Skip to content

Assumption

Prime #
637
Origin domain
Philosophy
Subdomain
logic and reasoning → Philosophy

Core Idea

An assumption is a proposition treated as true, for the purposes of some reasoning or activity, without being currently demonstrated within that reasoning. The structural commitment is that inference, modeling, design, and communication all rest on a layer of unproven inputs whose truth is loaded onto the conclusions or outputs that depend on them. An assumption is not a hypothesis (offered for test), not a premise in the narrow logical sense (which may be discharged within a proof), and not a fact (demonstrated independently); it is a load-bearing belief held without current justification because the activity must proceed. The structural signature is the transparent layer between what is given and what is concluded: every nontrivial inference, calculation, model, plan, or message has assumptions stacked beneath it, and they are often invisible until they fail — when a violated assumption makes a downstream conclusion go wrong in a way that looks inexplicable from any single layer, because the arithmetic was right, the data were clean, the procedure was correct, and yet the result is wrong.

Three features distinguish assumption as a structural pattern rather than just "a thing we believe." Assumptions have load: they bear the weight of some downstream commitment, and the load can be located by asking which conclusions would fall if the assumption were false. Assumptions exist whether or not the assumer is aware of them, so that surfacing implicit assumptions is itself a recognized competence across domains. And assumptions interact: an assumption can be defended by another assumption in a regress, protected by a procedure, or replaced by an evidenced claim, which is what empirical inquiry incrementally does. The pattern is tied to reasoning, which gives it a faint practice-bound flavor, but its vocabulary is bare and it travels through any inferential substrate.

How would you explain it like I'm…

The Hidden Bottom Block

When you build a tower of blocks, the bottom block is holding up all the others, even though you stopped looking at it. An assumption is like that bottom block in your thinking: something you treat as true so you can keep going, without checking it again right now. If that hidden block was wrong, the whole tower can fall over and you won't know why.

The Invisible Foundation

An assumption is something you take to be true so you can get on with figuring out an answer, even though you haven't proven it right now. It quietly holds up whatever you decide next, the way a foundation holds up a house you can't see. The tricky part is that assumptions can be invisible: you might not even notice you made one until it turns out to be wrong. When that happens, your answer breaks for no obvious reason, because every step you took was fine except the hidden one underneath.

Load-Bearing Belief

An assumption is a claim you treat as true for the sake of some reasoning, plan, or model, without justifying it inside that reasoning. It is not a hypothesis, which you set up specifically to test, and not a fact, which you have actually demonstrated; it's a belief you lean on because the work has to move forward. The key idea is that assumptions are load-bearing: real conclusions sit on top of them, so if one is false, the things depending on it collapse. They also exist whether or not you're aware of them, which is why spotting your hidden assumptions is a genuine skill. When a violated assumption breaks a result, the failure looks baffling, because the math was right, the data were clean, and yet the answer is wrong.

 

An assumption is a proposition treated as true for the purposes of some reasoning or activity without being demonstrated within that reasoning. Structurally it forms a transparent layer between what is given and what is concluded: every nontrivial inference, calculation, model, plan, or message stacks assumptions beneath it. It is distinct from a hypothesis (offered explicitly for test), from a premise in the narrow logical sense (which can be discharged inside a proof), and from a fact (demonstrated independently) — it is a belief held without current justification precisely because the activity must proceed. Three features make it a structural pattern rather than just a belief. First, assumptions have load: they bear the weight of downstream commitments, and you can locate that load by asking which conclusions would fail if the assumption were false. Second, they exist whether or not the assumer is aware of them, which is why surfacing implicit assumptions is a recognized competence across fields. Third, they interact — an assumption can be propped up by another assumption in a regress, shielded by a procedure, or replaced by an evidenced claim, which is exactly what empirical inquiry does incrementally. Their characteristic failure signature is the inexplicable downstream error: clean inputs and correct procedure, yet a wrong result, because a layer below silently gave way.

