Artificial Intelligence Bias in an Age of a Technical Elite (2019-2020)
The recent expanded use of machine learning techniques in real-world applications has been driven by data availability and processing power. Vast electronic data troves have become available to practitioners, making relatively old data analysis tools much more likely to solve difficult challenges, and massive improvements in processing power have significantly reduced the time taken to find solutions to these challenging problems.
This rapid growth in applied machine learning has come so fast that there is widespread potential for disruption and harmful misapplication. This is especially true since there is a troubling lack of diversity among highly skilled practitioners – the “technical elite.”
This Bass Connections project team will examine the disruptive and potentially harmful implications of machine learning, when applied by a not-so-diverse elite of highly skilled practitioners. The project will seek to open the closed world of applied machine learning to students and the public through development of a performance-driven workshop. Team members will be tasked with creating performance art that addresses the disruptive nature of the technology and its potential for harm if misapplied.
The project team will apply open source machine learning tools to develop a workshop in applied machine learning for performance artists. The goal is to distil the tools into a user-friendly set of interfaces that could be applied by artists.
The project will use the cross-listed multidisciplinary course Performance and Technology as a testing ground for tools in development and datasets. Team members will initially define a few core tools and application areas that are practical yet intriguing enough to reveal the potential for disruption and potentially harmful misapplication of machine learning.
Possible focus areas include tracking objects and people for identification and classification and filtering applicant pools for desirable traits through survey data analysis. These themes will set the stage for a great dramatic performance topic and present artists with an opportunity to use realistic machine learning tools to address these issues in believable ways.
Open source tools and datasets for machine learning for artists; performance pieces for academic and public audiences; artificial intelligence objects that will be available for future performance projects
Ideally, this team will consist of 12 undergraduates and 3 graduate students.
We would like to have one graduate student who is actively studying machine learning through their research and one or two additional graduate students from the arts.
Undergraduates on the team will include a mix of good coders and artists. The artists should be interested in the implications of machine learning but not skilled practitioners or even skilled coders. Their role is to “keep it real” so that the tools developed can be used by as diverse a group as possible.
At the end of this project, artists will be able to add a machine learning-based performance to their portfolio. The coders will have an application involving clients of a diverse makeup. It is possible that the resulting open source tools developed could become the basis for a startup or new research endeavor.
Participation during the summer components of the project is optional.
Summer 2019 – Summer 2020
- Summer 2019 (Optional): Subset of team begins work on skill-building with open source machine learning tools
- Fall 2019: Begin development of machine learning workshop
- Spring 2020: Begin Performance and Technology course; develop performance pieces
- Summer 2020 (Optional): Subset of team works on refinement or workshop tools
Independent study credit available for fall and spring semesters; summer funding available
Image: Artificial Intelligence 2018 San Francisco by O’Reilly Conferences licensed under CC BY 2.0
/faculty/staff Team Members
Martin Brooke, Pratt School of Engineering-Electrical & Computer Engineering*
Shaundra Daily, Pratt School of Engineering-Electrical & Computer Engineering*
Thomas F. DeFrantz, Arts & Sciences-African and African American Studies*
Matthew Kenney, Arts & Sciences-Art, Art History, and Visual Studies*
Cynthia Rudin, Arts & Sciences-Computer Science*