Power of Difference Assessment System; The Shifting Boundary Between Open-Source and Proprietary Software and its Role in the Future of AI
Yager, Connor, School of Engineering and Applied Science, University of Virginia
Ku, Tsai-Hsuan, EN-Engineering and Society, University of Virginia
Ibrahim, Ahmed, EN-Comp Science Dept, University of Virginia
The Sum is a Charlottesville-based non-profit, with the stated goal of helping people learn about their implicit biases and understand how to better communicate with people of other backgrounds. Their Power of Difference Assessment, or PDA, is a tool used to identify these implicit biases and collect data that professionals at The Sum can use to consult the assessment taker. Under the initial PDA system, a lack of automation prevented scaling to a larger user base, restricting the growth of the organization and its ability to collect data. Our team was tasked with enhancing the system with features that would allow it to serve as a tool in institutional studies, with the ability to support thousands of assessment-takers and collect meaningful data for use in research.
The genesis of the open-source software development methodology in the 1990s effectively divided the professional software development world into two distinct groups: those individuals and corporations who support free and open-source software and those who support the traditional proprietary model. Today, these lines have become blurred, as proprietary corporations pledge support for the open-source community while open-source developers demand the acceptance of stricter licensing options. With the advent of human-level artificial intelligence on the horizon, this shifting spectrum of code ownership and profitability has great ethical implications, as the most equitable ownership of a technology primed to change the lives of all people comes into question.
In the development of human-level artificial intelligence, the most destructive potential bug in its design is bias. In the same way that human bias is largely dependent on background, AI bias is based on faulty data. This data is the result of human-created algorithms that fail to accurately account for perspectives outside of those of its creator. The open-source vs. proprietary debate extends into this field, with the open-source community championing collective code-ownership, while proprietary supporters promise more rigorous safety testing and quality assurance. While it is difficult to make predictions about future technologies, evaluating the evolving relationship between open-source and proprietary technologies with respect to bias is an essential step in safety-engineering artificial intelligence.
BS (Bachelor of Science)
SCOT, Open-source vs. Proprietary, Artificial Intelligence, The Sum, Bias in AI
School of Engineering and Applied Sciences
Bachelor of Science in Computer Science
Technical Advisor: Ahmed Ibrahim
STS Advisor: Tsai-Hsuan Ku
Technical Team Members: Peter Felland, Amelia Hampford, Nuzaba Nuzhat, Sam Shankman, David Xue, Carl Zhang, William Zheng