Communication Errors and Repairs in the Automated Telephone Interviews

In: Sociologičeskij žurnal: Sociological journal, Band 25, Heft 2, S. 33-59

Abstract

This article presents the results of a robotic survey conducted by the Laboratory for Social Research Methodology (Institute for Social Analysis and Forecasting, RANEPA) with support from Voice Communication and the Social Validation. The goal of the survey was to select rural residents from an all-Russian sample. The 599 collected phone numbers were identified by the robot as rural. In order to accomplish this task, 64623 numbers were dialed, in 7310 cases connection was made and in 2703 cases the subscribers were asked the first question. Afterwards out of all identified cases (rural/city residents or "status not determined"), 300 were selected in order to conduct further encoding and listen to audio recordings. All of the respondents' answers were divided into two groups: coded (or having single-valued correspondence with the proposed closures), and non-coded, which the robot was supposed to fix. Five types of non-coded answers were identified: extended answer, question interruption, counter-question, alternative notice and a service frame. For each type of non-coded answer options for fixing and improving the script of the robotic questionnaire were proposed.