Frontiers in Fairness: Weighing fairness in the data science development cycle

If the front door is locked, text (802) 999-4944 and someone will let you in.

Summary:
Industries both old and new have benefited from leveraging AI to improve decision making. While the direct benefits of algorithmic decision making are clear, the broader societal impacts often go under studied. Data science practitioners can help lead in this area by moving beyond traditional performance metrics and pushing ourselves to define, measure, and support fair outcomes in our work.

This talk will take a broad view of fairness, accountability, transparency, and ethics (FATE) in data science. Starting with a discussion of how to conceptualize and define fairness, we will then move to exploring sources of bias in AI systems. Next, we will review specific metrics that provide visibility into dataset and prediction bias. To tie these topics together, we will discuss how FATE can be put into practice and baked into the data science development process. Finally, we will review a piece of recent regulation that outlines a fairness auditing methodology.

About the Speaker:
Josh Minot works in the insurance industry for Mass Mutual