Asking questions with computers: Interaction in the computer assisted standardized survey interview.

Abstract

Many social situations involve the focused exchange of information between two people. In such interactions, it is the job of the persons either to elicit information and/or to provide information to the other person. In U.S. society, these interactions increasingly involve communication with a computer. This is a study of one such social situation, the computer assisted survey interview, and the focused interaction that takes place within it. The objective has been to gain a better understanding of how the interviewer manages communication with both the respondent and the computer in a computer assisted interview. Through comparison with paper questionnaire interviews, this study addresses the question: How do data collection technology and instrument design affect interviewer behavior in the standardized survey interview? A fundamental difference between computer assisted personal interviews (CAPI) and paper personal interviews (PAPI) is that a computer assisted interview involves <italic>two</italic> interactions: one between the interviewer and respondent, and one between the interviewer and the computer. Computer assistance is also expected to increase standardization, through enforced delivery of survey questions and automatic tailoring of question text. Data from 38 CAPI and 14 PAPI National Health Interview Survey (NHIS) laboratory interviews are analyzed. Multilevel multivariate and behavior coding analyses are used to assess the effects of computer assistance, design, and respondent characteristics on interviewer behavior. Conversation Analysis (CA) is then used to help explain results. Findings show that computer assistance <italic> per se</italic> does not increase standardization in the survey interview. Instrument design influences whether or not the potential benefits of computer assistance are realized. Among factors examined, household size and text design characteristics (such as capitalization and emphasis) are found to be significantly associated with non-standardized interviewing behavior (not asking questions and having problems reading text). Findings also show that interviewers vary in their behavior across technologies, providing support for the notion that CAPI and PAPI interviews have different interactional substrates, that is, involve different task requirements, and call for different interviewer skills and resources.