Algorithmic sensemaking: Developing an LLM-based approach to digital government literature reviews

Abstract

Research on e-government and digital transformation is widely seen as conceptually fragmented. Existing literature reviews address this through either systematic literature reviews (SLRs), which offer depth but risk selection bias, or bibliometric methods, which scale well but yield generic, low-resolution findings. This poster proposes a quali-quantitative alternative, using the Specter language model to generate text embeddings for 34,023 Scopus abstracts on public sector digitalization (1990–2024), then visualizing them in a two-dimensional space to enable a close reading at scale through an ”algorithmic sensemaking” approach that bridges between content analysis and bibliometric methods. Using this approach, we identify three thematic regions, each with distinct topics and sub-topics: (1) core digital government discussions including applications (e-health, e-participation), (2) technical infrastructure (interoperability, cloud computing, blockchain), and (3) digital economy/business transformation (smart cities, organizational change). Tracing these topics over time shows the field shifting from narrow e-government/infrastructure concerns toward broader societal transformation.