Conversational Agents Trust Calibration

Impact of Voice Fidelity on Decision Making: A Potential Dark Pattern?

“…2 While CAs are still predominantly used for simple tasks such as checking the weather, playing music, or setting alarms [1], a growing number of users is expecting to use them routinely for purchasing products and services online [50]. Consequently, the increase in transaction-oriented interactions provided by the present-day CAs opens up ways for implementing dark patterns that can potentially affect users' decisions towards their disadvantage [24]. While the potential of dark patterns to influence users' interactions with voice-based systems has recently been highlighted by Owens et al [57], this area of research still remains understudied.…”

CUI@CHI 2024: Building Trust in CUIs—From Design to Deployment

“…Regardless of the type of interaction-transactional or social-it necessitates the establishment of trust [36]. This trust is paramount because it not only ensures the accuracy and reliability of information and services but also influences how comfortable users feel when sharing personal data or forging deep connections with these AI-powered systems [19]. In this workshop, our main goal is to unite dedicated researchers from the CHI community for an in-depth exploration of CUIs from a trust perspective.…”

“…This lack of research is perhaps most evident in the opacity associated with the fundamental nature of trust, and specifically its conceptualization [24]. While the term 'trust' finds frequent usage within HCI [4] and the more specialized literature on CUIs [19,21,27,43], a clear and generally accepted definition within this context remains elusive. In CUIs, trust can manifest in diverse ways: (1) it may encompass both cognitive elements rooted in thinking and judgment and affective aspects rooted in feelings [42], (2) it often comprises attitudinal [25] and behavioral components [41], (3) it can be user-centric or agent-centric (including the owner of the agent) [42], and (4) its nature is inherently contextdependent [4,44].…”

“…[10,12,29]. Specifically, in critical domains such as finance and health, overreliance on voice-based advice may lead to misguided financial decisions [19] or, in the case of health, the inadvertent disclosure of sensitive medical information without a complete understanding of the associated privacy risks, including potential breaches of confidentiality and unauthorized access to personal health data [9]. These concerns may be exacerbated by demographic factors, such as age, where older individuals may be more vulnerable to trust-related issues due to their potentially limited familiarity with AI technologies [34].…”

Speech as Interactive Design Material (SIDM): How to design and evaluate task-tailored synthetic voices?

“…finances (e.g., takeaway-ordering or purchasing products online) [5]. However, despite constant technological developments in speech technology and commercial deployment, there is still a paucity of research examining the voice design and its implications for the user [10].…”