Governing the Algorithm: A Mediation Analysis of Digital Transformation, Bureaucratic Discretion, and Service Quality in a Developing Democracy

Abstract

Governments worldwide are implementing digital transformation policies to enhance public service delivery. However, the impact of these algorithm-driven systems on street-level bureaucrats remains critically under-examined. This study investigates the complex pathways through which Indonesia's e-government policy affects bureaucratic work and service outcomes. This study employed a mixed-methods explanatory sequential design. Quantitative data were collected from 500 public officials across five Indonesian provinces. An E-Government Implementation Index (EGII) was constructed. We used Ordinary Least Squares (OLS) regression and a formal mediation analysis with bootstrapping to analyze the relationships between EGII, Perceived Bureaucratic Discretion (PBD), and Bureaucrat-Perceived Public Service Quality (B-PSQ). This was supplemented by 20 in-depth, semi-structured interviews to explain the statistical findings. Regression analysis confirmed a significant negative association between EGII and PBD (β = -0.47, p < 0.001) and a significant positive association between EGII and B-PSQ (β = 0.62, p < 0.001). The mediation analysis revealed that EGII has a strong, positive direct effect on B-PSQ (Effect = 0.57, p < 0.001) and a small but significant negative indirect effect through the reduction of PBD (Effect = -0.05, p < 0.01). Qualitative data revealed that officials feel constrained by "algorithmic cages" that, while improving efficiency, limit their ability to handle exceptional cases, thereby risking service equity for marginalized citizens. In conclusion, Indonesia’s digital transformation presents a complex trade-off. It successfully enhances administrative efficiency but simultaneously curtails the beneficial discretion of frontline bureaucrats, creating a small but significant drag on service quality. Effective digital governance requires a hybrid model that embeds algorithmic systems within a framework that empowers, rather than replaces, human judgment.