A rule that has no exceptions can become unjust. A rule with unlimited exceptions can stop being a rule.
Every civilisation encounters cases that do not fit the normal pathway. A person faces an emergency. A specialised skill cannot be recruited through the ordinary process. A legal rule contains a defined exemption. A public service must act before the standard paperwork is complete. A procurement system confronts a single-source emergency. A school, hospital, regulator or court faces circumstances the standard procedure did not anticipate.
Exception handling is therefore a core governance problem. The civilisation must preserve enough discretion to respond intelligently to reality while protecting predictability, fairness, accountability and the integrity of the ordinary rule.
This article owns that missing Civilisation Atlas layer. It is distinct from How Civilisations Set Rules and Resolve Disputes, which owns the rule system; How Civilisations Organise Power and Limit Authority, which owns discretion and authority; and How Civilisations Coordinate Across System Boundaries, which owns handoffs and boundary cases. Here the reader job is narrower: when should normal rules yield, who may authorise the exception, how is it documented, and how does civilisation stop exception from becoming privilege?
The short answer
Civilisations handle exceptions without breaking the rule when they make exceptions:
defined;
justified;
proportionate;
authorised;
time-bounded where appropriate;
traceable;
reviewable;
and measurable.
They distinguish an exception written into the rule from a discretionary waiver granted by an authorised decision-maker.
They distinguish emergencies from convenience.
They require reasons.
They track repeated use.
And when exceptions become common, they ask whether the rule itself is wrong.
1. Rules exist because consistency has value
Rules reduce uncertainty.
They tell people what is expected.
They constrain arbitrary authority.
They make similar cases more likely to receive similar treatment.
They let institutions scale beyond one person’s judgement.
2. Rules simplify reality
Every rule describes a category.
Eligible or ineligible.
Permitted or prohibited.
On time or late.
Compliant or non-compliant.
Reality is often more continuous than the category.
Exceptions appear at the edge.
3. Edge cases are not defects in reality
A rule can be well designed and still encounter rare conditions its authors did not predict.
Complex civilisation needs a mechanism for those conditions.
4. The first distinction: exception versus error
An exception applies when the unusual treatment is legitimate.
An error occurs when the rule was applied incorrectly.
Do not use “exception” to conceal mistakes.
5. The second distinction: exception versus exemption
An exemption is often defined in advance by law or policy.
If the conditions apply, the person or activity is outside the ordinary requirement.
A discretionary exception may require approval after considering the circumstances.
The governance burden differs.
6. The third distinction: exception versus waiver
A waiver can be permission not to comply with a requirement under defined authority.
The terminology varies across domains.
The analytical point is to know whether relief is automatic, conditional or discretionary.
7. The fourth distinction: exception versus override
An override is an authorised decision to bypass a normal control.
Overrides can be necessary in emergencies.
They deserve stronger logging because they alter the normal control path.
8. The fifth distinction: exception versus workaround
A workaround is often informal.
It may solve a real problem.
It may also bypass governance.
Repeated workarounds should be converted into formal exceptions or process redesign where justified.
9. A good exception preserves the purpose of the rule
The best test is not “does this case feel unusual?”
It is:
what purpose does the rule serve, and does the exception preserve or undermine that purpose?
10. Purpose comes before exception logic
If nobody remembers why the rule exists, nobody can judge whether relaxing it is safe.
Preserve rationale.
11. The exception must be tied to a condition
Urgent need.
Specialist scarcity.
Technical impossibility.
Humanitarian necessity.
Conflict between rules.
Defined statutory circumstance.
Without a condition, discretion becomes preference.
12. Conditions should be observable
“Special circumstances” is weak.
“The ordinary process cannot complete before the safety deadline” is stronger.
Observable criteria reduce arbitrary interpretation.
13. Exceptions should be proportional
Relax only what is necessary.
If one document is unavailable, do not remove every control.
If one deadline must move, do not waive unrelated standards.
14. Minimal deviation protects rule integrity
The smaller the deviation, the easier it is to return to normal operation.
15. Some exceptions should be built into the rule
Predictable recurring edge cases deserve predefined handling.
This improves consistency.
16. Predefined exceptions reduce decision burden
Front-line staff can act without seeking senior approval for known conditions.
This speeds service while preserving control.
17. Too many predefined exceptions make the rule unreadable
A rule with fifty exceptions may need redesign.