Structural Signature

the unproven proposition held as true — the activity that rests on it — the downstream conclusion loaded onto it — the load it bears — the transparency that keeps it invisible until failure — the dependence-on-other-assumptions that lets it be defended, protected, or discharged

An assumption is present when each of the following holds:

  • An unproven proposition (the held input). A claim treated as if true for the purposes of some reasoning without being currently demonstrated within it — distinct from a fact (demonstrated), a hypothesis (offered for test), and a narrow-logical premise (dischargeable within a proof).
  • A dependent activity (the resting structure). An inference, model, design, plan, or message that must proceed, and that rests on the proposition rather than re-deriving it.
  • A downstream commitment (the loaded conclusion). The conclusion or output whose truth is loaded onto the proposition, such that the conclusion would fall if the proposition were false.
  • Load (the weight-bearing invariant). The proposition bears the weight of some downstream commitment; the load is located by asking which conclusions would fail if it were false, and high-load assumptions are the consequential ones.
  • Transparency (the invisibility invariant). Assumptions exist whether or not the assumer is aware of them and are usually invisible until violated, when a downstream conclusion fails inexplicably though the explicit chain was sound — which is why surfacing implicit assumptions is itself a competence.
  • Interaction (the dependence invariant). An assumption can be defended by another assumption in a regress, protected by a procedure, or replaced by an evidenced claim; assumptions form a layer beneath the explicit chain, not isolated points.

The components compose into a transparent layer between what is given and what is concluded, so the trustworthiness of any conclusion is bounded by that of its weakest load-bearing assumption.

What It Is Not

  • Not a distributional assumption specifically. distributional_assumption is one species — an assumption about the probabilistic form of data (normality, independence, stationarity). Assumption is the genus: any load-bearing proposition held as if true, of which distributional ones are a statistics-bound case.
  • Not an axiom. axiom is held as foundationally true within a formal system and is never expected to be discharged or replaced; an assumption is held provisionally as if true for a purpose and is a candidate for justification, testing, or revision.
  • Not a hypothesis. A hypothesis is offered for test — its whole point is to be confronted with evidence. An assumption is relied on while reasoning proceeds, precisely so the activity need not stop to test it.
  • Not belief. belief_formation concerns what an agent holds as true and how it comes to hold it; an assumption is held as if true for the purposes of some activity, whether or not the assumer believes it, and exists even when the assumer is unaware of it.
  • Not a premise in the narrow logical sense. A logical premise is explicit and may be discharged within a proof; an assumption is typically transparent — invisible until violated — and lives beneath the explicit chain rather than within it.
  • Not a constraint. A constraint limits the space of admissible solutions; an assumption supplies an input the reasoning rests on. A constraint that turns out false leaves the solution space mis-drawn; an assumption that turns out false breaks a conclusion that loaded its weight onto it.
  • Common misclassification. Re-checking the explicit chain when a sound-looking conclusion fails — debugging the audited arithmetic instead of the unexamined input. Catch it by asking which proposition the chain relied on without deriving: surprising failures concentrate in the transparent assumption layer, not the visible computation.

Broad Use

The load-bearing-belief-without-current-justification structure recurs wherever reasoning rests on unproven inputs. In logic and mathematics it is the canonical case — axioms, conditional-proof hypotheses, the premises of an argument — and identifying it is the first move of rigorous proof. In statistics and modeling, every model rests on distributional, independence, stationarity, and exchangeability assumptions, and assumption-checking is institutionalized as residual diagnostics and sensitivity analysis. In engineering and design, assumptions about load, environment, and operating conditions underlie any design, and failure analysis often locates the broken assumption rather than a flaw in the design as designed. In software, assumptions about input format, encoding, concurrency, and downstream stability are the substrate of defects, and contracts and types are technologies for making them explicit and checked. In empirical science, background theory and instrument calibration are auxiliary assumptions of any experiment, and the holistic character of testing means a failed test can indict the auxiliaries rather than the hypothesis. In communication, shared common-ground assumptions about referents and mutual knowledge are the substrate of intelligibility. In law, implied warranties and background presuppositions are assumed and surface by violation. And in planning, any plan rests on assumptions about demand, competition, supply, and regulation, with assumption-mapping a recognized discipline. Across all of these the same pattern appears: a transparent layer of unproven inputs bearing the weight of everything concluded above it.

Clarity

Naming the category of assumption separates what is being said from what is being relied on by the saying. It exposes that the trustworthiness of any conclusion is bounded by the trustworthiness of its weakest load-bearing assumption — a pervasive point that resists intuition precisely because the assumptions are usually invisible. It also clarifies the difference between fixing the conclusion and fixing the assumption: a wrong conclusion may not be repaired by re-deriving within the same assumption layer, only by revising what was assumed. The vocabulary further separates assumptions from neighboring roles in useful ways. A belief is held as true; an assumption is held as if true for some purpose; a hypothesis is held as a candidate for test; an axiom is held as foundationally true within a formal system; a fact has been demonstrated. The same proposition can play any of these roles in different contexts, and the role determines the appropriate intervention — justify, discharge, test, accept, or verify. Naming the assumption layer thus turns a vague sense that "something we took for granted went wrong" into a precise location: the unproven input that bore the weight and failed.

