From Inquiry to Interaction: Evaluating Interaction Styles of AI-Based Conversational Agents

Recent advances in artificial intelligence (AI) have enhanced conversational agents (CAs), enabling more sophisticated and agentic interactions beyond simple question-answer formats. Emerging nonlinear and non-deterministic interaction designs influence user experience, yet their effects on user satisfaction remain unclear. This study examines how different CA interaction designs affect user satisfaction and perceived information gain. We conducted an online experiment simulating a vacation planning process supported by a CA. Five CA designs, each based on a large language model and differing in interaction style, were developed to test our hypotheses. Results show that users report the highest satisfaction and perceived information gain when CAs provide swift, clear answers while preserving user control throughout the conversation. The findings contribute to theory by clarifying how interaction design shapes user outcomes and offer practical guidelines for developing responsive, user-centered conversational agents.