Interaction Effect Mapping

Map how factors change one another's effects when combined so interventions are not evaluated only in isolation.

Version

v1 · 2026-08-24 · History

Solution archetype #

554

Problem family

Representation, Classification & Model Misfit

Problem subfamily

Relation, Interaction & Multicausal Structure

Essence

Interaction Effect Mapping is the practice of making combinations visible. It applies when factors that look understandable in isolation may behave differently when they coexist. The archetype asks a simple structural question: what changes when A and B happen together?

The answer may be synergy, antagonism, interference, independence, threshold dependence, saturation, or context-specific reversal. The point is not merely to produce an analysis table. The map should change design decisions: what to combine, what to separate, what to sequence, what to monitor, and what to stop assuming is additive.

Compression statement

When multiple factors are present or planned together and their combined effect may be non-additive, identify the factors, test or infer their combinations, classify the interaction effects, and redesign the intervention around the mapped pattern of synergy, antagonism, interference, independence, or conditional dependence.

Canonical formula: effect(A + B) is not assumed to equal effect(A) + effect(B); map delta_interaction = observed_combined_effect - expected_additive_effect.

When This Archetype Applies

Complete catalog groundingAt least one sufficient condition set is fully represented by existing primes or domain-specific abstractions.

Diagnostic problem

A system contains multiple potentially interacting factors, but decisions are being made as if each factor's effect can be understood independently. The hidden risk is that combinations may produce amplified benefit, destructive interference, masked harm, or context-dependent reversals.

What this problem means

The structural problem is hidden non-additivity. The system is being reasoned about as if each factor contributes its own separable effect, but the real outcome depends on the relation among factors.

This creates several recurring errors. Teams overestimate bundles because each part worked alone. They miss synergies because the right pair was never tested. They blame one factor for failure when the problem was the pairing. They generalize evidence from one context to another without noticing that a co-factor has changed.

Interaction Effect Mapping addresses that problem by shifting the unit of evaluation from isolated factors to factor combinations.

Applicability expression3 distinct conditions

any oneContext-varying effectsorNegative combined interactionorUnmeasured plausible interactions

Algebraic(A∨B∨C)

A=(a∧a′)∨(b∧b′)∨c∨(d∧d′)∨(e∧e′)∨(f∧f′)∨(g∧g′)∨(h∧h′)∨i∨j

′ context guard? connective not recorded∅ no catalog witness yet

groundedpartly groundedopen

3 conditions, all required.

3At least one of theselettered A–C

Any single one of these completes the pattern.

A

Context-varying effects · grounded · any one of 10

Observed effects vary sharply across contexts or populations. more

a

domainCarryover Effect— The validity threat in crossover and within-subject designs where residual influence from an earlier treatment persists into a later measurement window, biasing the contrast — its magnitude set by the unit's relaxation time against the inter-treatment gap.

AND both parts required

context guardThe carryover-induced contrast is sharp across treatment-sequence contexts.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

b

domainNovelty Effect— Recognise that a newly introduced stimulus draws an elevated response tied to its recency, not its intrinsic worth, which decays near-exponentially toward a steady-state baseline — so short-window evaluation systematically over-estimates the intervention's lasting effect.

AND both parts required

context guardThe observed novelty-effect contrast is sharp across early and steady-state contexts.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

c

domainIdiosyncratic Reaction— Classify a rare drug harm as a distinct causal category — dose-independent, qualitatively different in kind, and confined to a small biologically-defined susceptible tail — so it is understood not as too-much-drug but as uniform exposure meeting a heterogeneous responder population.

or — any one of these establishes this case

d

domainContraindication— Flag the sparse patient conditions under which a normally-indicated treatment becomes inadvisable, by naming the specific context in which its risk-benefit balance flips sign.

AND both parts required

context guardThe contraindication is established by observed past treatment results.

suppliesAn effect has been observed in past results.

or — any one of these establishes this case

e

domainSmall-Study Effects— The meta-analytic pattern in which smaller studies report systematically larger effects than larger ones, producing funnel-plot asymmetry that inflates the pooled estimate — a shared symptom of several biases, not a diagnosis of any one cause.

AND both parts required

context guardThe small-versus-large-study effect difference is sharp across study-size contexts.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

f

domainCohort Effect— An observed outcome difference associated with membership in groups defined by a shared birth or entry interval, kept distinct from aging, period shocks, and any unproven causal account of the cohort contrast.

