria, role, and verifier. It should avoid converting a local supervisor’s impression into a universal rating of reliability or potential.

Cultural responsiveness belongs in the design without becoming cultural stereotyping. Ask learners about language, prior experience, access, and how they understand the task. Do not assign a learning method because of an assumed national or family trait. Explain conventions necessary for the work, and separate those conventions from preferences that the organization could change.

There is a legitimate tension between portable evidence and confidentiality. An employee may be unable to publish a client record, a proprietary design, or an incident report. Leaders can arrange a redacted artifact, a structured verification, or an equivalent demonstration where appropriate. The record should disclose its limits rather than expose protected information for the sake of a more impressive portfolio.

The educational opportunity is substantial: work can provide meaningful problems and repeated performance under real constraints. But “real” does not mean automatically valid, unbiased, or transferable. Institutions should recognize the evidence while asking the same questions they would ask of a course: what was learned, what was demonstrated, what assistance mattered, and what remains unknown.

Remote is a delivery condition, not a pedagogy

Remote learning separates at least some teaching and learning from a shared physical classroom. It may use live meetings, materials accessed at different times, printed resources, or combinations of these. Blended learning includes both online and in-person elements. These descriptions identify arrangements; they do not establish quality.

Yu and colleagues’ meta-analysis reports a positive average learning-performance result for blended instruction across the included studies, with variation by design and context. ⁸ That is evidence about blended programs, not proof that fully remote learning is superior or that the medium alone produced the difference. It should be read alongside the older US review’s warning about additional time and instructional differences. ¹

For leadership decisions, I would disassemble the program rather than debate “online versus classroom” in the abstract. Identify the material learners encounter, the work they perform, the feedback they receive, the interaction they need, and the assessment used to establish capability. Then ask which elements require shared time, which require shared place, and which can be provided more flexibly.

A live online seminar can permit immediate questioning, but it still requires a workable schedule and connection. An asynchronous explanation can be revisited, but a learner needs a way to resolve confusion that the explanation does not address. A blended laboratory may keep physical practice on site while moving preparation elsewhere. None of these designs is automatically best; each has demands that must match the task and population.

The platform should serve the design. Content delivery, messaging, collaborative editing, submission, and feedback are functions, not learning outcomes. An activity log can help identify a technical problem or prompt a conversation. It cannot establish that the learner understood the material, worked independently, or had a usable opportunity to practice and demonstrate the required capability.

Assessment should be planned early. A remote course may need an oral explanation, a supervised practical component, an artifact with provenance, or a task performed under specified assistance conditions. If the intended capability involves physical equipment or direct interpersonal practice, the program must explain how those demands are met. Calling the course digital does not remove them.

Likewise, teachers need a realistic operating model. Who answers questions, reviews work, resolves access problems, and updates materials? Which activities require specialist expertise? How much work is expected, and when? Automation may perform a defined function, but it does not abolish the obligation to check quality or deal with cases the function does not handle.

A proposed remote program should therefore be describable without its brand names. A leader ought to be able to explain its teaching sequence, support, assessment, accessibility, and cost structure in ordinary language. If the case for adoption depends mainly on the platform being adaptive, intelligent, or data-driven, the important educational argument has not yet been made.

This is an argument for evaluating remote provision seriously, not for holding it to an impossible standard while ignoring weaknesses in existing classrooms. Compare complete, realistic alternatives against the same aims. Include the learners who would otherwise have no feasible route, and distinguish an improvement in access from a demonstrated improvement in learning.

Test the promises: access, cost, skills, and carbon

Remote provision has genuine affordances. A learner may avoid a journey, access a specialist course from another location, revisit an explanation, or arrange some work around other responsibilities. These possibilities can matter greatly. They become benefits in a particular program only when the design, resources, and learner’s circumstances make them usable.

Flexibility needs a precise meaning. Does the learner control the time of study, the pace, the location, or merely the time at which a recording is played? A course with fixed live attendance and tightly sequenced deadlines may be geographically remote but inflexible in time. Leaders should describe the actual arrangement instead of allowing the word “flexible” to carry several promises at once.

