Econometrica: Jan, 1974, Volume 42, Issue 1
A Metric Structure Derived from Subjective Judgments: Scaling Under Perfect and Imperfect Discrimination
https://doi.org/0012-9682(197401)42:1<55:AMSDFS>2.0.CO;2-1
p. 55-72
Robert Bartoszynski
Assume that to each pair (x, y) of elements of the set A there corresponds probability p(x, y) that y will be chosen as having more of a certain attribute, if the pair (x, y) is presented to the subject. It is shown that if the p(x, y) satisfy a certain set of (logically independent) conditions, then there exists an (essentially unique) measure f of the attribute of elements of A, such that probabilities p(x, y) depend only on differences between f(x) and f(y).