Challenging Vulnerability Perceptions towards Voice Activated Assistants: An Abstract

Artificial Intelligence (AI) refers to the ability of technology to acquire and apply knowledge and skills to a variety of settings. Intelligence is often likened to emotion, logic, reason, critique, creativity and ability to learn and/or problem-solve. Although these forms of intelligence are usually only perceived to be capable by humans, technology has advanced it applications and usage in being able to replicate these forms of intelligence for different purposes. For example, Amazon Alexa is perceived as being able to problem-solve through creative, logical and, occasionally, emotional, means. As such, AI agents can adopt different types of conversational styles when interact with users resulting in conversations characterised by a more social-oriented or task-oriented approach (Chattaraman et al. 2018). Although these capabilities are far beyond what technology has previously been capable of, AI is inherently limited surrounding its ability to plan, reason, move and empathise, thus leading to an apparent lack of user trust in their capabilities (PwC 2018). Accordingly, users are often reluctant to engage with AI due to negative perceptions surrounding the ability to perform tasks.

Alongside these perceived limitations of AI, users are often concerned over what happens once they have chosen to use it. Users are aware that Alexa is “always listening” and is capable of collecting vast amounts of consumer data, thus resulting in heightened privacy perceptions (PwC 2018). In this sense, vulnerability appears to be prevalent at both the pre-adoption and post-adoption of AI. Previous studies have observed that technology has the capabilities of increasing consumer vulnerability, which in turn hinders usage and adoption (Bucchia et al. 2020; Elms and Tinson 2012; Ratchford and Barnhart 2012). Yet, consumers vulnerability in interactions with AI conversational agents remains underexplored. Most importantly, the factors that can prevent individuals perception of vulnerability and then facilitate continuous usage are unknown.

By focusing on voice-activated assistants (VAs), this study wants to fill this gap and it aims to investigate how interaction with VAs can foster consumers trust and usage based on the AI conversational style. Specifically, we expect that VAs social-oriented conversations will increase consumers trust and usage. In addition, we suggest that this process will be mediated by the consumers perception of vulnerability such that when the VAs conversational style is social-oriented, consumer will feel less vulnerable and thus more willing to trust and use the AI agent. Finally, we propose that individuals’ locus of control will moderate such effect.

The study employs a mixed-method approach. Preliminary findings from the qualitative analysis offer a broad picture of how perceptions of vulnerability and trust are elicited in interactions with VAs. Further, the importance of the social nature of conversation and perceptions of vulnerability is clearly shown in this study.