Exploring the potential of machine learning to reduce administrative burden in participatory budgeting: a case study of Seoul

Purpose This study explores the potential of machine learning techniques to reduce administrative burdens in participatory budgeting, focusing on the case of Seoul, which received over 25,000 proposals between 2013 and 2021. Design/methodology/approach Similar to a coffee dripper, the budgeting process filters and refines numerous citizen inputs within budgetary constraints, requiring large-scale public deliberation to review, synthesise and prioritise these inputs. This process typically involves labour-intensive processing by both citizens and public officials, generating significant administrative burdens. This paper argues that machine learning techniques can automate redundant preparatory tasks, thereby enabling institutional resources to focus more on deliberation and decision-making. By combining unsupervised and supervised learning approaches, this study identifies 17 major topics within the proposal corpus and develops a proposal selection classifier that achieves an Area Under the Receiver Operating Characteristic score of 0.73, indicating reasonable predictive performance. Findings The findings demonstrate the potential of machine learning to alleviate learning costs through automated proposal categorisation and summarisation as well as compliance costs by providing predictive feedback that supports citizens in developing more robust proposals. Originality/value This exploratory study contributes to the emerging discourse on digital administrative burden, offering practical insights for mitigating administrative challenges in large-scale participatory budgeting.