Complexity can become its own access barrier.
18. Some exceptions need case-by-case discretion
Rare, high-consequence or context-sensitive cases may resist precise advance definition.
Discretion remains necessary.
19. Discretion should have an authority boundary
Who may approve?
Front line?
Manager?
Director?
Independent regulator?
The level should reflect consequence.
20. Low-consequence exceptions can be delegated
Small deviations with clear criteria should not wait for the top of the institution.
Delegation reduces bottlenecks.
21. High-consequence exceptions need stronger approval
Large financial exposure.
Safety implications.
Rights implications.
Precedent risk.
These justify higher-level review.
22. The approver should not always be the requester
Separation reduces self-interested exceptions.
Four-eyes review can be useful where abuse risk is material.
23. Independent approval is useful when conflicts of interest are high
Some exceptions should be reviewed by another function, regulator or oversight body.
24. Current Civil Service Commission practice illustrates controlled exceptions
The UK Civil Service Commission’s recruitment framework recognises defined circumstances in which fair and open competition may be excepted.
Its current guidance treats exceptions as genuinely exceptional, ties them to categories, requires justification, uses approval thresholds and recommends tracking and multi-stage authorisation.
The institutional details are specific to that system.
The general governance lessons transfer: defined grounds, reasoned approval, proportional authority and audit.
25. “Exceptional” should remain exceptional
If 40% of cases use an exception, the normal rule may no longer fit reality.
Track frequency.
26. Exception rate is a design metric
A rising exception rate can signal:
bad thresholds;
outdated rules;
changing environment;
or widespread gaming.
27. Exception clustering matters
Are exceptions concentrated in one region?
One team?
One demographic group?
One supplier type?
Patterns can reveal structural bias or abuse.
28. Exception reasons should be classified
Emergency.
Specialist need.
Data unavailable.
Conflicting obligation.
System outage.
Humanitarian grounds.
Other.
Taxonomy supports learning.
29. “Other” should not become the largest category
If it does, the taxonomy is weak.
30. Exception records need enough detail to be auditable
What rule?
Which condition?
Who requested?
Who approved?
When?
For how long?
What mitigation?
What outcome?
31. Documentation should be proportionate
A tiny, reversible exception should not require a forty-page submission.
High-consequence exceptions need deeper evidence.
32. Too much paperwork drives exceptions underground
If the formal route is impossible, workers create informal workarounds.
Governance should be usable.
33. Too little documentation destroys accountability
“Manager approved verbally” is fragile.
Important exceptions should leave a record.
34. Reasons constrain discretion
Writing the reason forces the decision-maker to connect facts, rule, purpose and outcome.
It also allows later comparison.
35. Reasons enable consistent future decisions
Case records can reveal how similar exceptions were handled.
This builds institutional memory.
36. Previous exceptions should inform, not automatically bind
Context differs.
Precedent is useful.
Mechanical copying can reproduce old mistakes.
37. Repeated similar exceptions may justify a new rule
Exception data is policy feedback.
If the same edge case returns, update the normal pathway.
38. Temporary exceptions need expiry
Emergency conditions end.
Temporary staffing needs change.
System outages are repaired.
Without expiry, temporary relief can become permanent drift.
39. Sunset dates are useful
Define when the exception ends unless actively renewed.
This reverses the default.
40. Renewal should require fresh evidence
Do not renew automatically because the exception already exists.
Ask whether the original condition remains.
41. Some exceptions are one-time by nature
A specific transaction.
A specific emergency.
A specific person.
Close them explicitly after use.
42. Open exceptions create control debt
Unclosed permissions accumulate.
People forget why they exist.
Review open exceptions periodically.
43. Emergency exceptions need rapid paths
Fire.
Cyber incident.
Medical emergency.
Disaster response.
The ordinary process may be too slow.
Pre-authorised emergency powers can preserve speed.
44. Emergency should have a trigger
Who declares the condition?
What evidence is required?
What authority activates?
What ends it?
45. Emergency exceptions should be narrower than emergency rhetoric
A crisis should not become justification for unrelated waivers.
Scope matters.
46. Emergency records can be completed after action if necessary
Speed can justify delayed paperwork.
It should not justify no paperwork.
47. Post-event review is essential
Was the exception necessary?
Did it work?
Was harm created?
