Coded Bias is a documentary directed by Shalini Kantayya and features some prominent women in the field of Artificial Intelligence(AI). It documents how gender, racial and class biases are baked into technological AI systems, how they impact certain communities and not others and what should be done to change this current state of affairs. Some of the speakers include:
- Joy Buolamwini
- Cathy O’Neil, PhD
- Silkie Carlo
- Virginia Eubanks
- Deborah Raji
- Timnit Gebru
- Safiya Umoja Noble, PhD
- Amy Webb
The documentary starts off with Joy Buolamwini, an MIT Media lab researcher who was using a face detector algorithm that could not recognize her until she wore a white mask. Joy went ahead to discover that this was not an isolated incident within the field of AI.
AI has for years been used to infringe on the rights of marginalized communities particularly women and people of color. This is because AI technologies that are used to predict the future have been created using skewed data sets thus resulting to skewed results. The algorithmic results show inequalities that persist because they are based on biased historical data. These biases are in turn embedded into the technology.
Some of the practices of discrimination discussed in the film include: biased credit card worthiness evaluations, home ownership & the mortgage crisis, college admission technologies, recruiting tools with AI algorithms that rejected women in Amazon, hospital technology that prioritized medical care for healthier white patients over sicker black patients, Microsoft’s Tay that turned into a misogynistic robot based on internet training, Facebook’s experiment during the 2016 elections, employee evaluation models, criminal risk assessment tools used by judges and HR hiring analytics & automations.
The common thing across these technologies is that they evaluate based on homogenous group of people and not unique features and identities. The result of their usage, Joy says, is ‘people suffering from algorithmic harm without accountability’. Algorithms classify really well, but when that is based on false historical information and the algorithms that have not been audited for accuracy, then systemic biases are going to be hard-wired into new technologies.
The film also discusses mass surveillance state, state surveillance and corporate surveillance. A segment of the documentary features an interviewee in China where facial surveillance is very embedded into most services to maintain social order and assign people social credit. It is definitely a must watch for anyone interested in learning about the impact of unregulated AI practices.
Coded Bias is a great and necessary watch. Educative and informed as told by women and black women. This group of people have both expertise in the study of algorithms and experience in by algorithmically marginalized or punished. Anyone in the digital world(which is everyone at this point) should be interested in this quest to evaluate algorithms, classify them as fair before they are released for use in society.
