Using Text-to-speech (TTS) for Audio Computer-assisted Self-interviewing (ACASI)

Abstract

We evaluate the use of text-to-speech (TTS) technology for audio computer-assisted self-interviewing (ACASI). We use a quasi-experimental design, comparing the use of recorded human voice in the 2006–2010 National Survey of Family Growth with the use of TTS in the first year of the 2011–2013 survey, where the essential survey conditions are largely unchanged. We examine substantive distributions of ACASI items, item missing data rates, interviewer observations, and time stamps. We find no negative effect of the transition to TTS. We discuss the advantages of using TTS for ACASI.