Master Thesis: A Personal Assistant for Healthy and Sustainable Food Recommendations
Published in AlmaLaurea, 2024
This thesis explores the development and evaluation of a conversational recommender system aimed at promoting healthier eating habits through personalized recipe recommendations.
Traditional recommender systems have revolutionized various domains such as e-commerce and entertainment, yet the domain of food recommendation, particularly in conversational interfaces, remains relatively underexplored. The unique challenges in this domain include the need to prioritize nutritional value over taste preferences and accommodate diverse cultural, dietary, and personal factors influencing food choices.
Our conversational recommender system engages users in multi-turn dialogue via Telegram and provides explanations on the nutritional and sustainability aspects of recommended recipes.
An experimental study involving 32 users demonstrates that the bot’s explanations positively influence users’ perception of recipe healthiness and sustainability, leading to reconsideration of dietary choices. Quantitative analysis reveals significant improvements in users’ perception accuracy post-explanation, indicating the persuasive potential of the system.
Moreover, qualitative feedback highlights user satisfaction with the recommendations and the clarity of explanations provided.
Overall, this research underscores the transformative potential of integrating healthy food recommender systems within chatbot interfaces as persuasive tools for promoting healthier dietary choices.