Using Network Science to Identify Trends in Food History. Sweeteners and Animal Proteins in Developing Countries: Colombia, 1977-2017.
Juan Carlos Sanchez HerreraFood consumption and dietary changes are shaped by larger historical, cultural, economic and political transformations in a particular geography. Dietary changes have been documented in different geographical contexts in particular as countries develop which in the past decades has led to negative consequences as non-communicable diseases and obesity rates rise, posing startling challenges to human and environmental health. Obtaining historical information on food consumption that reflects dietary changes presents challenges regarding biases, over accounting, and a lack of understanding on how ingredient combinations work, and which ingredients may substitute or complement each other. Food recipes have been used in different contexts and academic disciplines to overcome this challenge and approximate food consumption.
Food is one of the most interconnected systems ever created by humankind. Each dish we consume usually contains more than one ingredient and is brought to us through an intricate web of links connecting the ingredients on the dish to the places where they come from. Similarly, the dish itself —which can be recorded as a recipe¬— is a complex system in itself as ingredients are transformed through preparation techniques to accomplish the dish. The combination of ingredients in the form of a dish ensures that we obtain a varied diet necessary for our survival which if it is not balanced may lead to negative consequences. Hence, recipes are a source of information that allows a researcher to disentangle the complexity of food. However, studying large collections of recipes is a challenging task as the number of recipes that can be covered using qualitative methods is limited. Furthermore, recipes contain large amounts of information such as servings, cooking time, ingredient lists, images, directions, preparation techniques, etc., making them hard to approximate. The consequence is that several studies on food history fall short at analyzing cookbooks and limit their scope to the preface while leaving recipes untouched.
The purpose of this presentation is to demonstrate how network science can be used to identify changes in food consumption by using food recipes with a particular emphasis on developing countries which face challenges in dietary. This presentation focuses on identifying changes in two particular categories of food which are linked to non-communicable disease and obesity: sweeteners and animal proteins. The methods used are network science metrics such as eigenvector centrality and PageRank and their growth rate over time along with network embeddings to identify the periods in which these changes occurred to guide a deeper qualitative analysis into the rationale and sociocultural trends that explain the changes observed using network metrics. The primary source of information consists of 5,981 recipes encompassing 557 unique ingredients collected from Revista Carrusel, a magazine which has consistently published recipes since its launch in 1977 until today. The results present a mixed methods approach that uses network science to guide a deeper qualitative investigation of the primary sources. The scope of this work can illuminate the study of other primary sources beyond recipes.