AND both parts required

context guardThe observed between-cohort contrast is sharp across the compared populations.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

g

domainPerceptual Set— Explain how prior expectation, context, motivation, or expertise biases what is perceived from ambiguous or degraded input — the percept is the prior weighted by how decisive the sensory evidence is, so the expectation dominates precisely when the stimulus underdetermines it.

AND both parts required

context guardThe perceptual-set bias contrast is sharp across ambiguous and decisive sensory contexts.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

h

domainTask-Switching Cost— Isolate the performance penalty paid at the moment of changing between rule-sets — separate from either task's steady-state difficulty — by contrasting switch trials against repeat trials in the same mixed block, and split it into a preparation-reducible part and an irreducible residual.

AND both parts required

context guardThe switch-versus-repeat performance contrast is sharp across trial contexts.

suppliesThe observed effect varies sharply rather than remaining stable.

or — any one of these establishes this case

i

domainVentriloquism Effect— A sound is mislocalised toward a synchronised visual event when the brain infers a common cause, because vision's lower spatial uncertainty pulls the merged location estimate toward it by inverse-variance weighting — gated by temporal and spatial binding windows beyond which the percept splits.

or — any one of these establishes this case

j

domainWell-Travelled Road Effect— People underestimate the duration of familiar routes and overestimate equally long unfamiliar ones, because retrospective duration is reconstructed from the count of distinct, attention-demanding event traces memory retained — few for automatic familiar travel, many for novel decision-dense travel.

How this was matched — 2 shared + 2 branches

Observed effect magnitude or sign is sharply setting-dependent.

All of

• modalityAn effect has been observed in past results.
• comparisonThe observed effect varies sharply rather than remaining stable.

…and any one of

• domainThe sharp variation can occur across contexts.
• domainThe sharp variation can occur across populations.

or — any one satisfies this slot

B

Negative combined interaction · grounded

Two individually useful elements produce a combined effect below the specified no-interaction baseline. more

primeSynergy and Antagonism— Amplified or diminished effects.

or — any one satisfies this slot

C

Unmeasured plausible interactions · open

Dependency, compatibility, or dose interactions are plausible but have not been measured against a baseline. more

Other requirements and context (2)

Why these sit outside the expression

Supporting context — it may accompany or help interpret the situation, but it is not a load-bearing condition in a sufficient diagnostic set.

Application gate — it governs whether applying the archetype is appropriate or material, rather than defining the structural problem itself.

• Supporting contextMultiple interventions, components, or variables will be used together. more

• Application gateA team is choosing among bundles, sequences, or policy packages.

Coverage

2 of 3 conditions grounded · 1 open.

None of the 1 open conditions sit in the shared core — each falls inside one alternative branch, so grounding any one of them closes only that branch.

Read the methodology · Download the trigger-logic data

When to Use This Archetype

Use this archetype when multiple interventions, components, policies, actors, incentives, or conditions will operate together and isolated evidence is not enough. It is especially useful before scaling a bundle, deploying a multi-part program, combining system features, or interpreting inconsistent results across contexts.

It also fits situations where a component is praised or blamed even though its effect may depend on what it was paired with. For example, a coaching program may look weak unless paired with timely feedback, while a retry mechanism in infrastructure may become harmful when paired with aggressive autoscaling under load.

Do not use this archetype merely because there are many parts. The parts must plausibly modify one another's effects. A simple inventory, dependency list, compatibility checklist, or one-factor test is not enough.

Structural Problem

The structural problem is hidden non-additivity. The system is being reasoned about as if each factor contributes its own separable effect, but the real outcome depends on the relation among factors.

This creates several recurring errors. Teams overestimate bundles because each part worked alone. They miss synergies because the right pair was never tested. They blame one factor for failure when the problem was the pairing. They generalize evidence from one context to another without noticing that a co-factor has changed.

Interaction Effect Mapping addresses that problem by shifting the unit of evaluation from isolated factors to factor combinations.

Intervention Logic

The intervention begins by naming the factors clearly. Those factors may be treatments, design features, resources, incentives, environmental conditions, organizational roles, or technical components. The map then establishes a baseline expectation: what would we expect if effects were merely additive or independent?

