DiscipLink: Unfolding Interdisciplinary Information Seeking Process via Human-AI Co-Exploration

TL;DR

DiscipLink tackles interdisciplinary information seeking by combining human agency with large language models in a co-exploration workflow. It generates exploratory questions across disciplines, expands queries with discipline-specific terminology, and organizes retrieved papers around exploration themes to highlight connections. Contextualized clustering and information-scent annotations help users quickly screen and synthesize literature, while drag-and-drop collections support mixed-initiative workflows. Across two studies, DiscipLink improves outline quality, learning gains, and exploration efficiency, though users note EQ volume and integration challenges. The work demonstrates the potential of human–LLM collaboration to expand interdisciplinary inquiry and informs future improvements for nonlinear IIS support.

Abstract

Interdisciplinary studies often require researchers to explore literature in diverse branches of knowledge. Yet, navigating through the highly scattered knowledge from unfamiliar disciplines poses a significant challenge. In this paper, we introduce DiscipLink, a novel interactive system that facilitates collaboration between researchers and large language models (LLMs) in interdisciplinary information seeking (IIS). Based on users' topics of interest, DiscipLink initiates exploratory questions from the perspectives of possible relevant fields of study, and users can further tailor these questions. DiscipLink then supports users in searching and screening papers under selected questions by automatically expanding queries with disciplinary-specific terminologies, extracting themes from retrieved papers, and highlighting the connections between papers and questions. Our evaluation, comprising a within-subject comparative experiment and an open-ended exploratory study, reveals that DiscipLink can effectively support researchers in breaking down disciplinary boundaries and integrating scattered knowledge in diverse fields. The findings underscore the potential of LLM-powered tools in fostering information-seeking practices and bolstering interdisciplinary research.

Figures (8)

• Figure 1: The nonlinear model of IIS behavior includes three components foster2004nonlinear. Researchers do not go through these components with a certain order, but intertwine the activities in these components during IIS.
• Figure 2: The workflow of DiscipLink. We harness LLMs to support the three stages of the IIS process.
• Figure 3: The user interface of DiscipLink, including Orientation View for creating and tailoring EQs, Exploration View for explore retrieved papers for each EQ, and Collection View for organizing papers users found useful.
• Figure 4: DiscipLink provides diverse EQ suggestions to discover highly-scattered relevant information based on two types of seeds given by users: research topic of interest and paper of interest. We prompt LLMs to simulate as experts in specific academic fields to provide EQs. Users can also create EQs by themselves.
• Figure 5: DiscipLink generates a diverse set of queries to retrieve relevant papers for an EQ.