Silicus Traveler: An agent to simulate tourist behavior

Abstract

We introduce the Silicus Traveler, an innovative framework for a virtual tourist agent that simulates travel decision-making and behavior using large language models. Fidelity in these models hinges on high-quality, diverse, and representative data. Persistent challenges remain: cultural inclusivity, bias, and the inherent limitations of artificial systems. To address these, we propose conditioning the agent on Episode Records datasets (activities, situational context, and affect). We position the framework as a complement to human-centered research and as a reproducible pipeline that expands the methodological toolkit of tourism studies.