Manages Complexity

Assumption-management compresses a vast space of "things that could be true" to a small list of "things this work depends on being true." Once the load-bearing assumptions are listed, the analyst can ask the same three questions in any domain: which are strong (well-evidenced, unlikely to fail), which are weak (poorly-evidenced, plausible to fail), and which are load-bearing (their failure would invalidate the conclusion)? The cross-product of load and confidence yields an attention budget: weak-and-load-bearing assumptions are the intervention targets, while strong-and-non-load-bearing ones can be left alone. This is the same triage in every domain, and it is what makes the otherwise unbounded task of justifying a piece of reasoning tractable — rather than defending every proposition the work rests on, the analyst concentrates scarce attention on the few that are both shaky and consequential. By converting an open-ended worry ("what if we're wrong about something?") into a ranked, finite list, the pattern keeps the validation of complex inference, design, or planning within reach: the complexity of everything that could go wrong is reduced to the small set of assumptions that are both uncertain and load-bearing.

Abstract Reasoning

Recognizing the layer of assumptions supports the inference that any chain of reasoning is conditional on its assumptions, even when presented as unconditional. It supports the further inference that a surprising failure of a conclusion is more parsimoniously explained by a wrong assumption than by a flaw in the explicit chain — because the chain has been audited but the assumptions usually have not, so the unexamined layer is where surprising errors concentrate. The pattern licenses several cross-domain disciplines that all follow from its structure: sensitivity analysis (what range of assumption values yields what range of conclusion values), the inventory of implied warranties, the pre-mortem (what assumption must fail for the plan to fail), and the recognition that empirical theories are tested holistically rather than proposition-by-proposition. Each of these is a way of operating on the assumption layer rather than the explicit chain, and each is recoverable from the bare structure: because the conclusion is loaded onto the assumptions, perturbing them reveals the conclusion's true conditionality, and locating the load reveals which perturbations matter. The abstract leverage is thus the systematic interrogation of the conditional structure that every inference hides beneath its apparent unconditionality.

Knowledge Transfer

The transfers attach to the assumption-audit discipline itself, which carries across domains with minimal modification because it operates on the conditional structure of reasoning rather than on any field's content. The residual-check of model assumptions transfers to the load-margin-check of structural assumptions: list the assumptions, test the most load-bearing. The explicit-premise discipline of formal proof transfers to the named-warranty discipline of contracts, where unstated assumptions are the default failure source in both. The auxiliary-assumption awareness of experimental design transfers to the assumption-failure inventory of strategic planning, carrying the lesson that one cannot test a hypothesis, or pursue a strategy, without committing to the auxiliaries. And the hypothesis-discharge of conditional proof transfers to the precondition-checking of defensive code, where the assertions in a function are exactly the premises of its correctness claim. The deepest carry is a single universal principle made vivid by failure: the trustworthiness of any conclusion equals the trustworthiness of its least trustworthy load-bearing assumption, and that assumption may be invisible until it fails. A practitioner who has traced one rigorous-but-wrong conclusion — a sound analysis undone by an unrepresentative-population assumption — back to its broken assumption carries into every other domain the same move: surface the assumptions, locate the load, test or hedge the high-load ones, and accept the low-load failures. The intervention vocabulary is identical whether the broken assumption underlies a structural design, a regression, a negotiated agreement, a software release, or an act of communication, because in each the conclusion was conditional on an unproven input that bore its weight.

Examples

Formal/abstract

Ordinary least-squares regression is a precise instance of a conclusion resting on a transparent layer of load-bearing assumptions. The dependent activity is the estimation of a coefficient \(\hat{\beta}\); the downstream commitment is an inference such as "this predictor has a significant effect." Beneath it sits a stack of unproven propositions held as if true: linearity of the conditional mean, independence of errors, homoskedasticity (constant error variance), and — for the \(p\)-value — approximate normality of the sampling distribution. Each has load that can be located by the prime's test — which conclusion falls if it is false? Violate independence (say, the observations are a time series with autocorrelated errors) and the point estimate \(\hat{\beta}\) can remain unbiased while its standard error is badly understated, so the significance conclusion fails though the arithmetic was correct and the data were clean — the prime's signature of transparency: the assumption was invisible until it broke, and the failure looks inexplicable from inside the explicit chain. The interaction invariant appears too: the normality assumption can be protected by a procedure (a large-sample appeal to the central limit theorem) or replaced by an evidenced alternative (a bootstrap that re-derives the sampling distribution from the data). The discipline the prime prescribes is residual diagnostics and sensitivity analysis: list the assumptions, find the high-load ones (here, independence and homoskedasticity dominate the inference), and test or hedge those rather than re-checking the already-audited algebra.

