1. Work, mission, and institutional purpose

The word work conceals several different kinds of activity. It may describe effort through which people develop mastery, care for others, create culture, maintain infrastructure, or exercise responsibility. It may also describe labour performed under an employment contract, procedural activity required by an institution, or compensatory effort created by defective systems.

This distinction matters because effort is not inherently a problem. Some effort is constitutive of meaning. Hannah Arendt distinguished labour associated with biological maintenance, durable work that constructs a human world, and action through which people appear to one another politically.²⁸ Harry Braverman later examined how managerial control and deskilling altered the relationship between labour and knowledge.²⁹ David Graeber’s account of “bullshit jobs,” although contested and difficult to operationalise, captured a widespread intuition that some paid roles are experienced as purposeless even when economically sustained.³⁰

The present framework does not attempt to decide from outside whether an occupation is meaningful. It begins at the workflow level and asks what relationship a task has to the declared purpose of the institution.

Human effort can be divided into four functional classes:

• Mission work: directly advances the institution’s purpose.
• Enabling work: provides necessary coordination, maintenance, logistics, or infrastructure.
• Legitimate assurance: protects rights, safety, quality, continuity, accountability, or evidence.
• System-compensating work: exists because information, process, authority, or technology is defective.

Administration is not automatically burden. A clinical record can protect patient continuity; an audit trail can expose abuse; a review gate can prevent irreversible harm. The relevant question is not whether an activity appears bureaucratic. It is whether the activity creates proportionate mission value or compensates for avoidable system failure.

Work becomes system-generated burden when people are repeatedly required to compensate for defects in institutional coordination, information, authority, or technical design without proportionate mission value.

This makes burden mission-relative rather than merely subjective. A frustrating task may still be necessary. A painless task may still be wasteful. The test is functional: what human or institutional good would be lost if the task disappeared, and could that good be preserved through a better design?

The principle that follows is deliberately demanding:

Work must justify its claim on a human life.

That is not a rejection of work. It is a refusal to treat inherited procedures as morally self-justifying.

2. Administrative burden and system debt

Administrative burden is now an established field of public-administration research. Moynihan, Herd, and Harvey define it through learning costs, compliance costs, and psychological costs incurred when people interact with the state.³ Later work shows that these burdens are not evenly distributed: they can restrict access to rights and services most severely for people with fewer resources, less time, or greater need.⁴ Administrative burden is therefore not merely inconvenience. It can become a mechanism of inequality.

The concept can be extended carefully beyond citizen–state interaction. Institutions impose burden internally and externally whenever people must discover requirements, satisfy procedural demands, manage uncertainty, or absorb the psychological cost of opaque systems. OECD work on administrative simplification similarly treats reporting, permits, forms, and notification obligations as measurable costs and uses methods such as the Standard Cost Model to estimate time and expense.⁵

This dissertation introduces system debt as the structural origin of recurring burden:

System debt is the accumulated obligation imposed on people by unresolved defects in information architecture, process design, authority allocation, interoperability, and institutional memory.

Technical debt is one subset of this larger category. A brittle application may create manual workarounds, but system debt can also originate in contradictory policies, unowned processes, duplicate authorities, inaccessible archives, or incentives that reward documentation without use.

The cycle is predictable:

Structural defect
→ coordination or information failure
→ compensatory human action
→ repeated burden
→ normalisation
→ institutional dependence on hidden labour

Once the compensatory labour becomes routine, the institution may cease to perceive the underlying defect. The workaround becomes the workflow. People who remember exceptions, reconcile records, translate terminology, and chase approvals are treated as performing ordinary duties even when their labour exists only because the system has failed to encode its own operating conditions.

Healthcare supplies a strong empirical case. Arndt and colleagues combined electronic health-record logs with time-motion observation and found that primary-care physicians spent nearly two hours on EHR and desk work for every hour of direct clinical face time; the reported workday included substantial clerical, documentation, and inbox activity.¹ A 2024 AHRQ technical brief identified 135 studies and eleven distinct categories for measuring documentation burden, including EHR time, inbox work, after-hours work, administrative tasks, workflow fragmentation, and usability.²

The significance is not that documentation is unnecessary. Clinical records are essential. The evidence shows that the architecture surrounding those records can impose measurable burden and that no single measure captures it completely. System debt is often distributed across time, cognitive load, after-hours spillover, duplicated action, and lost attention.

