Abstract
As digital systems become more advanced, many everyday decisions are now made or supported by algorithms. These systems are designed to improve efficiency, speed, and performance, but they also raise an important question: how can automated systems remain fair, transparent, and easy to use for the people affected by them? This problem appears in many areas, including computing infrastructure and consumer markets. In shared computing environments, powerful resources like GPUs are expensive and often limited, however they are not used efficiently because of confusing tools and unfair access patterns.Currently, scheduling systems can leave GPUs running at only around half utilization, while users wait long periods to run jobs . At the same time, in consumer markets, algorithmic pricing systems are changing how essential goods like groceries are priced. These systems can adjust prices based on demand or user data, but they often lack transparency, leading consumers to feel that pricing is unfair or inconsistent . While these systems aim to optimize performance or profit, they can create confusion, reduce trust, and increase inequality if users do not understand how decisions are made. This shows that technical systems must balance efficiency with fairness and accessibility in order to be successful.
The technical project focuses on designing and developing a pop-up GPU scheduler that improves access to shared GPU resources in small-scale environments. GPUs are powerful but expensive, and many students and researchers rely on a single shared device. Current tools like command-line schedulers are difficult for beginners to use and can often lead to inefficient usage or unfair access. The proposed system introduces a web-based interface where users can log in, submit jobs, and view the status of the GPU in real time. Jobs are executed inside Docker containers to ensure safety and prevent conflicts between users. The scheduler uses a simple priority-based queue to determine job order while enforcing limits to prevent one user from monopolizing the GPU . Both user testing and system performance testing were used to evaluate the system. In user testing, participants compare the scheduler interface to traditional command-line tools, measuring how easily they can submit and run jobs. System testing checks how well the scheduler can handle multiple jobs at once and keep things running smoothly even when the system is under heavy load. Surveys and interviews are also used to measure user understanding and perceived fairness. The results show that a simple, visual interface improves accessibility and helps users better understand how resources are allocated. By making scheduling behavior more transparent, the system not only improves efficiency but also increases user trust and fairness in shared computing environments.
The STS research project examines how dynamic and algorithmic pricing in grocery markets affects consumer perceptions of fairness. Traditional pricing systems present the same price to all consumers, which creates a sense of stability and fairness. However, algorithmic pricing introduces systems that can adjust prices in real time based on factors such as demand, location, or purchasing behavior. This can lead to situations where different consumers pay different prices for the same product without knowing it. Consumers often view this as unfair, especially when it involves essential goods like food . The study analyzes evidence from academic research, policy reports, and real world examples to understand how these systems operate and how consumers respond to them. A key finding is that fairness depends not only on the price itself, but also on whether consumers understand how prices are determined. When pricing systems are opaque, consumers feel a loss of control and are more likely to distrust retailers. The project uses Actor Network Theory to explain that pricing outcomes are shaped by interactions between multiple actors, including algorithms, companies, consumers, and regulators. These interactions create systems where power and information are unevenly distributed. While dynamic pricing can improve efficiency and help businesses respond to market changes, it can also increase inequality and reduce trust if it is not transparent or well-regulated. The research shows the importance of communication, data privacy protections, and policy oversight in maintaining fairness.
Together, these projects show that the success of modern automated systems depends on more than just technical performance. In both GPU scheduling and algorithmic pricing, systems must be designed with attention to fairness, transparency, and user experience. A system that is efficient but difficult to understand or perceived as unfair will not be widely accepted or effective in practice. This work contributes to a deeper understanding of how technical design choices interact with social factors such as trust and accessibility.
