Intelligent Personal Assistants (IPAs), such as Apple’s Siri, Google Now, and Microsoft’s Cortana, have evolved from simple voice-recognition tools with very limited functions, to powerful platforms utilizing natural language processing, machine learning, and leveraging contextual and location data to provide intuitive, and increasingly predictive, results. IPAs are capable of automating tasks (such as adding calendar events, sending text messages, or placing calls) via voice commands, with more advanced features including providing search results, location-specific navigational aids, perform calculations, make reservations, identify nearby music, and even tell jokes.
Today’s IPAs resemble the autonomous “Serendipity Engine” that Google CEO Eric Schmidt envisioned in 2010 as taking past experiences, likes and dislikes and use them, along with geolocation information, to provide users information about things that might interest them wherever they might be, essentially delivering information before they realize they want it (Morrison, 2010). Siri, Cortana, and Google Now all have predictive and proactive features, including providing pre-loaded search results based on past history, automatically alerting you to traffic congestion in advance of a meeting, suggesting nearby coffee shops or bars based on location and time of day, or providing updates of a sporting event for which you have previously searched.
Accompanying the usefulness of IPAs is an increasing sense of “creepiness” (Douglas, 2013; Ingram 2013) with the amount of personal and locational information that must be tracked and processed to fully realize their benefits. To provide their contextual and predictive power, IPAs routinely track users’ web searches, calendar appointments, location via GPS, contents of their inboxes, and the plethora of personal information collected by Internet and search providers for which privacy experts have long expressed concern (see, for example, Rotenberg, 2008; Tene, 2009; Zimmer, 2008a; Zimmer 2008b).
While the privacy practices of IPA providers are slowly becoming more transparent (McMillan, 2013), it remains unclear how well users understand the privacy risks, how they evaluate such risks against the added convenience and usefulness of IPAs, and how concerns about these risks shape their behaviors and information-seeking activities. Following related work mapping individuals’ privacy preferences and their actions in the context of internet and social media platforms (Gross & Acquisti, 2005; Hoofnagle, et al., 2010; Tsai, et al., 2006), this research will explore how users of IPAs understand and balance the perceived costs and benefits in terms of generating and distributing their personal information.