Users often struggle to make rational privacy decisions within the notice and choice framework, primarily due to difficulties in understanding and processing privacy policies. Recent studies suggest that large language model (LLM)-based chatbots can improve privacy policy comprehension and privacy awareness. Meanwhile, the specific expectations and needs that users bring to LLM-based chatbots for rational privacy decision-making, and whether those needs are met, have been underexplored. Employing a technology probe and focus groups, we investigate the roles users expect from the chatbots and the needs that arise during interaction. We further interview three experts to corroborate our findings. Our study reveals a typology of three user-expected roles for LLM-based chatbots—Interpreter, Guardian, and Evaluator—along with satisfied and unmet needs within these roles. Finally, we outline implications that clarify where LLMs are—and are not—viable in privacy decision-making, while highlighting structural limitations of notice and choice.

With the advent of neural networks, it has become possible to generate synthetic voices that are nearly indistinguishable from real human speech (i.e., human-sounding voice).In contrast, earlier voice assistants used voices that were instantly recognizable as machine-generated, owing to their standardized, consistent, and highly intelligible qualities (i.e., artificial-sounding voice).Although people tend to prefer human-like voices, adopting human-sounding voices in voice assistants raises ethical concerns related to confusion or unintentional deception, particularly in voice-only contexts, even when their identity as systems is explicitly disclosed.To explore the voice design direction for future voice assistants, we examined how participants perceived and interacted when they were unexpectedly confronted with either an artificial-sounding or a human-sounding voice, both of which clearly identified themselves as voice assistants during an everyday phone call.Our findings reveal participants' experiences and conversational behaviors in each voice condition.Furthermore, we discuss how the voices of voice assistants should be designed and propose design implications that emphasize transparency and responsiveness in voice design.

Multilingual speakers tend to interleave two or more languages when communicating. This communication strategy is called code-mixing, and it has surged with today’s ever-increasing linguistic and cultural diversity. Because of their communication style, multilinguals who use conversational agents have specific needs and expectations which are currently not being met by conversational systems. While research has been undertaken on code-mixing conversational systems, previous works have rarely focused on the code-mixing users themselves to discover their genuine needs. This work furthers our understanding of the challenges faced by code-mixing users in conversational agent interaction, unveils the key factors that users consider in code-mixing scenarios, and explores expectations that users have for future conversational agents capable of code-mixing. This study discusses the design implications of our findings and provides a guide on how to alleviate the challenges faced by multilingual users and how to improve the conversational agent user experience for multilingual users.