Award-Winning Paper: “Privacy Attitudes of Smart Speaker Users”


For the tenth year, FPF’s annual Privacy Papers for Policymakers program is presenting to lawmakers and regulators award-winning research representing a diversity of perspectives, including those from students and academics. Among the papers to be honored at an event at the Hart Senate Office Building on February 6, 2020 is Privacy Attitudes of Smart Speaker Users by Nathan Malkin, PhD student in computer science at University of California, Berkeley, and his coauthors. The study surveys privacy attitudes of smart speaker users and presents an evaluation of users’ comprehension and use of existing privacy settings and controls.

In Privacy Attitudes of Smart Speaker Users, study authors Nathan Malkin, Joe Deatrick, Allen Tong, Primal Wijesekera, Serge Egelman, and David Wagner surveyed 116 owners of Amazon and Google smart speakers.

In an effort to understand whether smart speaker users are making informed decisions about the privacy consequences and controls offered by smart speakers, the authors used recordings of real interactions between the study participants and their devices. The authors found that “almost half did not know that their recordings were being permanently stored and that they could review them; only a quarter reported reviewing interactions, and very few had ever deleted any.” The authors found that the way smart speakers default to permanent storage of interactions with users places an “undue burden” on the user, and “is almost certain to result in most interactions going unreviewed.” However, the authors observe that, after the conclusion of the study, both Google and Amazon updated their voice assistants to allow for automatic data deletion after three or 18 months (Google) or the deletion of a day’s worth of recorded interactions (Amazon).

While more than 71% of smart speaker users had not raised privacy concerns related to their device in the past, the authors are careful to state that people are not “apathetic” about their privacy. Instead, people’s acceptance of smart speakers is tied closely to what is happening with their data, as well as the specific subjects in the speaker’s recordings. According to the study, more than 72% of users found recordings reviewed by a computer to be acceptable, while users were more likely to view human review of recordings as unacceptable. Additionally, users found certain interactions with smart speakers more sensitive than others. Recordings of children, financial information, sexual or medical topics, locations, and personally identifying information were viewed as particularly sensitive.

The authors conclude that privacy controls on smart speakers are underutilized. Based on the views of study participants, the authors suggest that voice assistants adopt shorter retention periods, despite the fact that users did not feel that their stored recordings presented a grave privacy danger. The authors may have revealed an important insight about individual perspectives on privacy: “people seem more protective of the privacy of others” than their own privacy.

If you’re interested in reading more about the attitudes of smart speaker users toward privacy, you’ll want to check out the full paper.

The Privacy Papers for Policymakers project’s goal is to put diverse academic perspectives in front of policymakers to inform the development of privacy legislation. You can view all of this year’s award-winning papers on the FPF website.