Tensions in Practice: Measuring more in tension with assuming more¶
An invented two-part total
A total of 10 could come from parts 4 and 6, or from 7 and 3. Repeating the same exact total cannot choose between these two declared possibilities. Measuring the first part can separate them; assuming that the first part is below 5 also leaves only one, but for a different reason. The assumption must not be presented as another measurement.
Separate the candidates with evidence
Obtain an observation on which their predictions differ.
Work within a justified restriction
Reduce the admissible possibilities using defensible prior constraints.
Why these aims pull against each other
A new channel costs measurement effort; a restriction can save that effort but makes uniqueness conditional on the restriction being true.
Choose an arrangement to see what changes and what remains difficult.
Arrows express the declared relations, not measured effect sizes. Examples and quantities are illustrative.
What this choice protects
What it costs
When it fits
Compare the arrangements
Measure a separating quantity
Measure the first component in addition to the total.
- What it protects
- The candidates predict different readings, making them distinguishable.
- What it costs
- The extra measurement needs access and a suitable instrument.
- When it fits
- The first component is observable and the reading is informative enough for the target.
Illustration note: Only two candidates and exact readings are assumed; this is not a noisy estimator.
Declare a restriction
Keep the total and assume the first component is below 5.
- What it protects
- The restricted two-candidate model has one remaining configuration.
- What it costs
- If the restriction is wrong, the excluded candidate could be the true one.
- When it fits
- The restriction has independent justification and is clearly reported.
Illustration note: An assumption eliminates a candidate; it does not create evidence that the first part was measured.
What this illustration does—and does not—establish
The source supplies the stated tension; the selected arrangements are bounded editorial illustrations. Costs and conditions remain part of the comparison.
- All candidate values are invented. In a larger candidate space, either step may still leave several possibilities.
- Uniqueness in an exact model does not establish practical precision under noise.
- The observation-versus-assumption distinction is about what resolved the ambiguity, not a universal ranking of measurement over domain knowledge.
Source entries
Identifiability
This source passage supplies the contextual tension. The concrete arrangements and schematic examples are editorial illustrations, not measured findings.
Channel Enrichment versus Restriction-by-Prior (Resolution Locus)
T3 — Channel Enrichment versus Restriction-by-Prior (Resolution Locus). A non-singleton class can be collapsed by enriching the observation channel (a new measurement) or by ruling out members on prior grounds — and the two differ sharply in honesty and cost. The failure mode is *silent prior smuggling*: collapsing the class with a default prior or modeling assumption nobody recognized as load-bearing, so apparent identification is an artifact of an undeclared restriction. Diagnostic: ask whether uniqueness came from new data or from an assumption; if a parameter is "identified" only because a prior ruled out the competing members, that prior is doing the work and must be declared and defended, not hidden in a default.
The source operation
Identifiability is the structural condition under which an *internal unknown* — a parameter, mechanism, causal effect, hidden state, or latent variable — is in principle recoverable from the *observable signal* the system makes available. The defining commitment is a *uniqueness* claim: across the space of admissible internal models, the mapping from internals to observations is one-to-one within whatever subspace the observation can see. When two distinct internal configurations produce the same observable distribution, the underlying object is *unidentified*, and no amount of additional data of the same kind can distinguish them — only a structural intervention (a new measurement channel, an experimental manipulation, a parametric restriction, a prior commitment) can.