Mapped back: The regression instantiates the full signature — unproven propositions, a resting inferential activity, a loaded conclusion, located load, transparency until violation, and assumptions that can be protected or replaced — and shows the prime's central claim concretely: the trustworthiness of the significance conclusion is bounded by its weakest load-bearing assumption, not by the correctness of the visible computation.

Applied/industry

A spacecraft-software defect makes the same structure vivid across an engineering and a software substrate at once. A navigation module computes thrust from a sensor reading, assuming the reading is in metric units; an upstream component, written by a different team, emits it in imperial units. The unproven proposition — "the interface speaks SI" — is a high-load assumption: the entire trajectory conclusion is loaded onto it. It is also perfectly transparent: every line of code is correct, every calculation executes, the data are clean, and yet the vehicle's path is wrong in a way that is inexplicable from inspecting either module alone, because the failure lives in the unexamined layer between them, not in the explicit chain. This is the prime's diagnosis exactly — surface the implicit assumption and the mystery dissolves. The interaction invariant shows the cure: the assumption can be made explicit and checked by a type or a unit-bearing contract at the interface, converting a silent load-bearing belief into a verified claim, which is what defensive contracts and typed boundaries are for. The same triage transfers to a business plan whose revenue conclusion rests on an unstated assumption that a key supplier's prices hold — weak-and-load-bearing, the prime's intervention target — and to an experiment whose result rests on an auxiliary calibration assumption, so a failed test may indict the instrument rather than the hypothesis. In each, the move is identical: inventory the assumptions, rank by load and confidence, and harden the few that are both shaky and consequential.

Mapped back: The units-mismatch case runs the prime end-to-end — a load-bearing, transparent assumption whose violation breaks a sound chain inexplicably, repaired by making it explicit and checked — and demonstrates the cross-domain transfer the prime promises: the same surface-locate-test discipline applies whether the broken assumption underlies a flight computer, a regression, a budget, or an experiment.

Structural Tensions

T1 — Surfacing versus Proceeding (Temporal Trade-off). Assumptions exist so that an activity can proceed without re-deriving its inputs; making them all explicit would halt the work. The tension is between auditing the layer and getting anything done. The failure mode at one pole is analysis paralysis — surfacing every assumption until no inference completes — and at the other blind momentum, proceeding on an unexamined stack. Diagnostic: triage by load, not by ease of surfacing; spend the audit budget only on assumptions whose failure would invalidate the conclusion, and consciously accept the low-load ones rather than either auditing all or ignoring all.

T2 — Load versus Confidence (Two-Axis Triage). An assumption's danger is the product of how much weight it bears and how likely it is to fail, but attention drifts to whichever axis is salient — usually the shaky-but-harmless or the load-bearing-but-solid. The failure mode is misallocated scrutiny: hardening a weak assumption that bears no load while a strong-seeming, untested, high-load one goes unchecked. Diagnostic: for each assumption ask both "which conclusion falls if this is false?" and "how likely is it false?" — only the intersection (weak and load-bearing) deserves intervention; vigilance on either axis alone wastes the budget and misses the real exposure.

T3 — Transparency versus Visibility (Failure Locus). The defining hazard is that assumptions are invisible until they break, so a failed conclusion looks inexplicable from inside the audited explicit chain. The tension is that the chain is where everyone looks and the assumption layer is where the error usually lives. The failure mode is re-checking the algebra: debugging the visible computation again and again when the fault is an unstated input the computation rested on. Diagnostic: when a conclusion fails though the explicit steps are demonstrably correct, stop re-auditing the chain and enumerate what the chain assumed — surprising failures concentrate in the unexamined layer, not the examined one.

T4 — Fix the Conclusion versus Fix the Assumption (Scopal Repair). A wrong conclusion cannot be repaired by re-deriving within the same assumption layer; only revising the assumption fixes it. The tension is that conclusion-level patches are local and cheap, assumption-level revision is structural and expensive. The failure mode is patching downstream: special-casing the broken output, which holds until the same violated assumption breaks a different conclusion. Diagnostic: ask whether the proposed fix changes any assumption or only the conclusion; if the input that bore the weight is untouched, the repair is cosmetic and the failure will recur through another path that loads the same assumption.

