Leveraging data to shrink EPFL's food-related carbon footprint

MegaBites – as its name indicates – will entail crunching data on how ingredients enter the campus food chain, from the preparation of meals to consumption and disposal. The project team will evaluate such factors as the ingredients sourced by on-campus restaurants, the content of the dishes they serve, the types of meals chosen by different consumer categories and the amount of organic waste produced. Their goal? To identify strategies for reducing the School’s carbon footprint by modeling its food system. The project just received the green light from EPFL’s Human Research and Ethics Committee to collect this key information, which will be done without encroaching on anyone’s personal data. MegaBites researchers will accessthe age group, gender and status (i.e., student, professor, other EPFL employee or non-EPFL member) of people who frequent our on-campus restaurants.

“We’ll use these data to test incentives for encouraging consumers to make more sustainable food choices,” says Vincent Moreau, a project manager at EPFL’s Data Science Lab (dlab), which is headed by Prof. Robert West. The overall goal is to cut EPFL’s food-related carbon emissions by between a quarter and a third in the next two years. Today, these emissions make up 13% of the School’s total.

The project is being spearheaded by dlab in association with EPFL’s Catering and Shops Unit (RESCO), and has the backing of EPFL’s Solutions4Sustainability Initiative. It’s designed to achieve two important objectives. “The most pressing one is to reduce our School’s environmental impact, and especially the impact of our food consumption,” says Moreau. “But we also hope that the data we collect will give us fresh insight into the EPFL community’s eating habits.” MegaBites is a broad-based project that aims to simulate a comprehensive view of EPFL’s food system, from its suppliers and restaurant kitchens to the different consumer categories – and even what ends up in cafeterias’ trash cans.

Data will play a pivotal role in these efforts. In 2019, Bruno Rossignol, the head of RESCO, launched an initiative to take a detailed look at every aspect of EPFL’s food system: how dishes are prepared, the carbon emissions associated with the ingredients, the 1.5 million meals served annually, the restaurants’ energy use, what the kitchens and consumers throw away, the nutritional values and carbon scores of the dishes served, and more.

Random and unannounced trials

By providing specific data on the preferences of each EPFL consumer category, MegaBites will help design and implement targeted incentives and then measure their effect on purchasing habits and food-related carbon emissions. “The entire EPFL community will contribute to MegaBites through the choices students and staff make at campus cafeteria daily,” says Moreau. “We plan to start by analyzing data collected between July 2022 and the end of 2023.” The research will include determining which dishes are preferred by different consumer categories – broken down by age group, gender and EPFL status – without collecting any identifiable data on individuals. The data will be obtained through the Camipro system.

These data will enable the research team to measure the effectiveness of the incentives and select the most pertinent ones. The incentives could include, for example, giving certain types of dishes more appealing names, lowering the prices of vegetarian meals, or revamping the way food is presented on a plate. “We plan to carry out randomized control trials, unannounced, during which EPFL’s IT department will supply us with data on food sales,” says Moreau. The testing will run from this spring to this fall.

“The foolproof way to reduce EPFL’s carbon footprint would be to require that only vegetarian meals are served,” says Moreau. “But that’s not what the School wants to do, given the many different cultures and culinary tastes within its community. That makes our job harder – but also more interesting from a research perspective. And the methods we’re developing, along with the results of our incentive-testing and analyses, could be useful to other universities around the world.”