Cost also needs a boundary. Chirikov and colleagues randomized second-year students in three Russian universities among in-person, blended, and online versions of two STEM courses. Final-exam results were similar across conditions; the online group reported lower satisfaction. The analysis estimated lower instructional costs for the tested model, while acknowledging platform startup, staff development, and coordination requirements. ⁹

That study is a concrete example, not a guarantee of cheaper tuition or lower total learner cost. For a proposed program, I would distinguish what the institution pays from what the learner pays. Include devices, connectivity, software, accessibility, support, assessment, travel still required, and time. A saving in one account may be a transfer of cost into another.

A cost-effectiveness comparison also needs an outcome. Cost per registration, cost per completion, and cost per demonstrated capability answer different questions. An inexpensive course that a learner cannot finish is not automatically a bargain. Conversely, a supported remote route may be worthwhile even when it is not the cheapest option, particularly when the relevant alternative is no usable access at all.

Personalization is another conditional promise. A course can offer alternative explanations or adapt practice to observed responses. Whether those choices improve the intended learning must be evaluated. A recommendation is not a diagnosis, and a smoothly completed sequence is not proof of independent mastery. The AI evidence in chapter 11 should inform the design without turning every adaptive feature into an established intervention.

Remote participation can also require digital and self-management skills. It does not follow that participation automatically teaches them. I would provide explicit orientation, practice with the necessary tools, a route for help, and opportunities to plan and review work. If the program claims a transferable digital capability, assess that capability rather than infer it from login frequency.

Global collaboration has a similar boundary. Connecting participants from different locations makes communication possible; it does not guarantee intercultural understanding. A proposed group task should specify the intellectual purpose, working language, contribution rules, time-zone arrangements, and procedure for resolving disagreement. The work should reveal what participants actually learned from one another, not merely where they lived.

Environmental claims require accounting rather than intuition. Caird and colleagues’ 2015 assessment of UK higher-education courses found lower estimated energy use and carbon emissions in distance-based models than in the campus-based models studied. The analysis considered travel, residential and campus energy, technology, and materials, and identified potential rebound effects. ¹⁰

The age and setting of that analysis matter. It does not establish the footprint of a current course, energy grid, device mix, or travel pattern. Before advertising an environmental benefit, an institution should define its system boundary and examine which activities are actually displaced. Keeping the same buildings operating while adding devices and online activity is a different proposition from replacing travel and additional residence.

The leadership conclusion is positive but conditional. Remote learning can expand usable options, and particular designs have produced credible results. The right response is to test the specific promise, retain the relevant limits, and state whose costs and outcomes were measured. That is more useful than either declaring remote provision transformative by definition or dismissing it because it is not a classroom.

Build the conditions before assigning the responsibility

A remote program should begin with an access check that is more demanding than “Do you have internet?” Ask whether the learner can use the required material, communicate when necessary, complete the task, and obtain feedback. The relevant device, connection, accessibility features, workspace, and time depend on the activity. A text-based assignment and a live technical simulation impose different demands.

I recommend testing a representative learning task before the formal course begins. The learner should be able to locate instructions, open materials, submit a small piece of work, and receive a response. The purpose is to discover barriers while the stakes are low. Completing this orientation does not certify readiness for every subsequent task, but it makes avoidable technical surprises easier to identify.

The institution needs an interruption plan. What happens when a connection fails during a meeting or assessment? How can a learner obtain materials if the platform is unavailable? Which deadlines can move, and who makes that decision? The answers should be visible before an incident, not negotiated only by learners with the confidence or connections to demand an exception.

Study conditions need the same realism. A dedicated quiet office is not a reasonable universal prerequisite for participation. Programs can consider downloadable materials, access to local study spaces, alternative meeting times, and tasks that do not require constant video. The appropriate arrangement will differ, but the institution should not design for an ideal household and treat every deviation as a motivation problem.

