Intelligent Support for Sustainability Awareness in Online Shopping

Sustainability concerns increasingly shape consumer choices, yet online
product descriptions often lack clarity or key sustainability informa-
tion. This creates information gaps that limit consumers’ ability to
make informed, responsible decisions—contributing to the persistent
intention–behaviour gap in ethical consumption.
While prior efforts in sustainability communication have focused on
labelling and user interface design, little attention has been paid to the
textual content of product descriptions and how it aligns with expert
assessments. This thesis addresses that gap by using generative AI to
generate sustainability awareness messages tailored to product context.
The work introduces a multi-stage framework that combines expert
knowledge with product-level data, structured through a sustainability
taxonomy covering health, environment, society, and economy. This
taxonomy guides the extraction of sustainability cues from product
descriptions, forming a Text-Based Green Profile (TGP) that serves as
a proxy for what consumers see. The TGP is then compared to expert
annotations to identify patterns of alignment and misalignment—some
of which reflect known decision-making biases. These patterns inform
the design of prompts for large language models (LLMs), which generate
context-aware sustainability messages. Finally, the generated messages
are evaluated through both automatic metrics and human feedback
from expert reviewers and a user study.
Findings show that the AI-generated messages were generally seen as
clear, relevant, and helpful. Participants reported improved awareness
of sustainability issues, with effectiveness varying by alignment pattern
and framing, highlighting the need for context-sensitive messaging. This thesis contributes a practical, scalable approach for generating
sustainability-focused messages to help close information gaps in online
shopping. It offers insights into how AI can support more transparent,
adaptive, and responsible communication in digital consumer environ-
ments.