Developing frameworks for an equitable future: from building decarbonization to generative modeling.
Name
DeSimone-zoed-smarchs-arch-2024-thesis.pdf
Description
Thesis PDF
Size
5.23 MB
Format
Adobe PDF
Checksum (MD5)
b0a65c74ccfcc9c423a96bbb09e8315e
Author(s)
De Simone, Zoe
Advisor(s)
Reinhart, Christoph
Wilson, Ashia
Date Issued
May 2024
Publisher
Massachusetts Institute of Technology
Abstract
In this thesis I develop computational frameworks to understand equity under two perspectives: building decarbonization policy and generative modeling.
Part 1 - Equitable building decarbonization
Buildings significantly contribute to global carbon emissions, necessitating urgent decarbonization to meet 2050 climate targets. The U.S. strives for net-zero emissions by 2050, supported by federal incentives promoting building upgrades. However, financing deep retrofits for all U.S. homes exceeds available public funds. This chapter proposes a model that examines long-term carbon reduction trajectories under various incentive policies, focusing on fairness and equity. Using Oshkosh, WI, as a case study, it explores the philosophical, economic, political, and mathematical dimensions of creating just and effective decarbonization policies that ensure healthy, low-carbon homes for all.
Part 2 - Equitable diffusion models
Generative Text-to-Image (TTI) models, while capable of producing high-quality images, often replicate training data biases. Traditional fairness views in machine learning, which consider fairness as binary, are challenged. This section introduces DiffusionWorldViewer, a novel framework with a Web UI that enables users to analyze the underlying worldviews of diffusion models and edit model outputs to align with their personal fairness perspectives, thus promoting a diverse understanding of fairness in AI technologies.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology. Department of Architecture
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
