The Carbon Footprint Wizard: A Knowledge-Augmented AI Interface for Streamlining Food Carbon Footprint Analysis

TL;DR

The paper tackles the challenge of making food carbon footprint analysis accessible by integrating lifecycle assessment data with retrieval-augmented AI to build a knowledge-augmented interface. It proposes a four-stage pipeline—ingredient processing, product matching, carbon footprint querying, and interactive exploration—implemented in a public demo. Key contributions include leveraging open LCA databases BONSAI, Agribalyse, and Big Climate Database within a chatbot framework, and employing human-in-the-loop product matching to improve accuracy. The work demonstrates the feasibility of translating complex LCA data into user-friendly, actionable insights while openly discussing data gaps and AI limitations. This approach has practical implications for researchers, policymakers, and consumers seeking transparent, iterative exploration of meal-level environmental impacts.

Abstract

Environmental sustainability, particularly in relation to climate change, is a key concern for consumers, producers, and policymakers. The carbon footprint, based on greenhouse gas emissions, is a standard metric for quantifying the contribution to climate change of activities and is often assessed using life cycle assessment (LCA). However, conducting LCA is complex due to opaque and global supply chains, as well as fragmented data. This paper presents a methodology that combines advances in LCA and publicly available databases with knowledge-augmented AI techniques, including retrieval-augmented generation, to estimate cradle-to-gate carbon footprints of food products. We introduce a chatbot interface that allows users to interactively explore the carbon impact of composite meals and relate the results to familiar activities. A live web demonstration showcases our proof-of-concept system with arbitrary food items and follow-up questions, highlighting both the potential and limitations - such as database uncertainties and AI misinterpretations - of delivering LCA insights in an accessible format.

Figures (6)

• Figure 1: Overview of our four-step methodology. Each step is represented by one color: ingredient processing in purple, product matching in orange, CF querying in black, and iterative exploration in green. The blue container shows the three databases. Arrows indicate data flow. The robot icons indicate LLMs. The person icons indicate input from humans.
• Figure 2: Recipe input interface with ingredient processing from natural language (shown on the left) to standardized quantities (organized as a two-column table on the right).
• Figure 3: Initial analysis showing ranked ingredients, impact range, and real-world equivalents as a natural language summary.
• Figure 4: The product selection showing multiple database matches for each ingredient, resulting from the semantic search in the product matching phase.
• Figure 5: Carbon footprint visualization through bar and pie charts showing ingredient impacts.