Active work is necessary for demonstrating what a learner can do, but “active” should not mean constant clicking or compulsory public performance. A careful written explanation, a worked problem, a critique, a design revision, or a physical demonstration may be appropriate. The task should create evidence relevant to the aim, and feedback should identify a possible next action.

Support needs named responsibilities and response expectations. Learners should know where to ask a subject question, report an access problem, discuss an assessment, or raise a welfare concern. These routes can be simple. Their credibility depends on whether someone can act, not on whether the program has a sophisticated help interface.

Self-management should be scaffolded where needed. A learner might begin with a short plan and a scheduled review, then take more control as they demonstrate that the arrangement works. The institution should not equate reduced contact with greater maturity. More independence should mean a greater ability to make and explain effective decisions, including the decision to seek help.

Evaluation must include the people the design fails to reach. Ask why learners did not begin, why they stopped, and whether the offered support was usable. A high satisfaction score among completers cannot describe everyone who was enrolled or excluded. Nonresponse and missing work should be investigated as limitations in the evidence, not silently counted as proof of learner deficiency.

The principle is straightforward: allocate responsibility to the party that can act on it. Learners can attempt work and communicate difficulties; institutions can change inaccessible materials, unclear rules, and unsupported processes. A remote program worthy of the name “open” should be able to show how both sides are supported and held accountable.

Connection is necessary for some routes, sufficient for none

The ITU’s 2025 estimates describe both the reach of internet access and the remaining gap: almost three-quarters of the world was online, while approximately 2.2 billion people remained offline. The report also identifies affordability and quality disparities. ² These are dated global estimates, not a live count of connected learners or a measure of educational readiness.

The educational question begins after that headline. What can a person actually do with the connection available to them? A shared phone, intermittent signal, expensive data allowance, inaccessible document, or unfamiliar interface may leave a nominally connected learner unable to complete a particular task. A program should investigate those conditions directly rather than infer them from a national penetration rate.

Connectivity statistics cannot establish that online education has produced equality, improved achievement, or removed geographic barriers. They describe an important part of infrastructure. An educational claim requires evidence about participation, support, learning, and the people who remain unable to use the provision.

My recommendation is to design more than one practical route where the task permits it. Low-bandwidth materials, downloadable text, accessible formats, asynchronous questions, and local access points can be considered alongside live interaction. These are options to assess against the learning purpose, not a universal technical prescription. Some tasks genuinely require specialized equipment or timely interaction, and that requirement should be stated honestly.

Infrastructure partnerships also need continuity. A donated device is not a complete access policy if repair, replacement, connectivity, accounts, and support are unresolved. A temporary subsidy should not be described as a permanent solution. Leaders should identify who owns each obligation and what learners can expect when the initial arrangement ends.

Language and local relevance belong in this discussion. A connection to a large collection of resources is not the same as access to an explanation a learner can use. Translation can help with words while leaving examples, assumptions, or requirements mismatched. Communities should have a role in selecting and adapting materials rather than merely receiving a global package.

The same principle limits claims about international reach. A platform may technically accept visitors worldwide while assessment, payment, accessibility, licensing, or legal arrangements restrict actual participation. If a program is advertised as global, explain what is available, to whom, under which conditions. Do not make a map of page views stand in for an account of educational opportunity.

Leaders should evaluate access as a sequence: receiving the invitation, entering the program, using the material, obtaining support, producing work, and having that work recognized. A barrier at any stage may make the earlier stages insufficient. The goal is not to maximize a connectivity statistic inside the institution. It is to make a credible learning route possible for a real person.

Internet access can support that goal, but it does not complete it. The point of expanding connection is to enlarge what people can learn and do, not to declare the gatekeeper problem solved once the login page loads.

Music deserves better than inflated transfer claims

Music belongs in education as a form of knowledge, expression, skilled practice, and cultural participation. A learner can develop a repertoire, interpret a work, compose, listen critically, or coordinate a performance with others. My case for access to these activities does not depend on proving that they raise mathematics scores or confer a general cognitive advantage.

