Artificial Intelligence: Comparing Survey Responses for Online and Offline Samples.

Over the past decade, an increasing amount of scholars and professionals have turned to the Internet to conduct research ranging from public opinion surveys to complicated psychology and economics experiments. While there has been a focus on whether online samples are reliable and accurate, less research examines the behavioral differences between online and offline respondents. This paper focuses on one such potential difference. I use an experiment to gauge whether online respondents augment their knowledge with outside research. In the experiment, I ask a sample of online and offline subjects to answer a battery of factual knowledge questions. The results show that voters in the online treatment group are significantly more likely to answer questions correctly - a relationship that strengthens when question difficulty increases. These findings suggest that some - but not all - respondents completing surveys online are supplementing their answers with research. [ABSTRACT FROM AUTHOR]