Next, the relevant combinations are selected. In small systems, full combination testing may be possible. In larger systems, the map may focus on pairwise combinations, high-risk combinations, high-frequency combinations, or combinations suggested by theory and prior failures.

The combined effects are then measured, simulated, inferred, or reviewed. The result is classified into actionable categories: reinforcing, antagonistic, interfering, independent, conditional, saturating, or uncertain. Finally, each category is connected to a design update: combine, separate, sequence, buffer, monitor, reduce intensity, or abandon.

Key Components

Interaction Effect Mapping shifts the unit of evaluation from isolated factors to factor combinations, exposing the non-additivity that single-variable thinking hides. The Factor List names the specific candidates for interaction — treatments, design features, incentives, or technical components precise enough to test or reason about — so the analysis does not dissolve into impressionistic talk about everything in the system. The Interaction Matrix is the structured representation of relations among those factors, recording which combinations have been considered and what is known about each. Effect Measurement defines the outcome being compared and grounds the map in evidence rather than impression, ideally tagging confidence, source, and time horizon since the same combination may improve one measure while worsening another.

Three further components convert observation into design. Interaction Type Classification sorts each pairing into actionable categories — reinforcing, antagonistic, interfering, independent, conditional, saturating, or uncertain — because raw observations without classification remain a collection rather than a guide. A Compatibility Check screens whether a given combination can be safely and meaningfully tested or deployed, acting as a guardrail that runs before interaction mapping rather than substituting for it. The Design Update Rule connects the map to action by specifying what to do when synergy, antagonism, interference, neutrality, or uncertainty is found; without this linkage the archetype collapses into analysis without intervention. Finally, the Review or Iteration Rule keeps the map honest over time, since scale, population, background conditions, and incentives can all change the interaction pattern after the initial study.

Component	Description
Factor List ↗	The factor list defines what may interact. A good factor list is specific enough to test or reason about. “Support” is usually too vague; “weekly coaching call,” “automated reminder,” or “peer accountability group” is more useful. The factor list prevents the map from becoming an impressionistic discussion about everything in the system.
Interaction Matrix ↗	The interaction matrix is the structured relation map. It records which combinations have been considered and what is known about their combined effect. It may be visualized as a table, graph, dependency map, or experiment design, but the component is the structured representation of relations, not the display artifact alone.
Effect Measurement ↗	Effect measurement defines what outcome is being compared. The same combination may improve one measure while worsening another, so measurement must match the purpose of the intervention. In uncertain settings, effect measurement should also show confidence, evidence source, and time horizon.
Interaction Type Classification ↗	Classification turns observations into design meaning. A combination that reinforces value invites bundling. A combination that cancels value invites separation or redesign. A combination that works only under a threshold invites tuning. Without classification, the map is just a collection of observations.
Compatibility Check ↗	A compatibility check screens whether a combination can safely and meaningfully be tested or deployed. It is a guardrail, not the archetype itself. Compatibility asks whether elements can coexist; interaction mapping asks how coexistence changes their effects.
Design Update Rule ↗	The design update rule connects the map to action. It defines what to do when the map shows synergy, antagonism, interference, neutrality, or uncertainty. This component is what keeps the archetype from becoming analysis without intervention.
Review or Iteration Rule ↗	Interaction maps can expire. Scale, population, background conditions, technology, incentives, and load can all change the interaction pattern. A review rule defines when to revisit the map and what signals should trigger revision.

Common Mechanisms

7 documented mechanisms across 5 implementation forms.

The grouping reflects forms represented among the mechanisms currently documented for this archetype; an absent form is not necessarily an impossible implementation.

Analysis, Modeling & Optimization · 1 mechanism

• Treatment Interaction Analysis — A method for evaluating whether an intervention's effect changes under different co-treatments, conditions, populations, or moderators.

Decision, Gate & Allocation · 1 mechanism

• Compatibility Screening — A checklist, rule set, or review procedure that filters unsafe, invalid, infeasible, or uninterpretable combinations.

Experiment, Test & Rehearsal · 2 mechanisms

• Factorial Experiment — Tests the focal factor, potentiating factor, and paired condition so interaction effects can be separated from isolated effects.
• Pairwise Combination Testing — A reduced testing method that checks two-factor combinations to detect likely interaction effects.

