10th U6+ Consortium of African Universities International Conference

Selective Dissemination of Information (SDI) services in academic libraries face mounting pressure from the exponential growth of digital scholarly content and the limitations of conventional rule-based systems. In Kenya, these challenges are compounded by manual workflows, static user profiles, and constrained ICT infrastructure, creating a significant gap between researcher information needs and the capacity of existing SDI systems. This paper aims to propose an AI-driven, clustering-based information mining model to enhance SDI services in Kenyan academic libraries. The model integrates K-means clustering, Natural Language Processing (NLP), TF-IDF vectorization, and dynamic user profiling to deliver personalized, timely scholarly content to library users. Grounded in Information Retrieval Theory and Recommender Systems Theory, the study employs a systematic literature review of peer-reviewed articles published between 2013 and 2024, sourced from databases including Scopus, Web of Science, IEEE Xplore, and Google Scholar, using search terms such as “selective dissemination of information,” “AI-driven recommender systems,” “K-means clustering library services,” and “dynamic user profiling.” A design-oriented methodology guided the construction of a five-layer architectural framework, comprising data ingestion, NLP pre-processing, clustering, dynamic profiling, and content delivery modules, validated through expert review against established information retrieval benchmarks. Evidence from existing literature demonstrates that AI-enabled clustering substantially reduces information overload while improving content relevance. Combining K-means clustering with dynamic user profiling ensures researchers receive targeted, discipline-specific scholarly updates in real time. Conceptual validation of the proposed framework against established precision and recall benchmarks from comparable AI-driven library systems indicates that the model is architecturally sound and operationally feasible within the Kenyan academic library context. Expert evaluation further affirmed the model’s modularity and scalability across institutions with varying ICT capacities. The study concludes that AI-driven SDI systems represent a viable and evidence-supported evolution for Kenyan academic libraries, transforming static, rule-based workflows into intelligent, adaptive, user-centred knowledge delivery platforms. The findings carry direct implications for policy and practice. Institutionally, library Managers should allocate budgets for AI infrastructure, pilot clustering based SDI systems, and establish metadata quality standards. Professionally, librarians need structured training in AI literacy and user profiling, which bodies such as the Kenya Library Association (KLA) should integrate into continuing professional development frameworks. At the policy level, regulatory bodies including the Commission for University Education (CUE) should develop ethical governance frameworks covering data privacy, algorithmic transparency, and equitable access. The ICT Authority of Kenya and technology vendors should support infrastructure upgrades,particularly internet connectivity and cloud computing access to sustain AI enabled SDI systems across diverse Kenyan universities. Funding agencies and university councils are further urged to support pilot implementations and longitudinal evaluations of these models.