Abstract
This capstone research will help adapt sustainability practices to the growing use of AI and computing resource intensive tasks. The use of artificial intelligence (AI) poses a threat to the environment and sustainability due to its high energy consumption, increasing demand for data centers, and growing carbon footprint. Large-scale data processing such as model training requires a large amount of electricity, water, and raw materials. Losing a focus on sustainability causes greenhouse gas emissions to rise which intensifies climate change, causing severe weather events like hurricanes, droughts, and wildfires.
I modeled the energy cost of database queries with a carbon-aware query optimization application about UVA study groups, which can be extended to any database heavy application. This application will track the energy cost of database queries and visualize carbon emissions throughout the day. The technologies used to build this application are a MySQL database and a PHP-enabled webspace. The purpose of this project is to show that any website or application that heavily involves databases can be carbon aware and help implement operational efficiency. It is important to consider the human and social dimensions of this technology because this directly impacts users’ awareness of the energy cost and carbon footprint of database operations they perform. Implementing carbon-aware query optimization into applications can help users make more sustainable choices, and reduce unnecessary energy consumption. However, an incorrect implementation may cause users to underestimate the environmental impact of their resource intensive tasks, which is not the intended effect. The societal concerns this technology addresses are climate change and environmental sustainability. This project could promote more applications, whether in the workplace or personally, to adopt carbon-aware computing practices, and lead to more discussions across teams on responsible and energy efficient software practices.
The relationship between corporate and individual contributions is analyzed using a comparative sociotechnical framework and the concept of power asymmetry. This approach compares two primary contributors: individual users, who generate demand through AI usage, and corporations, which design, deploy, and scale AI systems. Power asymmetry highlights how corporations have greater access to controlling demand, use of resources, and building new infrastructure, effectively causing them to have more influence over environmental outcomes. By examining this imbalance, this framework helps identify where responsibility for AI’s environmental impact is most concentrated.
To investigate this problem, I used a combination of secondary source analysis and comparative case studies. This included peer-reviewed articles, industry reports, and third-party estimates to quantify energy use, carbon emissions, and water consumption associated with AI systems. I conducted a comparative case study of large-scale data centers, including Microsoft’s facility in Mount Pleasant and data center clusters in Northern Virginia, to evaluate corporate infrastructure impact. Additionally, I analyzed and normalized energy consumption data across common AI tasks for users such as text generation, summarization, and image generation to compare their relative environmental costs.
The findings show that while both individual users and corporations contribute to the environmental impact of AI, corporations have significantly greater influence over environmental outcomes. Although individual AI queries consume relatively small amounts of energy, their cumulative impact is driven largely by corporate decisions regarding infrastructure expansion, system design, and large-scale deployment. The research also reveals a lack of transparency in corporate reporting of energy and resource usage, as well as the presence of the “myth of personal responsibility,” which shifts accountability toward users despite corporations having the most control over the growth of AI and data centers.
Putting the Capstone project and STS research together, the goal is to understand the importance of moving into a more sustainable world where both users and producers do their part to adopt sustainable practices. This would look like company operations and data centers becoming more carbon neutral and transparent with their carbon emissions, and users knowing the carbon contributions of their actions through carbon-aware applications. The capstone project shows how to make database applications carbon-aware, while the STS research reveals which groups shape the environmental outcomes of AI. They both help to further the computer science practice by promoting more sustainable AI systems that are technically necessary and socially beneficial.
