How Well Do My Results Generalize? Comparing Security and Privacy Survey Results from MTurk, Web, and Telephone Samples

Abstract

Security and privacy researchers often rely on data collected from Amazon Mechanical Turk (MTurk) to evaluate security tools, to understand users' privacy preferences and to measure online behavior. Yet, little is known about how well Turkers' survey responses and performance on security-and privacy-related tasks generalizes to a broader population. This paper takes a first step toward understanding the generalizability of security and privacy user studies by comparing users' selfreports of their security and privacy knowledge, past experiences, advice sources, and behavior across samples collected using MTurk (n=480), a census-representative web-panel (n=428), and a probabilistic telephone sample (n=3,000) statistically weighted to be accurate within 2.7% of the true prevalence in the U.S. Surprisingly, the results suggest that: (1) MTurk responses regarding security and privacy experiences, advice sources, and knowledge are more representative of the U.S. population than are responses from the census-representative panel; (2) MTurk and general population reports of security and privacy experiences, knowledge, and advice sources are quite similar for respondents who are younger than 50 or who have some college education; and (3) respondents' answers to the survey questions we ask are stable over time and robust to relevant, broadlyreported news events. Further, differences in responses cannot be ameliorated with simple demographic weighting, possibly because MTurk and panel participants have more internet experience compared to their demographic peers. Together, these findings lend tempered support for the generalizability of prior crowdsourced security and privacy user studies; provide context to more accurately interpret the results of such studies; and suggest rich directions for future work to mitigate experiencerather than demographic-related sample biases.

An empirical study on the perceived fairness of realistic, imperfect machine learning models

“…We surveyed 502 individuals in a convenience sample from Mechanical Turk. Table 2 summarizes our participants' demographics, which we compare to those of a recent study by Redmiles et al [20] comparing sampling methods. Our participant population skewed male.…”

“…In spite of this, most participants appeared to understand the mathematical concepts we presented to them. We tested participants' understanding of the graphs in our survey with a series of [20]. A dash indicates that the study did not use that particular category.…”

Work in Progress: A Comparative Long-Term Study of Fallback Authentication

“…While MTurk is widely used, Prolific is explicitly set up for academic research studies. Analyses have shown that MTurk yields reasonable results, even though not adhering to strict standards of experimental design [27]. Research has shown that participants from Prolific are more equally distributed across the world, and tend to answer more honest than people from MTurk while the data quality is comparable [23].…”

The importance of social factors in the association between physical activity and depression in children

“…One reason for the paucity of research on social components of youth activities may be previous methodological constraints. However, recent advances in technology such as Amazon's Mechanical Turk (MTurk) provide reliable and efficient online platforms for collecting psychological and other data [60,65]. MTurk offers a flexible, easy-to-implement method for "crowdsourcing" opinions [15], such as those about the dynamics of socialand non-social youth activities.…”