There is newer positive evidence that should not be dismissed. Jamey and colleagues’ 2024 meta-analysis examined music training and inhibitory control in children aged three to eleven. It included 22 longitudinal controlled studies, eight of which combined individual randomization with active control activities. The synthesis found a positive effect on the measured inhibition tasks, including in that randomized subset. ¹¹

The outcome is specific. It is not a finding that music improves every executive function, overall intelligence, mathematics, emotional understanding, or later employment. The included population was neurotypical; recommendations about clinical treatment need clinical evidence beyond this synthesis. Its results also do not show that the measured benefit persists indefinitely or identify one universally best music program.

The broader literature remains contested. Schellenberg and Lima’s 2024 review emphasizes weak or inconclusive evidence for broad causal nonmusical benefits and the importance of preexisting differences between people who do and do not receive training. ¹² The later, narrower inhibition analysis adds support for a particular outcome without settling every transfer question. An honest account can retain both findings.

This is exactly the discipline the book asks of other educational claims. Define the activity, population, comparison, outcome, and timescale. A neural response, a musician’s high grades, and a randomized change in an inhibition task are different kinds of evidence. They should not be placed in a single chain that ends with “music makes people smarter.”

For a school music program, I would begin with musical aims and fair access. What should learners have the opportunity to hear, make, interpret, and understand? Which instruments, approaches, and traditions are available? What support is needed for learners with different physical, sensory, financial, or prior-experience circumstances? These questions are worthwhile even when no far-transfer effect is claimed.

An ensemble provides a concrete setting for accountability. A learner may need to prepare a part, respond to agreed cues, listen to the group, and explain interpretive choices. Those are assessable demands of the activity. If the institution wants to claim improved teamwork outside the ensemble, it should collect evidence in that other setting. Participation alone does not establish the transfer.

The emotional and social experience of music should also be represented carefully. A learner may describe pleasure, frustration, belonging, or relief. Such accounts deserve to be heard without being converted into a clinical claim or a prediction of higher grades. A program can make space for expression and community without promising therapy or prescribing a single emotional response.

Practice habits offer a further teaching opportunity. An educator can help a learner identify a difficult passage, choose a strategy, inspect a recording, and plan the next attempt. The artifact can show musical progress. General time management or academic persistence would need separate evidence if asserted as outcomes. A plausible analogy is not a demonstrated effect.

Famous musicians or scientists are not a substitute for that evidence. A successful person who also played an instrument does not show that the instrument caused the success. Unverified quotations and stories about exceptional figures should not carry a causal argument. The relevant question is what comparison or observation supports the claim being made.

Music does not become less important when its benefits are described accurately. It becomes less vulnerable to a bargain in which access is defended only by effects on someone else’s preferred test. Leaders can support musical opportunity for its direct purposes while also evaluating specific additional outcomes when the evidence warrants it.

Language has value without a superiority claim

Knowing another language can make communication possible with people and materials that would otherwise be inaccessible to the learner. That direct capability is substantial. It can support family relationships, cultural participation, study, work, and the enjoyment of a text or conversation. Which opportunities follow depends on proficiency, context, and what the learner wants to do.

The claim of a general cognitive advantage is a different proposition and remains disputed. Yurtsever, Anderson, and Grundy’s 2023 quantitative analysis found that bilingual children outperformed monolingual children on executive-function tasks more often than expected by chance, using a direction-of-results approach. ¹³ This is affirmative evidence in the debate, not a causal estimate of the effect of taking a language course.

Lowe and colleagues’ 2021 meta-analysis pooled published and unpublished child studies and found a small initial advantage that was indistinguishable from zero after adjustment for publication bias. ¹⁴ The methods and questions differ from the later directional analysis. Reporting both is more informative than declaring that all recent research either proves or disproves a universal bilingual advantage.

The adult literature discussed elsewhere in the book also does not justify a blanket claim of executive superiority. ¹⁵ Nor do these childhood reviews establish delayed cognitive decline in old age, higher income for every multilingual person, or a predictable improvement in school grades. Those are separate outcomes requiring matching evidence.

