The feedback loop: A systematic review of how evaluation practices inform conversational agent design

Conversational Agents (CAs) have become increasingly integrated in our daily interactions. Despite this growing adoption, research on CA design remains fragmented, often lacking clear connections between design and its effects on user experience. Bridging perspectives from Human-Computer Interaction (HCI) and Information Systems (IS), this study provides a structured, design element-level synthesis of CA design practices and their effect on user experience. To do this, we systematically reviewed 97 empirical studies published in leading HCI and IS venues, following a rigorous three-stage methodology. Building on this, we proposed a CA design framework that identifies design dimensions and maps them with user outcomes to provide actionable guidance for practitioners. The resulting structured framework offers conceptual clarity and practical guidance, laying the groundwork for future empirical research and evidence-based design strategies in CA development.