Evaluating E-Commerce Sustainability: A Probabilistic–Behavioral Framework Using Bayesian Networks and Structural Equation Modeling
摘要
The rapid growth of e-commerce in developing economies has heightened concerns about logistics-related emissions, packaging waste, and consumer behavior impacting the environment. However, current sustainability evaluations tend to focus separately on operational and behavioral factors. This paper proposes a combined probabilistic-behavioral decision support model to assess the environmental sustainability of e-commerce activities in Bangladesh, utilizing a mix of Bayesian Belief Networks (BBN) and Partial Least Squares Structural Equation Modeling (PLS-SEM). The BBN component operates under uncertainty to model key operational subsystems — logistics, packaging, and waste management — as primary drivers of environmental impact, and evaluates scenario-based interventions through structured expert elicitation. Meanwhile, the SEM component explores correlations between environmental awareness, behavioral attitudes, and sustainability-oriented consumption tendencies based on consumer survey data. These models are integrated at the interpretation level, where behavioral insights help prioritize operational sustainability scenarios and leverage points. The findings indicate that last-mile delivery distance, vehicle type, and packaging material intensity are the most influential factors on environmental impact. Additionally, consumer environmental attitudes are moderately aligned with sustainable behavior. Sensitivity analysis can identify the most promising areas for operational and behavioral interventions to reduce emissions and waste. The proposed framework offers a clear, adaptable method for assessing sustainability in developing countries with limited data-providing practical insights for practitioners and policymakers — while acknowledging the exploratory and case-specific nature of the analysis.