Bias is the tendency for a whole set of responses to read high or low because of some inherent difficulty with the measurement process. (A chipped die may have a bias toward 6's, in that it comes up 6 much more often than we would expect.) In the study of HIV-related risk groups discussed above, people in the survey were asked intimate questions about their personal lives as part of a telephone interview. It there a possibility of bias in the responses? If so, in which direction? Will randomization in selecting the respondents help reduce potential bias related to the sensitive questions? Will randomization in selecting the respondents help reduce any potential bias?