Designing Conversational Agents for Search String Development in Literature Reviews
TL;DR
A conversational agent (CA) called STRINGI is developed that assists search string development through structured, guided, and pedagogically informed interactions, and bridges methodological prescriptions and operational support, enabling deployment while advancing research practices and offering potential for education.
Abstract
Developing rigorous search strings is essential for systematic literature reviews (SLRs), yet it remains challenging for young researchers and interdisciplinary teams who struggle with keyword identification, terminological heterogeneity, and database-specific syntax. While numerous tools support downstream SLR phases, search string development often lacks dedicated support. We address this gap by developing a conversational agent (CA) called STRINGI that assists search string development through structured, guided, and pedagogically informed interactions. Following a design science research approach, we conducted interviews to understand the problem space, aggregated design knowledge from multiple research streams, and instantiated 22 design features through systematic prompt engineering. We contribute a CA-based SLR support tool, a transparent design-features-to-prompt-transfer approach, and insights into the design of polyadic CA architectures. This offers potential for young researchers and students to both perform literature reviews as part of their work and develop their understanding of scientific principles and methods. The CA bridges methodological prescriptions and operational support, enabling deployment while advancing research practices and offering potential for education.