Record, Log & Register · 1 mechanism

• Interaction Matrix Table — A table or grid used to display factor combinations, combined effects, evidence confidence, and recommended actions.

Representation, Specification & Plan · 2 mechanisms

• Dependency Interaction Map — A map of how dependencies among components, services, processes, resources, or teams modify one another's effects.
• Design of Experiments Protocol — A planning protocol that determines which factors, levels, combinations, assignment rules, and measurement windows will be used to detect interaction effects efficiently.

Parameter / Tuning Dimensions

The main tuning dimension is scope: pairwise, selected bundle, full factorial, or higher-order mapping. A narrow scope is faster and cheaper; a broad scope is more complete but can become infeasible.

A second dimension is evidence strength. Some maps are experimental, some are observational, some are simulation-based, and some are expert-elicited. The draft should not pretend those have equal confidence.

A third dimension is factor granularity. Coarse factors make mapping manageable but vague. Fine-grained factors make mapping precise but may create combinatorial explosion.

Other important tuning dimensions include outcome metric, time horizon, context stratification, risk threshold, review cadence, and the action threshold for redesign.

Invariants to Preserve

Preserve clear factor identities. If the factors are vague, the map cannot explain what is interacting.

Preserve an explicit baseline. Interaction means deviation from an expected isolated or additive effect; without a baseline, every change can look like an interaction.

Preserve evidence quality. A suspected interaction is not the same as a tested interaction, and a local interaction is not necessarily universal.

Preserve action linkage. The map should inform combination, separation, sequencing, buffering, monitoring, or retirement. Otherwise it is analysis without intervention.

Target Outcomes

A successful interaction map reveals beneficial combinations that should be reinforced, harmful combinations that should be avoided or controlled, and neutral combinations that need not receive special attention.

It also improves attribution. Teams stop saying “this intervention worked” or “this feature failed” when the real explanation is that it worked with one co-factor and failed with another.

The strongest outcome is combination-aware design: the system is deliberately shaped around known or suspected interaction patterns rather than assembled from individually attractive parts.

Tradeoffs

Interaction mapping trades simplicity for realism. It gives a more truthful view of combined behavior, but it adds complexity, evidence burden, and decision overhead.

It also trades completeness against feasibility. Full combination testing may be impossible. Reduced mapping may be necessary, but it can miss higher-order interactions.

Finally, it can trade optimization against robustness. Designing around a synergy can improve performance but also create dependency on a fragile combination.

Failure Modes

The most common failure mode is spurious interaction mapping: noise or confounding is mistaken for a real relation. This can be mitigated through replication, confidence ratings, causal guardrails, and clear evidence labels.

Another failure mode is combinatorial explosion. Too many factors can make the map unusable. Prioritize high-impact, high-risk, high-frequency, theory-relevant, or irreversible combinations.

A third failure mode is pairwise blindness. Pairwise tests are useful, but they can miss three-way or higher-order interactions. Use targeted bundle testing where higher-order effects are plausible.

A fourth failure mode is static-map drift. A map made at one scale or in one context may fail later. Build in review triggers.

The final failure mode is analysis without action. Every interaction category should point to a design response.

Neighbor Distinctions

Interaction Effect Mapping is distinct from Catalytic Pairing. Catalytic Pairing is a downstream design move that deliberately uses amplification; Interaction Effect Mapping first discovers whether amplification exists.

It is distinct from Synergistic Bundle Design. Bundle design constructs a package around reinforcing effects; interaction mapping diagnoses whether proposed combinations reinforce, cancel, interfere, or merely coexist.

It is distinct from Antagonism Screening and Separation. Antagonism screening specializes in harmful negative interactions; interaction mapping covers the full interaction space, including positive, neutral, conditional, and uncertain relations.

It is distinct from Compositional Assembly. Assembly asks how to build a whole from parts; interaction mapping asks how the parts change one another's effects when combined.

It is distinct from a compatibility matrix. Compatibility tells whether coexistence is allowed or feasible. Interaction mapping tells what coexistence does.

Cross-Domain Examples

In public policy, a housing package may include subsidies, zoning reform, construction incentives, and tenant protections. The value of any one policy depends on what else is present, so the package should be mapped for reinforcement and interference before scale.

In infrastructure, caching, retries, rate limiting, and autoscaling may each improve reliability alone. Under load, however, their combined behavior can either stabilize the system or amplify failure. Interaction mapping identifies which combinations require tuning or separation.

