Bringing positive experiences to users is one of the key goals when designing conversational agents (CAs). Yet we still lack an understanding of users’ underlying needs to achieve positive experiences and how to support them in design. This research first applies Self-Determination Theory in an interview study to explore how users’ needs of competence, autonomy and relatedness could be supported or undermined in CA experiences. Ten guidelines are then derived from the interview findings. The key findings demonstrate that: competence is affected by users’ knowledge of the CA capabilities and effectiveness of the conversation; autonomy is influenced by flexibility of the conversation, personalisation of the experiences, and control over user data; regarding relatedness, users still have concerns over integrating social features into CAs. The guidelines recommend how to inform users about the system capabilities, design effective and socially appropriate conversations, and support increased system intelligence, customisation, and data transparency.

While previous studies of Conversational Agents (e.g. Siri, Google Assistant, Alexa and Cortana) have focused on evaluating usability and exploring capabilities of these systems, little work has examined users' affective experiences. In this paper we present a survey study with 171 participants to examine CA users' affective experiences. Specifically, we present four major usage scenarios, users' affective responses in these scenarios, and the factors which influenced the affective responses. We found that users' overall experience was positive with interest being the most salient positive emotion. Affective responses differed depending on the scenarios. Both pragmatic and hedonic qualities influenced affect. The factors underlying pragmatic quality are: helpfulness, proactivity, fluidity, seamlessness and responsiveness. The factors underlying hedonic quality are: comfort in human-machine conversation, pride of using cutting-edge technology, fun during use, perception of having a human-like assistant, concern about privacy and fear of causing distraction.