Assistant Professor Michigan State University, Michigan, United States
Overview: The overview of this proposal was to explore strategies and future recommendations to prevent bot and fraudulent responses on online surveys.Proposal text: Community-based research among underserved or marginalized populations often includes recruitment via social media, listservs, and other mechanisms because no previously established sampling frame is available. Yet, bot responses and fraudulent participants in online surveys is increasingly becoming problematic, particularly in studies which aim to provide an incentive to thank participants for their time. This presentation addresses lessons learned from a study exploring sexual satisfaction, aging, and health among gay men aged 50 or older in the Midwest. This population is considered hard-to-reach as this generation of gay men grew up in a time where being gay was highly stigmatized; furthermore, there are no available sampling frames that include adequate samples of older gay men. In addition, the timing of this study was such that the pandemic limited face-to-face interactions due to Covid-19 risk among older populations. A web-based survey was chosen as the data collection method for the study to protect respondents’ privacy as well as to decrease risk of transmission of Covid-19. During the data collection phase, the researchers encountered multiple very sophisticated instances of artificial intelligence (A.I.) and scammers attempting to take the survey for the incentive, making it extremely difficult to discern which survey responses were valid. While several security measures were enabled via Qualtrics, such as the prevention multiple entries and reCaptcha, A.I. and scammers were able to bypass these security measures with ease. The researchers developed a toolkit to decipher whether a respondent was legitimate or not. This toolkit included 1) asking potential respondents to email the researcher for a unique survey link; 2) examining the respondents’ email address and messages for clues about whether the respondent was legitimate; and 3) screening the respondent’s IP address location and origin for potential fraudulence. As A.I. becomes more sophisticated at bypassing security measures and answering relevant open-ended questions on surveys, researchers (especially those that lack cybersecurity knowledge) must be vigilant to ensure that the data collected via online surveys are valid responses while still protecting the privacy and confidentiality of true respondents.
