Will Landecker
London, E1 8PH, UK
Portland, ME 04101
2nd floor
11th floor
Boston, MA 02115
2nd floor
Talk recording
The phrase "it's in the data" has become synonymous with "it's objectively true." But datasets, and algorithms derived from them, are curated, designed, and built artifacts. They reflect the inevitable compromises that must be made during the design process. As our ability to collect data and build powerful algorithms grows, so does our ability to be misled by subtle and unintentional problems of construct validity, alignment, and evaluation. Will Landecker has worked three sides of this issue: a PhD in machine learning interpretability, a decade leading data science and machine-learning teams in Silicon Valley, and now an expert poring over algorithmic evidence in criminal cases. In this talk, he will share stories from experiences in academia, industry, and the justice system. We will see examples where those different sectors keep each other in balance, and others where they amplify each other's problems.
About the speaker
