Privacy of AI-Based Voice Assistants: Understanding the Users' Perspective - A Purposive Review and a Roadmap for Elevating Research on Privacy from a User-Oriented Perspective.

Intelligent voice assistants (IVA) are on the rise: They are implemented into new smart mobile and stationary devices such as smartphones, smart watches, tablets, cars, and smart speakers. Being surrounded by always-on microphones, however, can be perceived as a threat to one's privacy since audio recordings are saved in the cloud and processed for e.g., marketing purposes. However, only a minority of users adapts their user behavior such as self-disclosure according to their privacy concerns. Research has attempted to find answers for this paradoxical outcome through concepts such as the privacy calculus or the privacy cynicism. Moreover, literature revealed that a large group of users lacks privacy awareness and privacy literacy preventing them from engaging in privacy-preserving behavior. Since previous studies in the scope of IVAs focused primarily on interviews or cross-sectional studies testing models predicting user behavior, desiderata for elevating future research are presented. This leads to a more user-centric approach incorporating e.g., a motivational-affective perspective and investigation of causalities.