Advancing research impact with data science and artificial intelligence

: an evaluation of contributing factors, indicators, and methods

Abstract

This PhD thesis advances the field of research-impact science by addressing key methodological, conceptual, and technological gaps in how research impact is defined, measured, and predicted. Existing frameworks and metrics have predominantly focused on retrospective, publication-based evaluations, limiting their capacity to capture the multidimensional and forward-looking nature of research outcomes. This thesis develops and validates a multidimensional, standardised, and predictive framework that integrates bibliometric analysis, data-science tools, empirical modelling, and artificial intelligence (AI). Its overarching aim is to transform research impact assessment from a retrospective to a predictive science that supports proactive decision-making. The study is guided by five interrelated research questions (RQs):
• RQ1. How has research-impact science evolved beyond traditional bibliometricmetrics, and what limitations persist in aligning impact assessment with societal and economic priorities?
• RQ2. How can data-science approaches—particularly bibliometric analysis, topic modelling, and machine learning—reveal dynamic patterns in research impact by identifying emerging themes, interdisciplinary linkages, and influential entities?
• RQ3. Which factors determine research impact, and how can the interrelationships among Research, Researchers, Institutions, Publications, and Countries be empirically validated within a system-level model?
• RQ4. What challenges hinder the development of a standardised taxonomy of research impact indicators, and how can a validated classification enhance comparability and adaptability across disciplines?
• RQ5. How can an AI-driven framework transform research-impact assessment from retrospective evaluation to predictive, decision-support analysis that guides proactivepolicy and resource allocation?
Each research question corresponds to a distinct phase of the thesis. Chapter 2 addresses RQ1 through a comprehensive scoping review of metrics, frameworks, and generational transitions in research-impact science. It identifies four historical generations—ranging from citationbased metrics and multidimensional frameworks to data-driven and alternative-metrics models—and argues for the emergence of a fifth generation, characterised by predictive, AIenabled assessment. This conceptual analysis establishes the theoretical foundation for the subsequent empirical and computational investigations.

Date of Thesis submission	2025
Original language	English
Awarding Institution	Western Sydney University
Supervisor	Omar Mubin (Supervisor) & Abdullah Al Mahmud (Supervisor)