The fight against global warming is one of the most important challenges of this century. Regular reports, written by the Intergovernmental Panel on Climate Change (IPCC), have highlighted the effects of global warming and the need to reduce CO2 emissions to preserve the planet for future generations. These findings require that we modify the way we use energy and that we implement a transition plan to reach carbon neutrality by 2050.
However, modifying our behavior is not easy. We can indeed have different levels of sustainability per consumption domains. The level of sustainability in this context is linked to the level of pollution related to an act of purchase or to the use of a good or a service. For instance, an individual who buys local products in bulk in a store close to her home but uses her car to go to work every day, will not have the same sustainability levels for the food (high) and the mobility (low) consumption domains. Thus, knowing the levels of sustainability of an individual based on her different consumption domains is crucial to set up adequate interventions to help individuals and to guide public policy.
Many online tools enable to measure the carbon footprint of an individual. However, these tools usually only provide a global result enabling us to compare the average energy consumption of an individual to other individuals living in the same region or country. In our research work, we addressed the problem of predicting the level of sustainability of several consumption domains according to specific characteristics (e.g., demographic, psychological, etc.) of an individual. This prediction would therefore make it possible to automatically detect the profile of an individual in order to propose personalized positive incentives that could help her adopt a more sustainable lifestyle.
We used data that is related to the energy consumption of 5000 representative households of the Swiss population, acquired by the Competence Center for Research on Energy, Society and Transition (SCCER – CREST). More precisely, this center collected data related to three domains of consumption: heating, electricity and mobility, as well as demographic, psychological and social data through a series of questionnaires over several years.
On the one hand, we explored whether it was possible to predict the levels of sustainability of different consumption domains (e.g., the mode of transport used by an individual between home and work) on the basis of data that could not necessarily have a direct link with the individual’s consumption behavior (e.g., demographic and/or psychological data). To this end, different prediction techniques used in machine learning were assessed and compared. We were then able to show that it was possible to predict the sustainability levels of consumption domains, but also to highlight the key characteristics that could have the most influence on them, in particular the impact of the geographical context of the individual on the prediction of the sustainability level of the mode of transport used. This first analysis also confirms the relevance of the use of ensemble approaches in terms of prediction accuracy. Indeed, these approaches aim at building and using several predictive models versus using a single model.
On the other hand, we developed a framework to predict the sustainability levels of an individual by consumption domains and that includes different granularities in order to obtain her global profile in the form of a tree structure. The goal of this framework is to generate sustainability profiles of individuals in terms of consumption based on their own data (e.g., demographic and/or psychological data).
The possible perspectives resulting from this work allow us to think that it would be interesting to link this framework to several services capable of delivering positive incentives and to test it within a concrete business context in real time. More specifically, it would be possible to create positive incentives, such as ecological nudges, within applications used in the context of mobility, travel or food.
References:
Moro, A., & Holzer, A. (2020). A Framework to Predict Consumption Sustainability Levels of Individuals. Sustainability, Volume 12 (4) (Green Technology Innovation for Sustainability), 1-26.
Moro, A., & Holzer, A. (2019, 17 December). Supporting Green IS through a Framework Predicting Consumption Sustainability Levels of Individuals. Paper presented at International Conference of Information Systems (ICIS), Munich, Germany.
Author(s) of this blog post
Adrian Holzer is professor of information systems management at the University of Neuchâtel. He holds a doctorate in information systems from the University of Lausanne. He was an associate researcher at EPFL, co-head of the interdisciplinary platform at the University of Lausanne and researcher at the SNSF at the Ecole Polytechnique de Montréal. His research interests cover digital transformation in organizational, educational, and humanitarian contexts.
Arielle Moro carried out a PhD thesis in Information Systems at the University of Lausanne (UNIL). Her thesis aimed at exploring the tradeoff between predicting mobility and preserving location data privacy. Currently, she is a senior researcher (post-doc) at the Information Management Institute (IMI) and a lecturer at the University of Neuchâtel (UniNE). Her research interests focus on analyzing human behavior in various domains (e.g., mobility, sustainability, teaching, health) using machine learning techniques.
