ESRA 2027 sessions by theme

Back to Overview of Sessions

Session Details

Online surveys are an established mode of data collection in survey research. Recent advances in generative AI have created new challenges for ensuring the integrity of survey data. Increasingly sophisticated automated tools can generate plausible and coherent survey responses, making it more difficult to distinguish genuine human participants from fraudulent or synthetic respondents.
At the same time, what constitutes a “bot” is becoming harder to define. The term may encompass automated scripts and AI-generated respondents, but also AI-assisted human respondents and other forms of non-genuine participation. This ambiguity further complicates the reliable identification of fraudulent or synthetic responses.
Established data quality checks rely on combinations of indicators such as response times and patterns, consistency or plausibility measures, open-ended answers, attention checks, and IP information. Yet as automated tools become better at producing coherent, human-like survey data, such indicators may become less effective at distinguishing genuine from non-genuine respondents. Detection therefore involves two risks: fraudulent responses may remain undetected and genuine respondents may be falsely flagged. Deleting flagged cases can change the composition of the analytical sample if certain subgroups are more likely to be flagged based on their response patterns. When these groups also differ in the outcomes being studied, exclusion decisions can shift substantive survey estimates.
Detection methods therefore require evaluation not only in terms of identifying fraud, but also regarding the consequences of classification and deletion for data quality.
This session invites conceptual, methodological, and empirical contributions addressing bots, fraud, and synthetic responses in online surveys. We welcome work on definitions of non-genuine responding, detection and prevention strategies, validation and benchmarking of detection methods, consequences of exclusion decisions for survey estimates, and the implications of generative AI for survey integrity. Contributions addressing related challenges in digital trace data are also encouraged.