• Consumers can be encouraged to choose more sustainable delivery options. • Extra surcharges work best to get consumers to choose greener delivery. • Messages about the environment help a little, but not as much as surcharges. • Consumers feel better about extra surcharges when they think they’re fair. Logistic providers sometimes nudge consumers to participate in the delivery process and stimulate them to choose more sustainable delivery options to reduce the environmental impacts of last-mile delivery. Nevertheless, online retailers are wary of such interventions, fearing that this will reduce consumer satisfaction and jeopardize their competitive position vis-à-vis other online retailers and brick-and-mortar shops. Based on fairness theory, we developed and tested interventions to nudge consumers towards more sustainable delivery options while maintaining consumer satisfaction. A vignette-based experiment following a 2 (surcharge) × 3 (message) design was used to test the impact of surcharges in combination with a message grounded in fairness theory. Using logistic regression, we found that surcharges were the most effective in increasing the probability that the (free) more sustainable delivery option would be selected. A message focusing on the sustainability implications was also effective, but to a lesser extent. The combination of surcharges and messaging was found to decrease this probability. Using causal mediation analysis, perceived fairness was found to mediate the effect of the surcharge on satisfaction. The study findings support the notion that it is possible to design interventions that will stimulate consumers to select more sustainable delivery options without significantly affecting their satisfaction. Interventions based on price (dis)incentives remain the most effective when influencing consumer behavior in the last mile. Their negative effects need to be mitigated through appropriate communication, however.

Pedestrians, especially wheelchair users, face significant navigation challenges due to inadequate infrastructure. Addressing these challenges requires pathfinding services to offer optimal routes tailored to user needs. Such routes depend on detailed infrastructure data, often absent from crowdsourced maps. This study enhances OpenStreetMap (OSM) data and presents a multi-criteria, context-aware routing approach. It integrates Dijkstra’s algorithm, a Decision Tree (DT), and Ordered Weighted Averaging (OWA) for personalized navigation. The proposed framework makes two key contributions: (1) a Large Language Model (LLM)-driven OSM enrichment pipeline and (2) a custom OWA-weighted DT algorithm that adapts routes according to user preferences. Experiments on 190 routes in Calgary show the enhanced network reduces total walking distance by 25.89 km (12.45%) and travel time by 47.78 min (9.83%) compared to the original OSM-based network. The customized pathfinding method enables preference-aware routing based on criteria such as surface quality, slope, crossing safety, and lighting. Results show improvements in safety-focused profiles, with a 2.2% increase in safety-related attributes. Accessibility-focused profiles see a 5.3% increase in surface quality. Personalized routes remain efficient, averaging only 1.8% longer than the shortest available path. In 97.4% of cases, the distance stays within a 10% margin. This supports real-time routing with low computational overhead. Limitations include dependence on attribute completeness. Future research will incorporate real-time environmental conditions, expand preference-learning mechanisms, and evaluate generalizability across a wider range of urban contexts, supporting real-time routing with minimal computational overhead. • Enhances OpenStreetMap data using a multi-source enrichment pipeline. • Combines human expertise with Large Language Models to improve OSM geospatial data quality. • Cold-start routing framework combining rule-based paths, OWA, and a DT model. • Uses explicit and implicit preferences to create context-aware pedestrian routes. • Walking speed model that adapts to infrastructure and user demographic factors. • Tailored models and enriched data yield routes that better reflect user preferences.