T5 — Holistic versus Atomic Testing (Coupling). Assumptions interact — one is defended by another, protected by a procedure, or bundled with auxiliaries — so a failed test indicts a conjunction, not a single proposition. The tension is with the wish to falsify one assumption at a time. The failure mode is misattributed refutation: concluding the hypothesis is wrong when an auxiliary (calibration, instrument, background theory) actually failed, or vice versa. Diagnostic: list the auxiliary assumptions riding along with the one under test; if any could independently produce the observed failure, the test does not isolate the target, and the refutation must be apportioned across the bundle rather than pinned on one input.

T6 — Explicit Contract versus Residual Assumption (Completeness Limit). Making assumptions explicit — types, contracts, named warranties — converts silent beliefs into checked claims, but the move never terminates: every contract rests on further assumptions (that the type system is sound, that the checker ran, that the spec is right). The tension is between the value of explicitness and the impossibility of total explicitness. The failure mode is contract complacency: trusting a typed or contracted boundary as assumption-free, blind to the meta-assumptions the contract itself loads. Diagnostic: ask what the contract assumes to be true; a boundary that checks units still assumes the checker is correct and the spec captures intent, so explicitness relocates the residual assumption rather than eliminating it.

Structural–Framed Character

Assumption sits near the structural end of the structural–framed spectrum, with a frontmatter aggregate of 0.2 that records a structural pattern with a faint practice-bound tint. The core is bare relational structure: a load-bearing proposition held as if true, beneath a chain of reasoning, transparent until it fails. That skeleton — a transparent layer between what is given and what is concluded, with the conclusion's trustworthiness bounded by its weakest load-bearing input — is medium-neutral and travels through any inferential substrate.

Three diagnostics read flatly zero and set the structural baseline. The pattern carries no home vocabulary that must travel (vocab_travels 0.0): the same load-bearing-input structure appears as a unit convention in software, a load envelope in engineering, a common-ground referent in communication, and a distributional premise in statistics, each named in its own field's words. It carries no evaluative weight (evaluative_weight 0.0): an assumption is neither good nor bad, merely held — its danger is a product of load and confidence, not of any inherent approval or disapproval. And its origin is formal (institutional_origin 0.0): the notion of a proposition relied on without current demonstration is a piece of logic, not a product of any human institution.

What lifts it just off the pole is the two half-points, both honest. The pattern is faintly human-practice-bound (human_practice_bound 0.5): an assumption exists only relative to some reasoning or activity, so it presupposes an inferential process — yet that process need not be human, since any inferential substrate (a typed program, a formal proof, a model) carries assumptions, which is why the score is a half rather than a full point. And invoking it half-imports (import_vs_recognize 0.5): to call something an assumption is partly to recognize an input already bearing weight, but partly to take up the interpretive stance of auditing the conditional structure of a chain — a move tied to the practice of reasoning rather than read off a physical system. The 0.2 aggregate is the right reading: a structurally clean, vocabulary-free pattern whose only departure from the pole is its inseparability from the activity of inferring.

Substrate Independence

Assumption is a strongly substrate-independent prime — composite 4 / 5 on the substrate-independence scale. Its breadth is at the ceiling (domain breadth 5): the load-bearing-belief-without-current-justification structure recurs with the same force across logic and mathematics (axioms, conditional-proof hypotheses), statistics (distributional, independence, stationarity premises), engineering (load and operating-condition assumptions), software (input-format, encoding, unit conventions), empirical science (auxiliary and calibration assumptions), communication (common-ground referents), law (implied warranties), and planning (demand and supply assumptions) — distinct media in which the same transparent layer bears the weight of everything concluded above it. Structural abstraction sits at 4: the role — a transparent layer between what is given and what is concluded, whose trustworthiness bounds the conclusion's — is medium-neutral and the vocabulary itself is bare, but the pattern is inseparable from some inferential activity, so it presupposes a reasoning process (human or not) rather than running in a mute physical loop. Transfer evidence is concrete (4): the assumption-audit discipline — inventory, locate the load, test or hedge the high-load few — ports unchanged from regression residual-checks to structural load-margins to named contracts to pre-mortems, and the units-mismatch failure mode is recognizably the same defect across a flight computer and a spreadsheet. The composite of 4 records a pattern recognized across nearly every inferential domain, held just short of 5 only by its dependence on the activity of inferring.

