Computational Approaches for Studying Dietary Behaviors with Digital Traces

Human nutrition and dietary habits shape our health, daily life, societies, the environment, and life on earth in general. However, it remains challenging to understand and attempt to change dietary behaviors using traditional methods due to measurement and causal identification challenges.

In this thesis, we contribute computational and causal approaches leveraging large-scale passively sensed digital traces to shed new light on our dietary behaviors and derive novel scientific insights. We study dietary behaviors in two types of contexts: campus-wide and worldwide. Based on digital traces capturing behaviors of tens of thousands of people on campus and millions of internet users, we develop observational study designs that enable the isolation of causal effects of implicit behavioral interventions, including forming social ties, being exposed to the food choice of others, and stay-at-home interventions. The thesis is organized into three parts.

The first part presents studies based on situated on-campus food purchase logs. In the first study, we show how, when a person acquires a new eating partner on campus, the healthiness of their food choice shifts significantly in the direction of their new eating partner's dietary patterns. In the second study, we identify purchasing mimicry, i.e., copying the food choices of others in the purchasing queue, as a specific behavioral mechanism for how similarities in dietary behaviors between individuals occur on campus.

The second part of the thesis leverages online information-seeking traces (Google search query logs). Studying worldwide dietary behaviors, we identify and describe global shifts in dietary interests during the first wave of the COVID-19 pandemic, larger and longer-lasting than the typical changes during the end-of-year holidays in Western countries.

In the third part, we critically investigate the limits to how much computational approaches can reveal about dietary behaviors in the general population. We contribute a framework for reasoning about biases of digital traces and present a case study of food consumption in Switzerland. The bias estimates derived in the case study imply that researchers should aim to establish evidence of validity before relying on social media and tracking apps---two frequently used digital traces---as proxies for true food consumption in the general population.

The novel scientific findings and methodological advances presented here contribute to the existing knowledge about human dietary behaviors and inform the design of future food systems, policies, and behavioral interventions.