Should the emergency protocol change?
48. Exception authority can become privilege
If powerful actors receive exceptions more easily, rule legitimacy collapses.
Track who benefits.
49. Equal treatment does not require identical treatment
Different circumstances can justify different outcomes.
The difference should be tied to relevant criteria.
50. Relevant difference is the core fairness question
Why does this case deserve different treatment?
If the answer depends on status, influence or personal connection unrelated to the rule’s purpose, the exception is suspect.
51. Exceptions can protect substantive fairness
A rigid rule can produce absurd or harmful outcomes in unusual cases.
Controlled discretion allows civilisation to preserve the purpose rather than worship the wording.
52. Exceptions can also hide discrimination
Subjective discretion can reproduce bias.
Pattern analysis and review are necessary.
53. Comparable cases should be compared
If two similar cases receive different exceptions, record why.
Consistency does not mean automatic sameness.
It means explainable difference.
54. Appeal routes matter
A person denied an exception may need a review path.
This is especially important when the consequence is severe.
55. Appeals should not make every exception endless
Finality matters.
Review depth should be proportionate.
56. The exception approver should know the rule’s purpose
Approval training should include rationale, not only procedure.
Otherwise discretion becomes box-ticking.
57. Training should include examples and counterexamples
What qualifies?
What looks similar but does not?
Boundary examples calibrate judgement.
58. Case libraries can improve consistency
Anonymised examples help staff see how principles apply.
Update them as policy changes.
59. Case libraries can ossify discretion
Examples should guide reasoning, not become hidden new rules.
60. Exception processes need service levels
A legitimate exception route that takes six months may be useless for a time-sensitive case.
Approval speed should match the harm clock.
61. Slow exception processes create informal bypass
If the formal path cannot meet operational reality, people route around it.
Measure turnaround time.
62. Fast exception processes can increase abuse
Speed should not eliminate necessary evidence or separation of duties.
63. Tiered approval can balance speed and control
Low-risk exceptions: local approval.
Medium-risk: specialist review.
High-risk: senior or independent approval.
The categories should be defined.
64. Automation can support exception screening
Systems can check whether predefined conditions apply.
This can increase consistency.
Human review remains important for ambiguous or high-consequence cases.
65. Automated exceptions can hide policy choices
If software determines eligibility, the exception logic should remain inspectable.
Code is part of governance.
66. Algorithmic thresholds create edge cases too
A score of 69.9 and 70.0 can receive different outcomes.
Review mechanisms matter near consequential thresholds.
67. Human override of automated decisions needs governance
Who may override?
For which reasons?
Is override tracked?
Can the model learn from justified overrides?
68. Too many human overrides can reveal model failure
Exception data should feed redesign.
69. Too few overrides can reveal fear
Staff may trust the system too much or fear challenging it.
Override rates require interpretation, not simple targets.
70. Exception design should include data outages
What if required information is unavailable?
Does the service stop?
Can provisional treatment proceed?
What evidence must be reconciled later?
71. Provisional decisions are structured exceptions
Act now under limited conditions.
Verify later.
This can be appropriate when delay creates greater harm than uncertainty.
72. Provisional decisions need reconciliation
Temporary status should not remain unresolved.
Define the follow-up.
73. Exception handling can bridge system boundaries
A case may fit neither of two institutions perfectly.
A joint exception path can prevent abandonment.
The interface owner How Civilisations Coordinate Across System Boundaries explains the boundary architecture.
74. Cross-jurisdiction exceptions need special care
One institution should not casually waive another institution’s requirement.
Authority must be clear.
75. Mutual recognition can reduce the need for exceptions
If systems recognise equivalent qualifications or approvals, ordinary interoperability replaces case-by-case waiver.
Good interface design can shrink the exception queue.
76. Exceptions should not substitute for missing standards
If every organisation needs a bespoke waiver to connect, the system may need a shared standard.
77. Exceptions should not substitute for poor capacity planning
Repeated emergency overtime, procurement waivers or staffing exceptions can indicate chronic undercapacity.
Fix the system.
78. Exceptions should not substitute for leadership indecision
Sometimes managers approve exceptions because they refuse to change an outdated rule.
This pushes policy responsibility downward.
79. Front-line discretion can be humane and dangerous
Front-line workers see context.
They can adapt intelligently.
