Interpretable AI-enabled decision support for drinking-straw substitution using per-use greenhouse-gas indicators and user-review evidence.

Plastic drinking straws are a visible single-use plastic product, yet selecting suitable substitutes remains challenging because literature-derived climate evidence and reported user experience are rarely evaluated together. This study develops and demonstrates an interpretable AI-enabled decision-support workflow that integrates literature-derived per-use greenhouse gas (GHG) indicators with review-derived user evidence extracted from online customer reviews using natural language processing (NLP). Drinking-straw alternatives were used as an information-rich case study. The integrated assessment combined a GHG-derived score, a user-experience feature score, and rating-based consumer approval within a transparent multi-criteria decision analysis (MCDA) under four predefined decision-priority scenarios. Among the five shortlisted materials and within the evaluated dataset, the selected per-use GHG assumptions, review-derived user evidence, normalization procedure, and scenario-specific weights resulted in Silicone achieving the highest integrated MCDA score across all four scenarios, whereas Paper ranked lowest. Silicone combined a low per-use GHG indicator with the highest user-experience feature score and high consumer approval. Paper had the highest per-use GHG indicator and a moderate user-experience feature score, while lexical analysis identified recurring functionality-related expressions in its reviews. A shallow decision tree identified a 0.081 kg CO₂e/use threshold separating Paper from the lower-per-use-GHG reusable alternatives within the evaluated decision matrix. The study is not a new process-based life-cycle assessment or comprehensive sustainability assessment. Instead, it demonstrates decision support limited to per-use GHG indicators and review-derived user evidence; broader sustainability dimensions were outside the scope. Future studies may adapt and evaluate the workflow for other product categories using product-appropriate environmental criteria and relevant user-derived evidence.