Sustainable development

Sustainable development is an organizing principle that aims to meet human development goals while also enabling natural systems to provide necessary natural resources and ecosystem services to humans. The desired result is a society where living conditions and resources meet human needs without undermining the planetary integrity and stability of the natural system. Sustainable development tries to find a balance between economic development, environmental protection, and social well-being. The Brundtland Report in 1987 defined sustainable development as "development that meets the needs of the present generation without compromising the ability of future generations to meet their own needs". The concept of sustainable development nowadays has a focus on economic development, social development and environmental protection for future generations. Sustainable development was first institutionalized with the Rio Process initiated at the 1992 Earth Summit in Rio de Janeiro. In 2015 the United Nations General Assembly (UNGA) adopted the Sustainable Development Goals (2015 to 2030) and explained how the goals are integrated and indivisible to achieve sustainable development at the global level. The UNGA's 17 goals address the global challenges, including poverty, inequality, climate change, environmental degradation, peace, and justice. Sustainable development is interlinked with the normative concept of sustainability. UNESCO formulated a distinction between the two concepts as follows: "Sustainability is often thought of as a long-term goal (i.e. a more sustainable world), while sustainable development refers to the many processes and pathways to achieve it." The concept of sustainable development has been criticized in various ways. While some see it as paradoxical (or as an oxymoron) and regard development as inherently unsustainable, others are disappointed in the lack of progress that has been achieved so far. Part of the problem is that "development" itself is not consistently defined.

About this result

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Summary

CauseOccam: Learning Interpretable Abstract Representations in Reinforcement Learning Environments via Model Sparsity

Sergei Volodin

"I choose this restaurant because they have vegan sandwiches" could be a typical explanation we would expect from a human. However, current Reinforcement Learning (RL) techniques are not able to provide such explanations, when trained on raw pixels. RL algorithms for state-of-the-art benchmark environments are based on neural networks, which lack interpretability, because of the very factor that makes them so versatile – they have many parameters and intermediate representations. Enforcing safety guarantees is important when deploying RL agents in the real world, and guarantees require interpretability of the agent. Humans use short explanations that capture only the essential parts and often contain few causes to explain an effect. In our thesis, we address the problem of making RL agents understandable by humans. In addition to the safety concerns, the quest to mimic human-like reasoning is of general scientific interest, as it sheds light on the easy problem of consciousness. The problem of providing interpretable and simple causal explanations of agent's behavior is connected to the problem of learning good state representations. If we lack such a representation, any reasoning algorithm's outputs would be useless for interpretability, since even the "referents" of the "thoughts" of such a system would be obscure to us. One way to define simplicity of causal explanations via the sparsity of the Causal Model that describes the environment: the causal graph has the fewest edges connecting causes to their effects. For example, a model for choosing the restaurant that only depends on the cause "vegan" is simpler and more interpretable than a model that looks at each pixel of a photo of the menu of a restaurant, and possibly relies as well on spurious correlations, such the style of the menu. In this thesis, we propose a framework "CauseOccam" for model-based Reinforcement Learning where the model is regularized for simplicity in terms of sparsity of the causal graph it corresponds to. The framework contains a learned mapping from observations to latent features, and a model predicting latent features at the next time-steps given ones from the current time-step. The latent features are regularized with the sparsity of the model, compared to a more traditional regularization on the features themselves, or via a hand-crafted interpretability loss. To achieve sparsity, we use discrete Bernoulli variables with gradient estimation, and to find the best parameters, we use the primal-dual constrained formulation to achieve a target model quality. The novelty of this work is in learning jointly a sparse causal graph and the representation taking pixels as the input on RL environments. We test this framework on benchmark environments with non-trivial high-dimensional dynamics and show that it can uncover the causal graph with the fewest edges in the latent space. We describe the implications of our work for developing priors enforcing interpretability.

2021

The 2030 Agenda to assess the sustainability of campus operations: EPFL as a case study

