Bias in-Bias Out: The Injustice/Ambiguity of Algorithms

CALL FOR PARTICIPANTS Academic Learning Community: 

Fall 2018

Conveners: Gillian Hue, Neuroscience & Behavioral Biology,

Edward L. Queen, Center for Ethics

Computers increasingly control our lives.  While so self-evident as to be almost banal, the implications of this statement deserve greater social, cultural, intellectual, and political attention.  Computers, or, more precisely, the algorithms that they run, can determine if we are identified as potential terrorists, denied a mortgage, or even have our children are removed from our homes, Additionally, the algorithms themselves are structured in ways that highlight conflict and disagreement, often with horrid consequences. 

Behind these facts lies a disturbing and woefully underexamined reality.  Because these algorithms seem to operate rationally and mechanically, there exists an unexamined presumption that they are immune to bias and discrimination.  Not only does that seem not to be the case, but also once an artificial intelligence program starts we remain unaware not only of how it learns, but even of why it makes the decisions it does.  We are blind to the biases and prejudices the system teaches itself.  Do minor perturbations in a system (biases) at the beginning of the process result in massive distortions down the line or does the system learn to remove those biases?  We simply do not know.

This academic learning community seeks to engage the issues of potential biases in artificial intelligence with the goal of heightening awareness the centrality of artificial intelligence in our daily lives and its effects on all segments of society in order to enable faculty from a variety of disciplines to think and learn about this together and to find ways to incorporate this knowledge into their teaching and research.

Topics to be addressed include:

  1. Crime and Punishment—what role do AI-based systems play in decisions that determine punishment (broadly construed to include denials of jobs and loans, removal of children from a home, probation revocation or denial, arrest, sentencing, etc.)?
  2. The exacerbation of bias—on-line sexism, racism, bullying, incivility, and memes leading to violence.
  3. Privacy/Security—from on-line stalking to echo chambers to the dark web.
  4. AI/Machine based learning—the future is out of your control.  What happens when we cannot determine how or what a machine learns?

Academic Learning Communities are informal seminars that are intended to:

  • engage faculty in collaborative explorations of innovative research and teaching topics;
  • bring guest speakers to campus to enhance the curriculum and learning; and/or
  • help disseminate important research discoveries and innovative learning strategies to the broader community.

Particulars:

  • The Seminar will meet from 12:00 noon-1:30 on the following Thursdays: September 27, October 18, November 15, December 6.
  • Possible outcomes of the Academic Learning Community include: the development of a University Course or other team-taught courses, collaborative writing across disciplines, guest lecturing within one another’s courses, and use of findings and discussion to further a larger conversation not only throughout Emory University, including the Oxford campus, but also throughout the Atlanta metro area.
  • Each meeting will balance presentations by the facilitator or invited speakers with group discussions of relevant readings and presentations by seminar participants.
  • Up to 20 participants will be accommodated and will include both faculty and graduate students from across the university.

To apply, please fill out this form here.

The deadline for application is Friday, August 24.

 

Selections will be announced in early September. A limited number of spaces will be reserved for graduate students based on the relevance of their research to the topic.