Future Food

Project description

Our project addresses the challenge of promoting sustainable food behavior through AI-powered communication. We developed a chatbot prototype that educates users about future food values, such as environmental impact, nutrition, and food waste reduction. The research focuses on how conversational AI, fine-tuned with domain-specific knowledge, can improve user engagement and awareness about sustainability. The chatbot uses a custom-trained language model that delivers personalized, context-aware responses while maintaining privacy and ethical design. Through this project, we aim to demonstrate that intelligent digital tools can bridge the gap between sustainability education and everyday consumer behavior.

Context

The project lies at the intersection of digital transformation and sustainability, specifically within the food and health domains. With food systems accounting for over 30% of global greenhouse gas emissions, raising awareness about sustainable eating is a critical step toward a greener future. However, traditional education methods often fail to reach younger audiences or engage them effectively.

Our solution is an AI chatbot that leverages conversational interfaces and natural language understanding to deliver tailored information on sustainable food choices. The chatbot is built using open-source LLM technology (Mistral-7B) and is fine-tuned on a curated dataset covering topics such as plant-based proteins, food waste, and the carbon footprint of food production. The system includes personalization, contextual memory, and privacy-by-design principles. It serves both as an educational tool and a proof-of-concept for integrating ethical AI in sustainability communication. The intended users are students, young professionals, and other tech-savvy individuals who want to learn about sustainable diets interactively.

Results

The most important outcome of the project is a fully functional prototype of a conversational AI chatbot that educates users on future food values in a personalized, ethical, and engaging way. Our chatbot is powered by a fine-tuned Mistral-7B open-source language model enhanced through LoRA adapters. It was trained on a domain-specific dataset of sustainable food-related Q&As and conversational strategies designed to foster behavior change.

Key insights include:

• Users responded positively to personalization and ethical design: features like context memory and anonymous access increased comfort and trust.
• The chatbot effectively explained complex topics (e.g., carbon footprint, alternative proteins) in simple language tailored to the user’s interests.
• The hybrid design (rule-based flow + LLM) enabled both reliable factual responses and dynamic, human-like conversation.

We validated the chatbot through structured testing and feedback from early users, including surveys and interaction logs. Participants reported improved understanding of sustainability topics and high satisfaction with the interface and tone.

From a technology readiness perspective (TRL), the project currently stands at TRL 5–6: a validated prototype tested in a relevant environment. It is capable of being deployed to real users and extended with additional features such as multilingual support or gamified interaction.

This work demonstrates the potential of fine-tuned conversational AI to address global challenges through interdisciplinary design. The flexible architecture and modular backend allow future expansion into other domains like climate education or recycling behavior, making it a scalable and impactful digital solution.

About the project group

Me and Andrii are software students and, since semester 4, AI-specialized students. We are working on this project as part of our internship in Fontys (lectoraat project) under the guidance of Priyanka Darbari.