They can also create inconsistent treatment.
Bounded discretion needs principles and review.
80. Discretion should be wider where context matters greatly
Social work.
Medicine.
Emergency response.
Human circumstances resist full automation.
81. Discretion should be narrower where consistency is essential
Financial controls.
Safety interlocks.
Evidence preservation.
Some rules exist precisely to prevent ad hoc judgement.
82. The more dangerous the override, the stronger the gate
Proportional assurance applies.
83. Exception governance should distinguish permission and obligation
An exception may permit different treatment.
It may not require it.
Decision-makers should know the difference.
84. Exception language should be precise
“May” is not “must.”
“Except where” is not “unless preferred.”
Ambiguous drafting expands discretion unintentionally.
85. Exceptions can create precedent pressure
Once one person receives relief, others ask why they should not.
That is legitimate.
The institution needs a principled answer.
86. The answer should refer to relevant difference
Case A had condition X.
Case B does not.
That is stronger than “management decided.”
87. Secret exceptions undermine trust
Some details need confidentiality.
But the existence and governance of significant exception regimes should usually be visible enough for accountability.
88. Aggregate reporting can protect privacy
Publish counts, categories, approval rates and trends without exposing personal details.
89. Exception dashboards can become targets
Do not reward managers for low exception rates automatically.
They may deny legitimate cases.
Measure quality and pattern, not just volume.
90. Exception volume can be healthy after a new rule
Early use may reveal edge cases.
The key is whether the system learns and stabilises.
91. Exception volume can be unhealthy after years
Persistent high volume suggests the normal rule still mismatches reality.
92. Exception analysis should ask who never applies
A complicated exception route may be used mainly by sophisticated actors.
People with equal need but less knowledge may never request relief.
Access matters.
93. Information about exceptions should be discoverable
If a legitimate route exists, eligible people should be able to find it.
94. Discoverability should not invite gaming
Clear criteria help.
Opacity is not a good anti-abuse strategy.
95. Evidence requirements should match consequence
A minor deadline extension may require little proof.
A major safety waiver requires much more.
96. Evidence should be relevant
Do not collect information merely because the form has always asked for it.
97. Exception files should state uncertainty
Sometimes the evidence is incomplete.
Record that rather than manufacturing certainty.
98. Risk acceptance should be explicit
If the exception increases risk, who accepts it?
This is especially important in safety, finance and security.
99. Risk transfer is not risk elimination
An exception can move burden to another team, supplier or future date.
Track the transfer.
100. Compensating controls can make exceptions safer
If one control is waived, another may reduce risk.
Extra review.
Reduced scope.
Shorter duration.
Additional monitoring.
Compensating controls preserve purpose.
101. Compensating controls should be tested
Do not add ceremonial safeguards.
Ask whether they actually reduce the relevant risk.
102. Exception scope should be the minimum necessary
One transaction, not all transactions.
One month, not indefinite.
One location, not nationwide.
Scope limitation controls blast radius.
103. Exception scope should be visible to systems
Software and operators need to know where the normal rule still applies.
104. Exception inheritance is dangerous
A temporary exception can be copied into later contracts, procedures or software without the original rationale.
Track inheritance.
105. Migrations should revalidate exceptions
When systems change, ask whether old exceptions still apply.
Do not import legacy waivers automatically.
106. Policy revisions should include an exception review
Which old exceptions become unnecessary?
Which need incorporation into the new rule?
Which should end?
107. Rule design and exception design belong together
Write the normal path and the edge path at the same time.
This reduces later improvisation.
108. Good rules include explicit escalation boundaries
Normal decision.
Exception decision.
High-risk exception.
Appeal.
Each route should be visible.
109. Policy authors should observe exception operations
People who write rules should see how edge cases behave in practice.
Otherwise exception systems become permanent shadow policy.
110. Exception handling is a source of policy intelligence
It reveals where reality resists abstraction.
That makes exception data valuable.
111. Exceptions can reveal social change
New family structures.
New technologies.
New work patterns.
New risks.
Rules can age while society changes.
112. Exceptions can reveal technological change
Rules written around physical documents may generate exceptions when digital identity becomes normal.
Repeated exceptions may signal the need for modernisation.
113. Exceptions can reveal institutional mismatch
A case repeatedly falls between two agencies.
The problem may be interface design, not unusual citizens.