The educational response I recommend is to define language goals in terms of use. Can the learner understand the relevant material, ask and answer questions, explain an idea, interpret a text, or participate in the conversation they care about? A person may have different strengths in listening, reading, speaking, and writing. One global label should not conceal that profile.

Language learning can invite encounters with other perspectives, but vocabulary does not automatically create empathy. A course should make room for asking about context, noticing ambiguity, and checking understanding. Learners should not be taught that every speaker of a language shares one culture or viewpoint. Communication should enlarge the conversation rather than replace one stereotype with another.

Economic claims require particular care. Proficiency may be necessary for a role, valuable in a particular market, or irrelevant to another job. The book does not promise a universal wage premium or automatic promotion. A learner deserves concrete information about the role’s requirements and the evidence an employer will accept, not a generic assurance that multilingualism always pays.

The same caution applies to products. An app can provide exercises, recorded examples, reminders, or feedback. Those functions are not proof that regular use produces a certification-level capability. Assess the language the learner can actually use, including under conditions that differ from the app’s familiar prompts. A streak records participation in a routine; it does not define proficiency.

A hypothetical learning plan might begin with a specific purpose, such as understanding a family interview or conducting a workplace conversation. The learner could identify recurring language demands, practice with feedback, and attempt a new example after a delay. The plan should allow revision when the chosen material is too difficult, too narrow, or unrelated to the purpose. It is a proposed design, not a promised outcome.

Institutions should also recognize languages learners already know. A home language is not a deficit merely because it differs from the language of instruction. At the same time, a learner may need explicit support for the vocabulary and conventions of a new academic or professional setting. Respect for an existing repertoire and instruction in an additional one are compatible.

Language is therefore a strong example of the evidence-over-proxies argument. Assess the communication and understanding being claimed. Do not require learners to justify linguistic and cultural participation through exaggerated benefits to unrelated abilities. The direct human possibilities are enough to warrant a serious place in education.

A decision exercise: compare complete learning routes

The following is a hypothetical planning exercise, not an account of a program that produced measured gains or savings. Its purpose is to show how a leadership team could make a decision without confusing a delivery mode with its outcomes.

Suppose an institution wants to offer an introductory data-analysis course to learners who combine study with employment or care responsibilities. It is considering two routes: a classroom course with scheduled meetings, and a remote course with asynchronous materials, scheduled support, and an in-person final demonstration where required. Both routes aim at the same capability: interpreting a dataset, checking an analysis, explaining a conclusion, and identifying limitations.

The first task is to specify that capability well enough to assess. Learners should know what kinds of data and tools are in scope, what counts as an acceptable explanation, and which assistance is permitted. The institution should identify a meaningful final task and a later check before choosing a platform. Otherwise it may choose the assessment that is easiest for the purchased system rather than the one the learning claim requires.

The next task is to describe each route as a complete service. For the classroom route, include preparation, instruction, practice, feedback, room access, travel, and assessment. For the remote route, include material production, accessible formats, platform support, learner orientation, feedback, communication, equipment, and assessment. List responsibilities that are shared across routes. A comparison that counts staff time on one side and ignores it on the other is not a fair cost comparison.

Now draw three separate cost accounts: the provider’s, the learner’s, and the wider system’s. The provider may pay for staff, facilities, licensing, maintenance, and administration. The learner may pay for transport, connectivity, equipment, care arrangements, or time away from paid work. The wider system may subsidize some of those costs. A cost does not disappear because it is paid by a family or another public service.

Distinguish startup from recurring expenditure. Producing materials or preparing staff for a new route is not the same as supporting each additional learner. Nor should an optimistic assumption about future enrollment make today’s startup cost invisible. Present alternative enrollment scenarios and show which costs change with scale. The purpose is not to predict one perfect number, but to make the decision’s dependence on assumptions visible.

