'AI' Literature Search Tools and Scholarly Diversity Project

Two datasets used for the 'AI' Literature Search Tools and
Scholarly Diversity Project. This study examines whether 'AI'
literature search tools amplify or lessen any biases compared to
conventional literature search platforms. Adopting algorithmic
ethnography, we analysed 800 search results consisting of the top 20
journal articles in four topics in higher education across 10 literature
search platforms: five AI-driven tools and five ‘non-AI’ traditional
databases. Metrics included first-author gender, geographic affiliation,
numbers of citations and journal impact factors. We found that the search
results from 'AI' platforms exhibited significantly greater
geographic diversity but showed no consistent advantage in terms of gender
balance. Both 'AI' and ‘non-AI’ tools prioritised high-citation
numbers and high-JIF publications. This work contributes to the literature
that looks beyond perceptions of 'AI' literature search tools
and examines their outputs. The insights gained will be useful for
everyone making decisions regarding which platforms to use when conducting
searches of academic literature and for those advising them. Description