114. Exceptions can reveal classification bias
If one group disproportionately needs exceptions, the category may fit them poorly.
Investigate before blaming users.
115. Exceptions can reveal overregulation
If sensible ordinary activity requires repeated waivers, the base rule may be too restrictive.
116. Exceptions can reveal underregulation too
If staff repeatedly add ad hoc safeguards, the base rule may be too weak.
117. The best exception system reduces its own unnecessary workload
Learn from repeated cases.
Improve the rule.
Automate stable cases.
Reserve human judgement for genuine novelty.
118. Exception specialists can become a hidden elite
If only a few insiders understand how to obtain relief, access becomes unequal.
Document the process.
119. Specialist review should support, not monopolise, judgement
Build capability in ordinary teams for low-risk cases.
120. Exception committees can become slow
High assurance can turn into delay.
Use cadence and delegated thresholds.
121. Exception committees can also become permissive
Familiarity can normalise deviation.
Independent audit helps.
122. Audit should sample ordinary approved exceptions too
Do not review only scandals.
Pattern quality matters.
123. Audit findings should change policy
Otherwise audit becomes documentation theatre.
124. Exception governance needs ownership
Who owns the overall exception regime?
Not every case—
the system.
Someone should monitor patterns, policy fit and control health.
125. System ownership prevents fragmented discretion
Without it, every department invents its own approach.
Consistency disappears.
126. Exception regimes need a charter
Purpose.
Authority.
Scope.
Categories.
Approval levels.
Evidence.
Duration.
Appeal.
Reporting.
Review.
127. The charter should name what cannot be excepted
Some requirements are hard floors.
Rights.
Safety barriers.
Legal prohibitions.
Evidence integrity.
Not every rule is waivable.
128. Non-waivable rules should be explicit
Front-line staff should not have to discover them after an attempted exception.
129. Conflict between non-waivable rules requires higher authority
Sometimes obligations collide.
That is no longer an ordinary exception.
It may require legal or policy resolution.
130. Exceptions should not be used to conceal policy conflict
If two rules systematically contradict, fix the rules.
131. Exception handling can preserve innovation
New technologies may not fit old rules.
Controlled pilot exceptions can enable learning while protecting the public.
The innovation owner How Civilisations Innovate Without Breaking What Works explains how bounded experiments should work.
132. Innovation exceptions should be reversible
Limit scope.
Define stop conditions.
Monitor outcomes.
Do not turn experimentation into permanent bypass.
133. Regulatory sandboxes are institutionalised exception spaces
They create limited conditions under which novel models can operate with tailored oversight.
The lesson is controlled deviation, not deregulation without boundary.
134. Sandbox exit matters
What happens after the test?
Return to ordinary rules?
Change the rules?
Stop the activity?
Exit should be designed before entry.
135. Crisis exceptions and innovation exceptions are different
One responds to urgent harm.
The other creates a controlled learning environment.
Do not use the same approval logic blindly.
136. Humanitarian exceptions solve another problem
Some rules create unintended harm in exceptional human circumstances.
Relief may preserve the ethical purpose of the wider system.
137. Humanitarian exceptions still need criteria
Compassion and consistency are not enemies.
Criteria help similar cases receive similar consideration.
138. Exception governance is a trust system
Citizens need to believe:
exceptions exist when genuinely necessary;
they are not reserved for insiders;
and the normal rule still matters.
139. Too few exceptions can make the system cruel
Rigid rules can produce absurd outcomes.
Legitimacy falls when institutions refuse to see obvious context.
140. Too many exceptions can make the system corruptible
If every rule is negotiable, access and influence replace predictability.
141. The healthy middle is governed discretion
Rules provide the floor.
Discretion handles genuine edges.
Records and review keep discretion accountable.
142. The Civilisation Atlas exception test
For any exception, ask:
Which rule is being varied?
What purpose does that rule serve?
What circumstance makes the ordinary path inappropriate?
Is the exception predefined or discretionary?
Who has authority to approve?
Is the approver independent enough?
What evidence is required?
Is the deviation the minimum necessary?
What risk increases?
What compensating controls exist?
What is the scope?
What is the duration?
Does it expire automatically?
Who reviews renewal?
Can the decision be appealed?
How is the exception recorded?
How is privacy protected?
How often is this exception used?
Who benefits from it?
Are patterns uneven?
