Reflective Language at the Point of Purchase: Mindful Conversational Agents for Sustainable E-Commerce

Abstract

Consumers often want lower-impact options but lack timely, personalized guidance at the moment of purchase. We study a conversational eco-agent embedded in checkout that provides individualized impact summaries and, when appropriate, a single comparable lower-impact substitute; a variant adds one or two brief reflection prompts. In a four-week randomized field experiment in a mobile-first Indian checkout, the eco-agent reduced basket-level carbon dioxide equivalent (CO₂e) by 0.62 kg relative to the control, and the reflection variant reduced CO₂e by 1.05 kg. Lower-impact substitution events increased from 3.6% in control to 7.9% with the eco-agent and 11.1% with reflection, while time-to-checkout rose modestly, and post-purchase satisfaction and Net Promoter Score were non-inferior. A preregistered text-based mediator showed that reflective language accounted for 26% of the eco-agent effect and 42% of the incremental reflection effect. The article contributes a replicable, ethically transparent checkout intervention and provides mechanism-level causal evidence linking reflective language to sustainable substitution in a live retail setting.