Our Latest Research

We conducted a survey using 12 total AI voices from three leading Text-to-Speech (TTS) providers: Rime, ElevenLabs, and Google. After placing nearly 100,000 calls to respondents across the country, we evaluated which voices led to the most survey completions, fewest "Do Not Call" requests, and fewest respondent hangups during the introduction.

Following up on an earlier study alongside NORC at the University of Chicago, we take a closer look at calling patterns for respondents who complete surveys inbound. We study the relative completion rates between outbound vs inbound calls and show that respondents sometimes call outside of standard business hours -- even as early as 5 AM or as late as 11 PM!

Alongside ICF, we helped conduct the largest continuously conducted public health telephone survey in the world, the CDC's Behavioral Risk Factor Surveillance System (BRFSS) in Florida and Ohio. This study involved a human agent to AI interviewer handoff, and some respondents reported greater comfort speaking to AI about sensitive topics than a human interviewer.

Published in collaboration with Professor Trent D. Buskirk, this paper weighs the positives and negatives of AI interviewers compared to traditional IVR systems for both qualitative and quantitative data collection. We find that AI interviewers outperform traditional IVR systems and that, for now, AI interviewers are a better fit for quantitative than qualitative data collection.

The first published research on using an AI voice interviewer at scale. Conducted in partnership with SSRS, this research showed that respondents report positive experiences when being interviewed by a Miravoice AI Interviewer, and that some even prefer it to traditional quantitative CATI surveys.

In partnership with NORC at the University of Chicago, we demonstrated that Miravoice AI interviewers can conduct lengthy quantitative survey interviews at scale while delivering a positive respondent experience. The study also found that a significant number of participants preferred to complete surveys by calling an inbound number, including during non-standard hours.