Intervention Stack Accretion

Version

v1 · 2026-08-24 · History

Prime #

935

Origin domain

Systems Thinking & Cybernetics

Subdomain

intervention dynamics → Systems Thinking & Cybernetics

Core Idea

A system accumulates a growing set of concurrently active interventions, each justified at add-time, whose joint cost grows super-linearly and whose removal is structurally harder than their addition. Three commitments drive it: asymmetric add/remove costs, combinatorial interaction (the \(2^N\) joint-effect space), and constituency formation around each intervention.

How would you explain it like I'm…

The Pile Only Grows

Imagine every time something goes a little wrong, you add a new rule to fix it, and you never take old rules away. Soon you have a giant pile of rules nobody can keep track of, and taking one away feels scary because what if you needed it? Adding is easy, but cleaning up almost never happens on its own. So the pile of rules just keeps growing.

Easy to Add, Hard to Remove

Intervention Stack Accretion is what happens when a system keeps piling on fixes — rules, patches, safety steps — that all stay switched on, and the pile only ever grows. Adding a new fix is easy: you just have to show it might help with a problem you feel right now. Removing one is hard: you'd have to prove it's no longer needed, and you can't see what would happen without it while it's still in place. On top of that, every fix grows fans, paperwork, and other things depending on it, so people resist removing it even if it stopped helping. Because adding is easy and removing is hard, you get a one-way ratchet. The only way to shrink the pile is a special clean-up job — a sunset review, a refactor, a debt sprint — that someone has to deliberately start.

The One-Way Fix Ratchet

Intervention Stack Accretion is the pattern where a system accumulates a growing set of concurrently active interventions — each individually justified when added — whose joint behavior and management cost grow disproportionately to the simple sum of parts, and whose removal is structurally harder than their addition. Three commitments define it. First, asymmetric add/remove costs: adding needs only a showing that it might help a felt local need, while removing needs a showing that it is no longer needed — harder, because the counterfactual of life without it is unobservable while it's in place. Second, combinatorial interaction: with N active interventions the joint-effect space isn't N items but the 2^N subsets and their interactions, so complexity scales super-linearly. Third, constituency formation: each intervention acquires stakeholders, dependencies, and audit trails that resist removal regardless of whether it still does net good. Together these form a one-way ratchet — the stack only grows under ordinary operation — and reversing it takes a distinct named operation (deprescribing, sunset review, refactor, technical-debt sprint) that must be deliberately instituted.

 

Intervention Stack Accretion is the structural pattern in which a system accumulates a growing set of concurrently active interventions, each individually justified at the time of addition, whose joint behavior and management cost grow disproportionately to the simple sum of parts, and whose removal is structurally harder than their addition. Three commitments define it. First, asymmetric add/remove costs: adding an intervention requires showing only that it might help against a felt local need, while removing one requires showing it is no longer needed — structurally harder, because the counterfactual of what happens without it is unobservable while the intervention is in place. Second, combinatorial interaction: with N concurrently active interventions the joint-effect space is not N items but the 2^N subsets and their pairwise and higher-order interactions, so management complexity scales super-linearly. Third, constituency formation: each intervention, once present, acquires stakeholders, dependencies, audit trails, and institutional memory that resist its removal independently of whether it still does net good. Together these produce a one-way ratchet: the stack only grows under ordinary operation. Reversing it requires a distinct named operation — deprescribing, sunset review, refactor, regulatory simplification, technical-debt sprint, change moratorium — which must itself be deliberately instituted and is resisted by the same forces that resist any individual removal. The pattern is sharper than a generic directional-asymmetry ratchet because it specifies what is accreting: discrete interventions on a shared system, each with its own justification, audit trail, and constituency; and sharper than a generic rising-load pattern because it specifies the add-remove asymmetry and combinatorial interaction rather than mere monotonic accumulation.

Broad Use

• Clinical medicine: polypharmacy — medications accumulate until interactions and prescribing cascades exceed any single drug's benefit.
• Regulation and law: rules accrete across administrations; sunset clauses exist because the default is monotonic accretion.
• Software: feature flags, dependency stacks, and observability layers accrete into technical debt.
• Security: defence-in-depth controls accrete into alert fatigue and conflicting requirements.
• Organisations: change programmes, OKRs, and training mandates accrete into "initiative fatigue."
• Curricula and tax codes: required courses or exemptions accrete until the combined whole is unworkable.

Clarity

It separates intervention quality from intervention quantity: each item can be individually positive while the joint stack is net-negative.

Manages Complexity

It names three designable quantities — the add/remove asymmetry, the interaction matrix, and the constituency — and identifies a distinct named inverse-operation as the only force countering the ratchet.

Abstract Reasoning

It trains the reasoner to ask, at add-time, what evidence will be required to remove this later, how cost scales with the interaction matrix, which constituency will form, and whether a named inverse-operation exists.

Knowledge Transfer

• Pharmacy → software: deprescribing protocols are the same move as refactor sprints and deprecation policies.
• Software → law/curriculum: the "debt" metaphor and its instruments (budget, scheduled paydown) port to regulatory and curricular debt.
• Medicine → security: Beers-style "controls of concern" lists are a direct port of evidence-based deprescribing criteria.

Example

A patient starting at 55 on one antihypertensive accumulates a statin, an SSRI, a PPI for drug-triggered reflux, a benzodiazepine for SSRI-induced insomnia, and more — until at 75 they are on eleven drugs, several treating other drugs' side effects, removed only by a deliberate deprescribing review.

Relationships to Other Abstractions

Current abstraction Intervention Stack Accretion Prime

Parents (1) — more general patterns this builds on

• Intervention Stack Accretion is a kind of Ratchet Effect Prime

  Intervention Stack Accretion is a specialization of Ratchet Effect, retaining the parent's defining structure while adding the child's specific commitments.

Children (1) — more specific cases that build on this

• Technical Debt Prime is a kind of, typical Intervention Stack Accretion

  Technical Debt is typically a specialization of Intervention Stack Accretion, retaining the parent's defining structure while adding the child's specific commitments.

Hierarchy paths (3) — routes to 3 parentless roots

• Intervention Stack Accretion → Ratchet Effect → Path Dependence → Dependency

Show alternative paths (2)

Not to Be Confused With

• Intervention Stack Accretion is not Ratchet Effect because a bare ratchet names only directional asymmetry, whereas this prime adds discrete interventions, combinatorial interaction, and constituency formation.
• Intervention Stack Accretion is not Lock-in because lock-in is per-item reversal cost, whereas accretion's distinctive content is the additive accumulation across many items and their super-linear joint cost.
• Intervention Stack Accretion is not Cascade because a cascade is a trigger pattern (one event causing the next), whereas accretion is the accumulated state and many additions are independent.