The effect of interviewer image in a virtual-world survey

Abstract

When survey respondents are consciously or unconsciously influenced by the characteristics of the interviewer, bias in the survey estimates may result. The effects of interviewer characteristics such as gender and race on survey estimates have been considered previously, but isolating and experimentally manipulating a single interviewer characteristic has been infeasible. With the advent of online virtual worlds, it is now possible to conduct experiments focusing on individual physical interviewer characteristics. We conducted an exploratory study, surveying 60 individuals in Second Life, an online virtual-world community in which the respondent and interviewer were both represented as avatars—three-dimensional representations of real-life individuals. To explore the effect of interviewer appearance on reported health behaviors and attitudes, we randomly assigned half of the survey respondents to a “thin” interviewer and half to a “heavy” interviewer. The data suggest that those who reported to the heavy interviewer were less likely to say their own avatar was attractive, reported less frequent real-life exercise, and reported a higher real-life body mass index, although because of the small number of participants, we did not detect statistically significant differences. The findings suggest that interviewer appearance may have a biasing effect on reports in virtual-world surveys—and perhaps in real-world surveys. Despite the lack of statistically significant findings, the study illustrates the future potential, benefits, and challenges to surveying and conducting methodological research in a virtual world.