Culture as Consensus: A Theory of Culture and Informant Accuracy

TLDR

The paper presents and tests a formal mathematical model for analyzing informant responses to systematic interview questions. The authors develop a Bayesian model that simultaneously estimates each informant’s cultural competence and the true answers to questions, handles true‑false, multiple‑choice, and fill‑in‑the‑blank formats, and provides guidance on the number of informants needed for desired confidence. The model yields accurate results with as few as four informants in familiar domains, and the authors discuss its implications.

Abstract

This paper presents and tests a formal mathematical model for the analysis of informant responses to systematic interview questions. We assume a situation in which the ethnographer does not know how much each informant knows about the cultural domain under consideration nor the answers to the questions. The model simultaneously provides an estimate of the cultural competence or knowledge of each informant and an estimate of the correct answer to each question asked of the informant. The model currently handles true‐false, multiple‐choice, andfill‐in‐the‐blank type question formats. In familiar cultural domains the model produces good results from as few as four informants. The paper includes a table showing the number of informants needed to provide stated levels of confidence given the mean level of knowledge among the informants. Implications are discussed.