Deploying Conversational Agents in Virtual Research Environments: Approaches and Lessons Learned

Abstract

Conversational agents have the potential to streamline tasks, provide support, and enhance user experience across various domains including Virtual Research Environments (VREs). The recent progress in conversational artificial intelligence and Large Language Models (LLMs) offers novel strategies for the development of these agents. This paper reports on the potential benefits, the challenges and the approaches resulting from concrete experiences in developing and equipping D4Science-based VREs with suitable conversational agents. The paper presents three successive implementation approaches and the resulting agent solution, each designed to address the limitations identified in the preceding iteration and to leverage the advantages offered by newer implementation and development options. The proposed approaches led to the progressive refinement of the agent design and functionality, resulting in DAVE, a conversational agent capable of securely interacting with multiple D4Science services and supporting a wide range of user workflows. The iterative process highlighted critical requirements—including authentication handling, usability, and extensibility—that can inform the design of conversational agents in similar research infrastructures. The study shows that conversational agents can effectively lower the barrier to accessing VRE functionalities and enhance user engagement. The resulting design principles and lessons learned provide a foundation for future work aimed at extending DAVE with an enhanced feedback mechanism and locally hosted LLM integration, and conducting systematic usability evaluations within active research communities.

Ähnliche Arbeiten