비형식 평생학습 참여 예측을 위한 설명가능한 AI 의사결정: 디지털․고령사회에서의 참여자 예측과 포용적 교육정책
Explainable AI Decision Models for Non-formal Lifelong Learning Participation: Predictive Insights and Inclusive Policy in a Digital Aging World
조은지(인천대학교 경영대학 경영학부); 이현서(인천대학교 글로벌정경대학 GTS학부); 유효정(인천대학교 글로벌정경대학 GTS학부); 김경원(인천대학교 글로벌정경대학 GTS학부)
51권 1호, 31~50쪽
초록
As society navigates rapid aging and digital transformation, understanding the complex, interactive factors driving adult participation in non-degree, non-formal lifelong learning is critical for designing inclusive educational policies. This study leverages machine learning, deep learning, and Explainable AI (XAI) to predict non-formal learning participation and interpret its core determinants. Utilizing national survey data (2018–2022), we quantified these factors across demographic and time-series dimensions, highlighting non-formal education as a crucial driver of social engagement and quality of life for older adults. Our findings indicate that program accessibility, workplace size, informal physical activity, and life satisfaction significantly boost participation rates, whereas socioeconomic vulnerability and outdated learning media act as primary barriers. Furthermore, the post-pandemic era reveals a strengthened reliance on digital learning platforms and a crucial motivational shift among older learners—moving from economic-driven goals to the pursuit of health and cultural enrichment. This study expands prior research by quantitatively elucidating the interplay of structural, behavioral, social, and psychological factors. We propose a four-pronged policy framework: (1) expanding structural safety nets for vulnerable groups and SME workers; (2) enhancing generation-specific digital literacy and learner autonomy; (3) building community-linked learning models; and (4) designing wellbeing-centric programs. Algorithmic evaluations confirm that CatBoost delivers the most stable and generalized predictive performance, proving that interpreting key determinants enables accurate future forecasting. By directly linking model interpretability with predictive precision, this study establishes a robust, data-driven methodological framework for policy design. Ultimately, it offers empirical evidence for bridging educational divides and tailoring lifelong learning ecosystems, achieving both theoretical innovation and practical policy impact.
Abstract
As society navigates rapid aging and digital transformation, understanding the complex, interactive factors driving adult participation in non-degree, non-formal lifelong learning is critical for designing inclusive educational policies. This study leverages machine learning, deep learning, and Explainable AI (XAI) to predict non-formal learning participation and interpret its core determinants. Utilizing national survey data (2018–2022), we quantified these factors across demographic and time-series dimensions, highlighting non-formal education as a crucial driver of social engagement and quality of life for older adults. Our findings indicate that program accessibility, workplace size, informal physical activity, and life satisfaction significantly boost participation rates, whereas socioeconomic vulnerability and outdated learning media act as primary barriers. Furthermore, the post-pandemic era reveals a strengthened reliance on digital learning platforms and a crucial motivational shift among older learners—moving from economic-driven goals to the pursuit of health and cultural enrichment. This study expands prior research by quantitatively elucidating the interplay of structural, behavioral, social, and psychological factors. We propose a four-pronged policy framework: (1) expanding structural safety nets for vulnerable groups and SME workers; (2) enhancing generation-specific digital literacy and learner autonomy; (3) building community-linked learning models; and (4) designing wellbeing-centric programs. Algorithmic evaluations confirm that CatBoost delivers the most stable and generalized predictive performance, proving that interpreting key determinants enables accurate future forecasting. By directly linking model interpretability with predictive precision, this study establishes a robust, data-driven methodological framework for policy design. Ultimately, it offers empirical evidence for bridging educational divides and tailoring lifelong learning ecosystems, achieving both theoretical innovation and practical policy impact.
- 발행기관:
- 한국경영과학회
- 분류:
- 경영학