GenAI: A Shifter of Information Retrieval Paradigms

Since the emergence of Information Retrieval systems in the late 1950s, achieving optimal search results has been a critical goal for librarians and information specialists. Traditional IR methods rely heavily on the balance between 'recall'—retrieving all relevant documents—and 'precision'—excluding irrelevant ones. However, these metrics present limitations, particularly due to their binary view of relevance and the inherent trade-off between recall and precision.

Historically, IR has employed Boolean logic for query construction, a method that, while powerful in some respects, often results in high recall at the expense of precision, requiring significant manual effort to filter results.

The advent of Generative AI (GenAI) marks a paradigm shift in IR, with the potential to address many of these challenges. Early implementations of GenAI-driven search tools, such as Scopus AI, have demonstrated remarkable improvements in precision, driven by natural language processing rather than complex Boolean queries. By allowing users to input research questions as prompts, GenAI can yield more relevant results, simplifying the search process and reducing the need for manual filtering.

However, GenAI is not without its limitations. The potential for missing relevant documents remains a concern, particularly in tasks requiring exhaustive retrieval, such as systematic or scoping reviews. While GenAI represents a promising evolution in IR, it should be viewed as complementary to traditional methods rather than a full replacement.

This presentation will explore the transformative role of GenAI in IR, its current capabilities, and the ongoing need for further refinement to ensure it can reliably meet the rigorous demands of academic and professional research. We will draw upon experience and insights gained from a series of workshops on using GenAI for IR that we have done with librarians from Denmark, Norway, South Africa and Scotland.