Most online surveys still look and feel much like paper questionnaires. In particular, although the web is an interactive medium, web surveys rarely exploit this interactive capability. One exception is a series of demonstrations that online survey respondents seem to become more conscientious when prompted (usually with a brief textual message) in response to behaviors like item nonresponse, nondifferentiation, and answering very quickly. While these earlier studies have found that interactive intervention can reduce the occurrence of these behaviors, the underlying mechanism-why the intervention works-remains unclear. To shed light on this, we conducted two experiments to explore why respondents might change their behavior after being prompted. Part of the explanation lies in whether the benefits of the intervention are specific to the targeted behavior or lead to a general increase in conscientious responding. The findings show that intervention can lead to genuine improvement in respondent behaviors, but it also runs the risk of producing socially desirable bias in survey answers.