Quantifying the qualities of language.

Abstract

We here quantify the qualities of language. Specifically, we derive numerical values, and associated variability scores for statements of general probability as well as for anchor terms commonly expressed on Likert-type scales. Our results show significant inter-responder variability in the understanding of such terms. We also demonstrate scale compression such that average estimates are poorly aligned to the a priori expectations of equal response spacing in commonly used verbal response instruments. Terms further show intriguing propensities such that people agree more about statements of agreement, but manifestly disagree more about statements of disagreement. The valence of these terms is thus systematically related to their degree of common understanding. The accuracy of language, especially when used in conjunction with modern advanced technologies, proves vital for effective communication. Our work demonstrates that the tool of language still remains a fairly blunt one, but one that can be sharpened by systematic quantitative evaluation, as we demonstrate here.