Design and Evaluation of an AI-Based Recipe Generator for Reducing Food Waste Among University Students in Finland

Abstract

The objective of this thesis was to design and evaluate an AI-based recipe generator aimed at reducing household food waste among university students in Finland. Household food waste in Finland is influenced by factors such as limited cooking skills and inefficient meal planning. This highlights the need for practical, accessible tools tailored to the Finnish context, particularly for individuals with limited cooking experience. The study examined how artificial intelligence could be utilized to generate practical cooking recipes based on ingredients available to the user, supporting more efficient food usage and minimizing unnecessary waste.

The study was conducted by developing a prototype system based on a pretrained language model. The system was designed to accept user-provided ingredient inputs and generate corresponding cooking recipes. In order to improve the quality, relevance, and usability of the generated outputs, prompt engineering techniques were applied. In addition, post-processing methods were implemented to enhance the structure and readability of the generated recipes. The prototype was tested using a variety of ingredient combinations to assess its general functionality.

An informal, small-scale inquiry was conducted with university students to explore their food consumption habits, including purchasing behaviour, cooking practices, and causes of food waste. The findings were used to guide the system design, ensuring that the proposed solution aligns with real-world user needs and challenges.

The results showed that the system was able to generate creative and generally useful recipe suggestions based on limited input ingredients. However, inconsistencies in structure and variation in output quality were observed. Despite these limitations, the findings suggest that AI-based recipe generation can serve as a promising tool for encouraging efficient ingredient usage and reducing food waste at the household level.

In conclusion, the study demonstrated the potential of artificial intelligence in supporting sustainable food practices. Future work could focus on improving output consistency through more advanced models and conducting larger-scale user evaluations to better assess practical impact.