In education, tutoring, adaptive software, parent communication, and assessment cadence may reinforce one another for some students while creating overload for others. Mapping interactions by context helps avoid one-size-fits-all conclusions.

In organizational design, incentives, autonomy, reporting lines, and review cadence often modify one another's effects. A performance system that works under one governance arrangement may backfire under another.

Non-Examples

A single A/B test of one feature against a control is not Interaction Effect Mapping unless it examines how the feature's effect changes with another factor.

A dependency inventory is not Interaction Effect Mapping unless it evaluates how dependencies modify outcomes when active together.

A compatibility checklist is not Interaction Effect Mapping by itself. It may prevent invalid combinations, but it does not explain synergy, antagonism, interference, or conditional effects.

A copied best-practice bundle is not Interaction Effect Mapping if the local combination effects are not examined.

Related Abstractions

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (3)

• Factorial Design: Multiple variables tested together.
• Relation: Describes associations or dependencies.
• Synergy and Antagonism: Amplified or diminished effects.

Also references 4 related abstractions

• Causality: Cause-effect relationships.
• Composition: Arranges components into a cohesive whole.
• Cross-Impact Analysis: Interacting trends.
• Effect Size: Magnitude of effect.

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

Factorial Interaction Mapping · mechanism family variant · recognized

Maps interaction effects through a planned factorial or fractional-factorial design that estimates main effects and interaction terms.

• Distinct from parent: The parent archetype can use experiments, observation, simulation, expert review, or dependency analysis; this variant specifically uses factorial experimental logic.
• Use when: {'condition': 'Factors can be experimentally varied or systematically simulated.', 'reason': 'The design can directly estimate interaction terms rather than relying only on judgment.'}; {'condition': 'The cost and risk of testing combinations are acceptable.', 'reason': 'Factorial designs can become expensive as factor count grows.'}.
• Typical domains: agronomy and product optimization, program evaluation, engineering and operations
• Common mechanisms: Factorial Experiment, Design of Experiments Protocol

Pairwise Interaction Mapping · scale variant · recognized

Maps two-factor interactions as a tractable approximation when full combination mapping is infeasible.

• Distinct from parent: The parent archetype can include bundles and higher-order interactions; this variant intentionally restricts attention to pairwise combinations.
• Use when: {'condition': 'The number of factors makes exhaustive testing impractical.', 'reason': 'Pairwise coverage provides a manageable first pass.'}; {'condition': 'Most plausible risks are expected to arise from direct two-factor interactions.', 'reason': 'The approximation is defensible when higher-order effects are unlikely or low consequence.'}.
• Typical domains: software compatibility, policy package review, operations planning
• Common mechanisms: Pairwise Combination Testing, Interaction Matrix Table

Dependency Interaction Mapping · domain variant · recognized

Maps how dependencies among processes, services, resources, or organizational units modify one another's behavior when active together.

• Distinct from parent: The parent maps any non-additive factor relation; this variant focuses on dependency-mediated interactions.
• Use when: {'condition': 'System components are individually acceptable but interact through shared dependencies.', 'reason': 'Failures and amplifications often occur through dependency pathways rather than direct pairing alone.'}; {'condition': 'A change in one component alters the effect, risk, or load of another component.', 'reason': 'The dependency relation is functioning as an effect modifier.'}.
• Typical domains: software infrastructure, supply chains, organizational workflow
• Common mechanisms: Dependency Interaction Map, Compatibility Screening

Near names: Interaction Effect Analysis, Combined Effect Mapping, Treatment Interaction Analysis, Dependency Interaction Maps.

Editorial Notes

Problem Classification

Classification: Representation, Classification & Model Misfit → Relation, Interaction & Multicausal Structure

Problem kernel: factor interactions are omitted from additive decisions

Rationale: Components can amplify, suppress, or reverse one another, so isolated effect estimates do not represent combined causal structure.

Independent corroboration: The earliest necessary condition in the frozen evidence is: A system contains multiple potentially interacting factors, but decisions are being made as if each factor's effect can be understood independently. That is a relation interaction and multicausal structure problem because Object-centered or additive descriptions hide direction, composition, interaction, nonlocal influence, and multiple causal pathways among entities.

Review outcome: Independent reviewer agreement; high confidence.