Computer-Assisted Self-Interviewing (CASI)

Computer assisted self-interviewing (CASI) is a technique for survey data collection in which the respondent uses a computer to complete the survey questionnaire without an interviewer administering it to the respondent. This assumes the respondent can read well (enough) or that the respondent can hear the questions well in cases in which the questions are prerecorded and the audio is played back for the respondent one question at a time (audio computer assisted self-interviewing—ACASI).

A primary rationale for CASI is that some questions are so sensitive that if researchers hope to obtain an accurate answer, respondents must use a highly confidential method of responding. For a successful CASI effort, the survey effort must consider three factors: (1) the design of the questions, (2) the limitations of the respondent, and (3) the appropriate computing platform.

Unless one has a remarkable set of respondents, the sort of instrument needed for CASI (or any self-administered interview) will be different from what one uses when a trained interviewer is administering the questionnaire. Having a seasoned interviewer handling the questioning offers a margin of error when designing questions. When the researcher is insufficiently clear, she or he essentially counts on the interviewers to save the day. Their help comes in a variety of forms. The interviewer can explain a question the respondent asks about. Good interviewing technique requires the interviewer to avoid leading the respondent or suggesting what the expected or “correct” response is. The interviewer can also help salvage bad questions when the respondent's answer reveals that, although the respondent showed no overt confusion about the question, it is clear that the respondent either did not understand the question or took the question to be something other than what was asked. The interviewer can also help out the questionnaire designer when, during a complex interview, it becomes clear to the interviewer that something has gone wrong with the programming and the item occurs in a branch of the questionnaire where it should not be. The interviewer can then try to put things right or at least supply a comment that will help the central office sort out the problem.

In all of these cases, the interviewer plays a crucial role in improving data quality. With a self-administered survey, regardless of mode, the safety net of a trained interviewer is not available. The circumstances of a self-administered interview put a real premium on clarity, computer assisted or not. The need for clarity is all the higher because there are no learning-curve effects for CASI—interviewers may do hundreds of cases, but with CASI essentially each respondent does just one. Thus, the question wording itself needs to be clear and self-contained so the respondent does not need to ask clarifying questions. Many surveys provide “help” screens to interviewers that have supporting information about a question, but that is not a good idea with CASI—using help screens violates the “Keep it simple” rule. Anything the respondent needs to see or read should be on the display screen the respondent sees, with no additional scrolling, clicking, or pressing of function keys. One should also be wary of question fills or elaborate skip patterns, since a simple error by the respondent can produce myriad problems that ripple through the rest of the instrument.

...