Perspective Accuracy Assessment for Everyday DisagreementsCould the other side sign what you just said about them?

Methodology: Thirty-nine original exercises across five formats: sixteen single-choice restatements, eight confidence ratings on a six-point scale with half of them worded the other way round, six select-every-that-applies exercises, five binary claims and four matching exercises. One response instruction is declared for the whole instrument and it is a KNOWLEDGE instruction - which restatement the person who holds the view would accept - never what the respondent would say. The confidence ratings carry their own instruction and are scored as a probability, never as an answer. Construct statement: it measures whether somebody can state a position they do not hold in a form its holder would accept, and whether they can tell when they have failed to. It does not measure who is right, political or social opinions, intelligence, verbal ability, empathy, agreeableness, or willingness to change one's mind. No item uses a partisan political subject; every disagreement is an ordinary one about a rota, a repair, a price, a booking or a rule. Scoring is calibration. Accuracy on the sixteen restatements is chance-corrected against an authored answer prior per option rather than against a uniform draw. The eight paired confidence ratings are scored with the Brier score, decomposed into calibration error and discrimination, and reported as a skill score against a respondent who is equally sure about everything - a reference that has perfect consistency and no information in it. A calibration claim from fewer than eight items is noise, which is why there are eight. Reliability is McDonald's omega estimated from item count and a stated assumed inter-item correlation, printed with what it permits; where it will not carry a number, the refusal is printed. Every reported figure carries its standard error band, drawn on the chart rather than described in a footnote. Constructs and sources: the ideological Turing test as a scored, judge-rated behavioural measure of perspective taking, with published pass rates of 54, 64 and 71 per cent across three topics and 1,200 blind judges (Brand, Brady and Stafford 2025); naive realism and why acceptance by the holder is the right criterion (Ross and Ward 1996); the finding across twenty-five experiments that instructing people to imagine another's perspective made them LESS accurate while asking the person directly made them substantially more accurate (Eyal, Steffel and Epley 2018); the illusion of explanatory depth (Rozenblit and Keil 2002), cited for the measurement claim only, because the intervention claim built on it has failed preregistered replication and this report says so on the page; speakers' overestimation of their own clarity (Keysar and Henly 2002); the ordering rule for engaging a position - restate, agree, learn, then object - which is Rapoport's and is normative rather than empirical (Rapoport 1960; Dennett 2013); the Brier score and its decomposition into reliability, resolution and uncertainty (Brier 1950; Murphy 1973); overconfidence and the calibration of subjective probability (Lichtenstein, Fischhoff and Phillips 1982); acquiescence and balanced keying (Baumgartner and Steenkamp 2001); and the Barnum effect (Forer 1949). All items are original works. No commercial critical-thinking instrument, mediation framework, argument-mapping product or report layout is reproduced or implied, and no affiliation with any assessment publisher exists or is claimed.