The Quantity of Interest
V&V is organized around the specific output a decision depends on; error and uncertainty are measured for that quantity, not in general.
Verifying What Matters
A simulation produces enormous amounts of output, but a decision usually depends on a few specific numbers: a peak temperature, a growth rate, an integrated yield. These are the quantities of interest. Verification and validation are organized around them, because error and uncertainty are meaningful only for a specific output. A code can be accurate for one quantity and inaccurate for another on the very same run.
Why the Choice Is Central
- Convergence and order of accuracy can differ between quantities; a local peak may converge slower than an integrated total.
- Uncertainty and sensitivity rankings depend on which output is examined.
- The validation metric compares the quantity of interest, not the whole field, to experiment.
Choosing It Honestly
The quantity of interest must be defined before the analysis, driven by the decision it supports, not chosen after seeing results. Choosing the output that happens to converge best or agree best with data is a subtle form of cheating that inflates apparent credibility. The decision defines the quantity; the quantity defines what must be verified, validated, and bounded.
Local Versus Integrated Quantities
Integrated quantities, such as a total energy or an average, often converge faster and are less sensitive than local, pointwise quantities, such as a peak value at a single location. A study that reports clean convergence for an integrated quantity says little about a local peak that the decision actually depends on. Each quantity of interest needs its own convergence, uncertainty, and validation evidence; borrowing another quantity's evidence is unjustified.
Framing the whole V&V effort around explicit quantities of interest keeps it focused and honest. It ties every convergence study, every error bar, and every validation comparison to a number a decision-maker cares about, and it prevents the diffuse claim that a code is generally good from standing in for the specific evidence that the number in question is trustworthy.