Then examine access before recruitment. Invite prospective learners to try a representative task and discuss the demands of each route. Ask about scheduling, the required device, accessibility, the place where work would be done, and what happens during an interruption. Do not turn this inquiry into a selection filter that excludes people with difficulties. Use it to identify what support the route must provide or what claim of openness the institution cannot yet make.

Some needs may favor one route; others may favor the alternative. A learner may need access to equipment on site but benefit from remote preparation. Another may have reliable technology but no feasible live meeting time. The planning team should consider whether a common learning standard can coexist with more than one participation arrangement. Uniformity of attendance is not automatically fairness, and unlimited choice is not automatically feasible.

Assessment must preserve comparability without pretending that conditions are identical. Record the assistance, tools, time, and supervision permitted in each demonstration. If one route offers more practice, that difference belongs in the interpretation of its outcomes. If one assessment allows repeated attempts and the other does not, the resulting scores should not be described as though they came from the same conditions.

The evaluation should distinguish immediate performance from retention and transfer. A final task can show capability at the end of instruction. A later task can examine what remains and whether the learner can apply it to a changed dataset. The team should define what counts as a meaningful improvement before seeing results. It should also specify what an uncertain or mixed result would imply for the next decision.

Participation needs its own account. Report who was invited, who enrolled, who began, who completed, and whose outcome evidence is missing. Ask about reasons where feasible and with appropriate consent. If a route works well for those who finish but is unusable for many of the intended learners, that is part of the program’s result. The institution should not solve the problem statistically by discussing completers alone.

The comparison should include burdens as well as benefits. Ask learners and staff about the workload, accessibility, usability, and support they experienced. A demanding task may be educationally justified, while an avoidable administrative obstacle is not. Neither a favorable satisfaction rating nor a favorable test result should erase the other kinds of evidence. The leadership team needs to decide which tradeoffs it is willing to defend.

Environmental accounting requires another explicit boundary. Identify the journeys and accommodation arrangements actually changed by the course. Consider whether rooms remain heated and lit regardless of attendance, whether devices are newly purchased, and whether online activity supplements rather than replaces other activity. Use current, documented conversion assumptions where estimates are made. Do not apply an old study’s average percentage reduction to this new program.

An access improvement can also change the environmental comparison. A course may serve learners who previously had no feasible route, adding activity rather than replacing an existing journey. That does not make the access improvement undesirable. It means the institution must distinguish the ethical case for serving additional people from the separate numerical claim that emissions were reduced. Combining them into one promotional statement conceals the tradeoff.

Data governance should be designed alongside evaluation. Collect what is needed to compare the stated outcomes and investigate access, not everything the platform can record. Explain what will be retained, who can inspect it, and whether participation in optional research affects access to the course. Do not use private study behavior to construct a permanent ranking of learner seriousness.

Before the pilot begins, agree on decision rules. What result would justify continuation? What difficulty would require redesign? Which safety or access failure would require a pause? Who has authority to decide, and how will learners be informed? A pilot without an ending rule can become a permanent program whose evidence is never strong enough to challenge the initial commitment.

Finally, report the decision in language that matches the evidence. The institution might conclude that one route provided a usable option for a specified group at a stated cost, while learning differences remained uncertain. It might find comparable immediate performance but need more information about later use. Or it might identify a support problem serious enough to delay expansion. No outcome is assumed in this exercise.

The lesson is not that every institution needs an elaborate research department before teaching remotely. It is that the ordinary questions of purpose, cost, access, evidence, and responsibility should be answered coherently. Start with a decision small enough to examine, document what was actually provided, and allow the result to change the plan. That is how an institution can learn alongside its learners.

Evidence that can travel

Portable evidence needs five elements: the capability claim, the artifact or observation, the conditions under which it was produced, the criteria applied, and trustworthy provenance. It should also state what remains unknown. A verified project can be powerful evidence of design and execution while saying little about unrelated domains.

Interoperability should protect learner control. Portability cannot mean that every institution gains permanent access to a complete behavioral history. Learners need selective disclosure: the ability to present evidence relevant to the claim without exposing everything collected along the way.

