Which panel summary answers which question? Aggregating two-coordinate expert evidence, with an extended sensitivity analysis of a neutrosophic VIKOR panel

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

For two-coordinate annotations (μ, λ) of favourable and contrary evidence, the coordinatewise mean and the panel mean of common scores give a panel split between full support and full opposition the same summary as a panel of two neutral experts. Keeping the coordinates is what allows such panels to be told apart, and which summary does so depends on the question asked of the panel. This methodological note makes that dependence explicit. An organizing lemma identifies the summary that answers each family of threshold queries: the knowledge-order meet and join answer bounds on each kind of evidence, the extremes of the certainty and contradiction degrees answer queries about gluts, gaps and polarisation, and neither answers the other's queries. An extended sensitivity analysis of a published neutrosophic VIKOR panel (3 experts, 5 alternatives, 12 criteria), possible because the source printed every expert's score matrix, reproduces the published ranking exactly. The published winner is first under the mean at each of the five VIKOR parameter values examined, under three of five combination rules and more often than any other alternative across 27 expert resamples, while majority and optimistic rules favour another alternative; all 29 intermediate combined values reflect split ratings. An exploratory simulation compares extremes and moments of equal dimension across seven noise models and three classifier families, with Bonferroni-adjusted intervals over 312 contrasts. Moments were ahead in most cells, plain extremes only under uniform noise and trimmed extremes under contaminated and heavy-tailed noise, so no summary was preferable in general.