Privacy, Power, and Invisible Labor on Amazon Mechanical Turk

Abstract

Tasks on crowdsourcing platforms such as Amazon Mechanical Turk often request workers' personal information, raising privacy risks that may be exacerbated by requester-worker power dynamics. We interviewed 14 workers to understand how they navigate these risks. We found that Turkers' decisions to provide personal information during tasks were based on evaluations of the pay rate, the requester, the purpose, and the perceived sensitivity of the request. Participants also engaged in multiple privacy-protective behaviors, such as abandoning tasks or providing inaccurate data, though there were costs associated with these behaviors, such as wasted time and risk of rejection. Finally, their privacy concerns and practices evolved as they learned about both the platform and worker-designed tools and forums. These findings deepen our understanding of both privacy decision-making and invisible labor in paid crowdsourcing, and emphasize a general need to understand how privacy stances change over time.