Meaningless effort is therefore not best defined as work someone dislikes. It is better defined as mission-external or disproportionate human effort generated by a system’s unresolved debt.

3. Documentation, coordination, and context loss

Institutions can possess enormous quantities of information while remaining unable to use it coherently. A fact may have been recorded without being retrievable; retrieved without being interpretable; interpreted without the necessary authority; or acted upon without preserving the reason for action.

The relevant chain is:

Capture → storage → retrieval → interpretation → authority → action

A failure at any stage can force people to reconstruct context that the institution has already encountered.

Documentation serves several legitimate functions. It may be operational, evidentiary, coordinative, memorial, protective, or regulatory. Burden emerges when documentation is collected without becoming usable institutional memory. A form may satisfy a reporting obligation while failing to support future action. A decision may be logged without its assumptions. A case may be handed between teams with its status but not its reasoning. A file may exist in several versions without a canonical state.

Context loss converts prior work into future work. The institution repeatedly pays to rediscover what it once knew, and the payment is usually made in human attention.

This is especially visible in fragmented digital estates. Each application may function correctly within its own boundary while the person remains responsible for remembering the relationships between them. The human becomes the integration layer: translating names, reconciling states, identifying the current record, and carrying exceptions that no schema represents.

The problem is not solved by producing more documentation. Additional records can increase future burden when they are duplicated, unsearchable, unversioned, poorly scoped, or detached from decisions. Conversely, high-quality documentation can reduce burden by preserving provenance, rationale, ownership, and status.

A useful distinction is therefore:

recorded information ≠ available context
available context ≠ understood context
understood context ≠ authorised action

The purpose of institutional memory is not to preserve everything indefinitely. It is to preserve enough structured context that people do not repeatedly reconstruct the same operating reality.

This is where the Garden Institute’s theory of burden meets Nilsoft’s technical programme. The Garden asks what continuity has been lost and what human cost follows. Nilsoft asks how state, evidence, decisions, and ownership can be made durable and inspectable. Doctrine asks which records may legitimately be created, retained, combined, disclosed, or used to act upon a person.

Context preservation is therefore not a purely technical convenience. It is simultaneously a burden problem, a memory problem, and an authority problem.

4. Verification as hidden labour

Automation is often evaluated by the time required to produce an output. That measure can be profoundly misleading.

A generated document may be faster to draft but slower to verify. An automated classification may reduce data entry while creating an exception queue. A recommendation may accelerate a decision while transferring responsibility to a reviewer who lacks access to the evidence or logic behind it. A system may save time for one department by shifting correction work to customers, citizens, contractors, or after-hours staff.

This is verification labour: the work required to establish that an automated or institutional output can be trusted, accepted, corrected, or acted upon.

Verification includes:

• factual checking;
• policy and eligibility checking;
• source validation;
• correction and rework;
• exception review;
• duplicate-execution detection;
• approval;
• incident response;
• audit reconstruction;
• confirmation that a state change actually occurred.

Research on generative AI in knowledge work suggests that AI changes rather than simply removes cognitive work. A 2025 CHI study of 319 knowledge workers found that participants described a shift from information gathering toward verification, from problem solving toward response integration, and from task execution toward task stewardship. Because the study relied on self-report, it should not be read as direct evidence of cognitive decline; it does support the narrower claim that verification and stewardship become central components of AI-assisted work.¹⁹

The EU AI Act explicitly recognises automation bias. For high-risk systems, Article 14 requires human-oversight measures that help reviewers understand limitations, interpret outputs, avoid over-reliance, override or reverse outputs, and stop the system safely.²⁰ This is significant because “human in the loop” is not inherently protective. A reviewer without time, evidence, independence, competence, or actual authority may provide ceremonial oversight while retaining legal or moral responsibility.

The correct unit of evaluation is total burden:

Total Human Burden
=
execution time
+ coordination time
+ documentation time
+ verification time
+ rework time
+ exception-handling time
+ recovery time
+ after-hours spillover

Net Burden Reduction
=
baseline total human burden
− post-intervention total human burden

Cognitive intensity, autonomy, frustration, and perceived meaningfulness should be measured separately rather than converted into fictional equivalent hours. NASA-TLX, for example, provides a multidimensional instrument for subjective workload across mental, physical, temporal, performance, effort, and frustration dimensions.²⁷

The central warning is simple:

Automation that removes execution but expands verification may be pseudo-automation rather than liberation.