Léa Barbey

The EPFL sustainability unit was created in 2007 to coordinate, support and propose sustainable development actions on the campus. Its 2018-2020 strategy introduced a project called CORE SUSTAINABILITY, aiming at the progressive integration of sustainability within the organization, at the heart of EPFL's core missions (education, research, innovation), but also in the various Vice-Presidencies and their departments, thus at the heart of the organization. The objective is to empower all stakeholders, by committing them to sustainability goals that they should integrate to their management and reporting tools, in order to multiply the achievements and the exposure in this field. It is important to address the question of how this integration should be conducted. This master project would analyze the use of the UNO's reference frame, namely the 2030 Agenda and its well-known seventeen Sustainable development goals (SDGs). The working hypothesis would be as follows: 1. Are SDGs an effective tool for conducting sustainability integration strategies in universities? 1.1 Do SDGs make it possible to define indicators that stakeholders consider relevant? 1.2 Do SDGs bring stakeholders together around a common notion for sustainable development? It is necessary to give a clear definition of the different notions used in that working hypothesis. "Effective" is defined as "that achieves its goal, that leads to useful results", which is understood in the sense of creating a common vision, that could then easily be implemented. For conducting this research project, and because of the limited time available, a focus on three illustrative SDGs will be done (for example: 13. Climate action, 5. Gender equality, 12. Responsible consumption and production or 7. Energy, 9. Industry, innovation and infrastructure, 4. Quality education). This choice should however not erase the general theoretical frame from our minds. The methodology for this project would consist in a comparative case study on processes and reference frame choices for implementing sustainability (and more specifically by means of sustainability's indicators definition) at a university campus level. The different steps would be as follows: a) Literature review b) Review of sustainability policies at the universities c) Application of the MONET methodology (similar to the canton de Vaud methodology) to define indicators relative to those SDGs for the administrative services of EPFL: (i) distribution of the targets to the respectively responsible services; (ii) review of existing indicators at the EPFL campus level (cf. GRI reporting in 2012 and 2014) and matching to the targets at issue; (iii) participative definition (by the services) of indicator(s) for each target, based on the existing indicators listed d) Analysis of the proposed methodology for indicators definition: (i) qualitative data collection (survey and interviews with key stakeholders); (ii) qualitative data analysis (are the defined indicators seen as relevant by the stakeholders? Do stakeholders share a common vision of the notion of sustainability?) e) Comparative analysis of another methodology for indicators definition (in another university): (i) quantitative analysis: are the selected indicators similar? (ii) qualitative analysis: are the defined indicators seen as relevant by the stakeholders? Do the stakeholders share a common vision of the notion of sustainability?

2020

Quantification and Classification of Handprint-based Net Positive Resource Efficiency Measures in the Swiss Industry Sector

Florian Elias Landerer

The task of the master project is to analyse a data set with information of several hundred companies from Switzerland. The goal is to find in the Swiss industry sector a classification/typology for the nature of resource efficiency of innovating companies and comparing their impacts in Switzerland and abroad.

2018

Environmentally friendly

Environment friendly processes, or environmental-friendly processes (also referred to as eco-friendly, nature-friendly, and green), are sustainability and marketing terms referring to goods and services, laws, guidelines and policies that claim reduced, minimal, or no harm upon ecosystems or the environment. Companies use these ambiguous terms to promote goods and services, sometimes with additional, more specific certifications, such as ecolabels. Their overuse can be referred to as greenwashing.

Upcycling

Upcycling, also known as creative reuse, is the process of transforming by-products, waste materials, useless, or unwanted products into new materials or products perceived to be of greater quality, such as artistic value or environmental value. Upcycling is the opposite of downcycling, which is the other part of the recycling process. Downcycling involves converting materials and products into new materials, sometimes of lesser quality. Most recycling involves converting or extracting useful materials from a product and creating a different product or material.

Environmental technology

Environmental technology (envirotech) or green technology (greentech), also known as clean technology (cleantech), is the application of one or more of environmental science, green chemistry, environmental monitoring and electronic devices to monitor, model and conserve the natural environment and resources, and to curb the negative impacts of human involvement. The term is also used to describe sustainable energy generation technologies such as photovoltaics, wind turbines, etc.

Sustainable Food Systems: Connecting Farmers and Consumers

Explores a student project connecting farmers and consumers through sustainable food systems.

Ecological economics

Ecological economics, bioeconomics, ecolonomy, eco-economics, or ecol-econ is both a transdisciplinary and an interdisciplinary field of academic research addressing the interdependence and coevolution of human economies and natural ecosystems, both intertemporally and spatially. By treating the economy as a subsystem of Earth's larger ecosystem, and by emphasizing the preservation of natural capital, the field of ecological economics is differentiated from environmental economics, which is the mainstream economic analysis of the environment.

Topics in environmental studies

Environmental studies (EVS or EVST) is a multidisciplinary academic field which systematically studies human interaction with the environment. Environmental studies connects principles from the physical sciences, commerce/economics, the humanities, and social sciences to address complex contemporary environmental issues. It is a broad field of study that includes the natural environment, the built environment, and the relationship between them.

Topics in environmental policy

Environmental policy is the commitment of an organization or government to the laws, regulations, and other policy mechanisms concerning environmental issues. These issues generally include air and water pollution, waste management, ecosystem management, maintenance of biodiversity, the management of natural resources, wildlife and endangered species. For example, concerning environmental policy, the implementation of an eco-energy-oriented policy at a global level to address the issues of global warming and climate changes could be addressed.