<p>This paper proposes a multitasking linear neural network model to analyze both their recycling behaviors and preferences simultaneously. It utilizes latent representations to share information across tasks for improvement in efficiency and accuracy. The model appropriately handles diverse data types commonly found in surveys, including categorical and continuous variables. It creates vector embeddings of consumer demographics and latent representations of individual consumers, enabling a deeper understanding of consumer demographics and their impact on recycling habits. The model's performance is evaluated against traditional methods like linear regression and non-linear neural networks, demonstrating its superior predictive capabilities and efficiency. Furthermore, the model's ability to generate interpretable insights and market segmentations based on demographic inputs and latent representations is showcased, offering valuable theoretical implications for marketing analytics and managerial implications for targeted recycling interventions.</p>