Author: Rumman Chowdhury
Publisher: Parity
Publication Year: 2021
Summary: The following talk addresses the question: Why does technology need ethics? Dr. Chowdhury describes 2 kinds of bias, a quantitative one and a qualitative one, specifically data and reporting bias and design bias. She poses the question: what are we assuming this AI can do? As data scientists, we make assumptions about our data and what it is measuring and we forget what it cannot measure quantitatively. An average person, however, thinks of bias in terms of -isms (sexism and racism). “Data is not an objective truth, it is reflective of pre-existing institutional cultural and social biases.” As data professionals, it is common to say “the data says so;” however, we must keep in mind that the data does not come without context, the data comes because people have made decisions. Some questions to answer: How is the data being collected, and might that introduce bias? What assumptions are we making about our model and its applicability to the question? Humans are inconsistent and socialized into a behavior that tells us to act a certain way. Creation of fancy models is not the solution, it is the creation of unbiased models are provide true solutions.