Implications for leaders

• Accept direct evidence from outside the institution against published criteria.
• Create structured, portable records for workplace and community learning.
• Evaluate remote programs as complete designs, not as a delivery mode alone.
• Budget for devices, accessibility, support, space, and time, not connectivity alone.
• Preserve learner ownership and selective disclosure when evidence travels.

Questions to carry forward

1. What valuable learning does your institution currently refuse to see because it happened elsewhere?
2. What context must travel with an artifact for another institution to trust it?
3. Which learners are nominally connected but practically unable to participate?

Notes

1. 1Means, Barbara, Toyama, Yukie, Murphy, Robert, et al.. 2010. “Evaluation of Evidence-Based Practices in Online Learning: A Meta-Analysis and Review of Online Learning Studies”. U.S. Department of Education.↩
2. 2International Telecommunication Union. 2025. “Measuring Digital Development: Facts and Figures 2025”. ITU.↩
3. 3Zimmerman, Barry J.. 2002. “Becoming a Self-Regulated Learner: An Overview”. Theory Into Practice, vol. 41, no. 2, 64–70.↩
4. 4Erickson, Heidi H., Watson, Angela R., Greene, Jay P.. 2024. “An Experimental Evaluation of Culturally Enriching Field Trips”. Journal of Human Resources, vol. 59, no. 3, 879-904.↩
5. 5Wang, Xueshen and Wei, Yun. 2024. “The influence of parental involvement on students’ math performance: a meta-analysis”. Frontiers in Psychology, vol. 15, 1463359.↩
6. 6Xu, Jianzhong, Guo, Shengli, Feng, Yuxiang, et al.. 2024. “Parental Homework Involvement and Students’ Achievement: A Three-Level Meta-Analysis”. Psicothema, vol. 36, no. 1, 1-14.↩
7. 7Blume, Brian D., Ford, J. Kevin, Baldwin, Timothy T., et al.. 2010. “Transfer of Training: A Meta-Analytic Review”. Journal of Management, vol. 36, no. 4, 1065-1105.↩
8. 8Yu, Qing, Yu, Kun, Li, Baomin, et al.. 2025. “Effectiveness of blended learning on students’ learning performance: a meta-analysis”. Journal of Research on Technology in Education, vol. 57, no. 3, 499-520.↩
9. 9Chirikov, Igor, Semenova, Tatiana, Maloshonok, Natalia, et al.. 2020. “Online education platforms scale college STEM instruction with equivalent learning outcomes at lower cost”. Science Advances, vol. 6, no. 15, eaay5324.↩
10. 10Caird, Sally, Lane, Andy, Swithenby, Ed, et al.. 2015. “Design of higher education teaching models and carbon impacts”. International Journal of Sustainability in Higher Education, vol. 16, no. 1, 96-111.↩
11. 11Jamey, Kevin, Foster, Nicholas E.V., Hyde, Krista L., et al.. 2024. “Does music training improve inhibition control in children? A systematic review and meta-analysis”. Cognition, vol. 252, 105913.↩
12. 12Schellenberg, E. Glenn and Lima, César F.. 2024. “Music Training and Nonmusical Abilities”. Annual Review of Psychology, vol. 75, no. 1, 87-128.↩
13. 13Yurtsever, Asli, Anderson, John A.E., Grundy, John G.. 2023. “Bilingual children outperform monolingual children on executive function tasks far more often than chance: An updated quantitative analysis”. Developmental Review, vol. 69, 101084.↩
14. 14Lowe, Cassandra J., Cho, Isu, Goldsmith, Samantha F., et al.. 2021. “The Bilingual Advantage in Children’s Executive Functioning Is Not Related to Language Status: A Meta-Analytic Review”. Psychological Science, vol. 32, no. 7, 1115-1146.↩
15. 15Lehtonen, Minna, Soveri, Anna, Laine, Aini, et al.. 2018. “Is Bilingualism Associated With Enhanced Executive Functioning in Adults? A Meta-Analytic Review”. Psychological Bulletin, vol. 144, no. 4, 394–425.↩