Appendix B — Failure Modes and Conditional Traps
Not every abandoned approach was a permanent dead end, and not every successful approach was universally good. This catalog asks whether a technology failed because its mechanism was impossible, because its scale or context changed, because institutions did not support it, or because externalities were ignored.
B.1 Scientific and Epistemic Failures
| Failure mode | What it looks like | Better response |
|---|---|---|
| Unfalsifiable patching | Each contradictory result is absorbed by an unmeasured exception | Specify the observation that would force a change; preserve failed data |
| Authority substitution | A famous text or institution substitutes for current evidence | Treat inherited knowledge as a hypothesis with a known error history |
| Benchmark capture | Optimisation targets the public score rather than the real objective | Keep hidden, rotated, real-world evaluation and outcome audits |
| Scale extrapolation | A successful bench result is treated as an industrial rate | Measure the full process, distribution, maintenance, and failure recovery at target scale |
| Single-study synchrony | Several narratives repeat one disputed observation | Preserve provenance and seek independent replication or direct data |
| Category error | A wrong explanation is discarded together with a useful practice | Separate mechanism, observation, and engineering response |
Historical examples include phlogiston, spontaneous generation, geocentric astronomy, and humoral medicine. Their failures did not make every associated observation, instrument, or sanitary practice worthless. The useful lesson lies in separating testable mechanism from the practice that happened to work before the mechanism was understood.
B.2 Technology and Scale Failures
| Failure mode | Mechanism | Design response |
|---|---|---|
| False minimalism | A locally simplest process is mistaken for the simplest total system | Compare boundaries, labour, waste, failure recovery, and maintenance—not one component |
| Unfalsifiable precision | A polished number is repeated without source, range, or conditions | Attach conditions, uncertainty, measurement method, and verification date |
| Pilot-to-production leap | A small batch succeeds while controls, materials, and operators do not scale | Define control points, process capability, maintenance, and stop criteria before scale-up |
| Material monoculture | One input failure disables the whole system | Use qualified alternatives, strategic stocks, and tested substitution |
| Central-point fragility | Standardisation removes duplication but creates a single failure | Standardise interfaces while preserving redundant routes and recovery capacity |
| Technology lock-in | Incumbent infrastructure captures a process known to be inferior | Use staged replacement, interoperability, and explicit migration cost |
| Maintenance blindness | Acquisition and demonstration are funded; inspection and spares are not | Fund the full lifecycle before declaring success |
| Safety debt | Hazards are accepted because failure appears unlikely | Quantify consequence, exposure, controls, residual risk, and stop-work criteria |
| Unmaintainable optimisation | Peak performance depends on exceptional inputs or tuning | Design for variation, operator error, worn parts, and ordinary supply |
Roman numerals, logographic writing, classical alchemy, canal transport, and mechanical calculators all worked within particular constraints. None should be described categorically as “waste.” The question is whether a recovery programme retains the capability and abandons the unnecessary complexity.
B.3 Institutional Failure Modes
| Failure mode | Mechanism | Countermeasure |
|---|---|---|
| Monopoly on literacy | Gatekeeping turns knowledge into rent and blocks error correction | Broad education, redundant archives, open standards, succession plans |
| Metric gaming | Targets replace outcomes | Multiple measures, independent audit, public uncertainty, rotating evaluations |
| Debt without productive use | Obligations compound while assets and cash flow do not | Transparent credit, stress tests, recovery rules, productive-use review |
| Centralised calculation failure | Decision-makers lack dispersed local information | Feedback, local capability, experiment, and accountable aggregation |
| Licence and gatekeeping capture | Entry barriers protect incumbents rather than public safety | Transparent criteria, independent review, time limits, appeals |
| Fiscal illusion | The sponsor pays while environmental or health costs are socialised | Full-cost accounting, pollution control, maintenance reserves, lifecycle procurement |
| Successorship vacuum | Critical knowledge dies with one person or workshop | Apprenticeship, written procedures, duplicate equipment, supervised rotation |
| Redundant failure | Backups share the same supplier, protocol, or belief | Physical and logical diversity; independent recovery tests |
B.4 Use of the Catalog
Before committing to a large programme, write four lines:
- Falsifier: what result would show this path is not working?
- Evidence boundary: what has been demonstrated, at what scale, and where?
- Full-cost boundary: which labour, material, waste, health, environmental, and failure costs are included?
- Recovery route: what still works if this programme is abandoned?
If any line cannot be answered, the next task is measurement and institutional design—not construction at larger scale.