Does repeated use indicate rule redesign?
143. The exception chain
A useful operating model is:
normal rule → edge condition → eligibility for exception → evidence → authorised judgement → bounded deviation → compensating control → expiry → review → rule learning
144. The exception failure chain
A common failure pattern is:
rigid rule → repeated real-world mismatch → informal workaround → hidden discretion → unequal access → untracked privilege → loss of trust
145. Another exception failure chain
The opposite failure is:
vague rule → easy exception → convenience approvals → exception normalisation → rule erosion → inconsistent treatment → corruption risk
146. The exception repair chain
A stronger sequence is:
define purpose → define grounds → assign authority → require reasons → limit scope → track use → audit patterns → update the base rule
147. Misconception: exceptions weaken rules
Poorly governed exceptions do.
Well-governed exceptions can protect the rule’s purpose by handling legitimate edge cases.
148. Misconception: fairness means no exceptions
Fairness can require treating relevantly different cases differently.
The difference needs reasons.
149. Misconception: discretion is arbitrary
Discretion can be principled, bounded and reviewable.
Arbitrariness is discretion without sufficient governance.
150. Misconception: more documentation always means more control
Excess bureaucracy can push decisions off-record.
Control must be usable.
151. Misconception: low exception rates prove good policy
People may be unable to access the exception path.
Interpret rates with context.
152. Misconception: high exception rates prove flexibility
They may prove the base rule is broken.
153. Model limit: legal exceptions are jurisdiction-specific
This article does not determine when a statutory exemption applies.
Legal advice belongs to the relevant jurisdiction.
154. Model limit: medical exceptions require clinical standards
Clinical judgement is domain-specific.
This framework concerns governance, not treatment decisions.
155. Model limit: security exceptions can contain sensitive information
Audit and transparency should protect legitimate confidentiality.
156. Transfer check: recruitment
A highly specialised expert is needed urgently.
Can ordinary competition produce a candidate in time?
Which exception ground applies?
Who approves?
What duration?
How is future normal recruitment restored?
157. Transfer check: procurement
A critical component has one supplier after a disaster.
Which competition requirements may be varied?
What price safeguards remain?
How long does the emergency route last?
158. Transfer check: school policy
A general attendance rule encounters a student with a documented medical need.
Does the exception preserve the educational purpose while adapting the pathway?
159. Transfer check: digital identity
A person cannot complete the standard verification process.
What alternative proof route exists?
How is fraud risk controlled?
Who owns the edge case?
160. Transfer check: AI decision support
The model recommends denial.
A human reviewer sees unusual evidence outside the model.
Can the decision be overridden?
How is the override recorded?
Does the case feed model improvement?
161. Final answer
How do civilisations handle exceptions without breaking the rule?
They preserve the ordinary rule because consistency, predictability and fairness matter.
They acknowledge that reality still produces legitimate edge cases.
They define exception grounds as clearly as possible.
They scale approval authority with consequence.
They require reasons, evidence, scope and duration.
They use compensating controls when a normal safeguard is relaxed.
They make important exceptions traceable and reviewable.
They protect equal access to the exception path so discretion does not become privilege.
They track patterns rather than treating every case as isolated.
And when exceptions stop being exceptional, they repair the rule itself.
A mature civilisation is neither rigid nor arbitrary. It knows when the rule should hold, when reality justifies a narrow deviation, who may authorise that deviation, and how to return from the edge case without losing the integrity of the system.
Evidence and scope
UK Civil Service Commission, Recruitment Principles and current exceptions guidance. Used as a concrete contemporary example of an exception regime with defined categories, justification, approval thresholds, documentation, tracking and review.
GOV.UK statutory-exception guidance. Used only to illustrate the general distinction between a predefined legal exception and a permission that requires separate authorisation; the legal details are jurisdiction-specific.
The governed-discretion model, exception-health metrics, exception failure chains and Civilisation Atlas exception test are original analytical teaching constructions for this library.
Continue through the Civilisation Atlas
Return to the Civilisation Atlas, ground the rule system in How Civilisations Set Rules and Resolve Disputes, connect discretion to authority through How Civilisations Organise Power and Limit Authority, connect edge cases to system seams through How Civilisations Coordinate Across System Boundaries, or connect controlled experimentation through How Civilisations Innovate Without Breaking What Works.