  • Composite substrate independence — 4 / 5
  • Domain breadth — 5 / 5
  • Structural abstraction — 4 / 5
  • Transfer evidence — 4 / 5

Relationships to Other Primes

One-hop neighborhood: parents above, mutual partners to the right, children below.Assumptionsubsumption: Epistemic Mode Of A PropositionEpistemic ModeOf A Propositionsubsumption: Distributional AssumptionDistributionalAssumption

Parents (1) — more general patterns this builds on

  • Assumption is a kind of Epistemic Mode Of A Proposition

    The file: assumption is ONE VALUE of the mode dimension (held-as-if-true provisionally). Dimension-to-value, like temperature-to-hot. Clean child; nearest neighbor (0.82).

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

  • Distributional Assumption is a kind of Assumption

    A distributional assumption IS an assumption — the statistics-bound species (a load-bearing premise about the probabilistic form of data). assumption is the genus; the child's own DfN already names 'assumption in the broader sense' as its parent.

Path to root: Assumption → Epistemic Mode Of A Proposition

Neighborhood in Abstraction Space

Assumption sits in a moderately populated region (45th percentile for distinctiveness): it has near-neighbors but no dense thicket of synonyms.

Family — Logical Moves & Precondition Gating (10 primes)

Nearest neighbors

Computed from structural-signature embeddings · 2026-06-14

Not to Be Confused With

The near-identical embedding neighbor is distributional_assumption (similarity 0.997), and the relation is a clean genus/species one that should be stated explicitly because dedup may need to reconcile them. A distributional assumption is a load-bearing proposition about the probabilistic structure of data — that errors are normal, that observations are independent, that a process is stationary or exchangeable. It is a fully legitimate, sharply scoped pattern within statistics and modeling. But everything that makes it work — that it is held as if true, that it bears load, that it is transparent until it breaks, that its failure invalidates a conclusion while the arithmetic remains correct — is inherited from assumption-in-general. The genus covers unit conventions in software, load envelopes in engineering, common-ground referents in communication, supplier-price stability in a business plan, and instrument calibration in an experiment; the distributional species covers only the probabilistic subset. Keeping the parent/child relation explicit prevents two errors: treating all assumptions as if they were statistical (and so reaching for residual diagnostics where the real exposure is a units mismatch or a missing warranty), and treating distributional assumptions as a wholly separate concern rather than as the statistics-bound instance of the universal load-bearing-input pattern.

A second genuine confusion is with axiom. Both are propositions accepted without proof at the point of use, but they differ in standing and in expected fate. An axiom is foundational within a formal system — it is held as true by stipulation, is not a candidate for evidence, and is never expected to be discharged or revised so long as one works inside that system; changing it means changing the system. An assumption is provisional — held as if true so an activity can proceed, with the standing expectation that it may be justified, tested, hedged, or replaced as inquiry advances. The empirical sciences run on assumptions, which is why they revise; formal systems rest on axioms, which is why they are stable until redefined. Confusing the two produces either misplaced rigidity (defending an assumption as if questioning it were a category error) or misplaced revisionism (treating the axioms of a formal system as empirical bets).

A third confusion is with belief_formation. A belief is a proposition an agent holds as true, with a psychology of how it was acquired; an assumption is a proposition treated as if true for the purposes of some reasoning, decoupled from whether anyone believes it. A modeler can assume normality while actively disbelieving it, holding it provisionally because the activity must proceed and the approximation is good enough. Crucially, assumptions exist whether or not the assumer is aware of them — the transparency invariant — whereas a belief is, by definition, something the agent holds. The discriminating question is whether the proposition is doing load-bearing work in a chain of reasoning (assumption) or describing the agent's doxastic state (belief); the same sentence can be one, the other, or both depending on the role it plays.

For a practitioner these distinctions route the right repair. Confusing assumption with distributional assumption narrows the audit to statistical diagnostics and misses non-probabilistic load-bearing inputs. Confusing it with axiom misjudges whether the input is revisable. Confusing it with belief obscures that the dangerous assumptions are precisely the ones nobody consciously holds — the transparent, unexamined inputs that fail silently. The unifying move is to inventory the load-bearing propositions a piece of work rests on, classify each by role (foundational axiom, provisional assumption, tested hypothesis, held belief, established fact), and concentrate scrutiny on the provisional, high-load, transparent ones — because those are where sound chains break inexplicably.

Solution Archetypes

No catalogued solution archetypes reference this prime yet.