Positionality-Aware Machine Learning

Explore positionality-aware machine learning in this comprehensive tutorial from the FAT*2020 conference. Delve into the concept of positionality and its impact on AI systems through real-world examples like the US Supreme Court, ICD, and ImageNet. Examine the goals of mechanical objectivity and understand the importance of deliberation, repetition, and feature discovery in AI development. Gain insights into the generalized workflow of positionality-aware machine learning and analyze it from technical and systems perspectives. Learn how to create more equitable and context-aware AI systems by considering the diverse positions and perspectives that influence data collection, model development, and implementation.

Introduction
Project Overview
Tutorial Structure
Examples
US Supreme Court
ICD
ImageNet
Goals
Mechanical objectivity
Positionality
Summary
Spectrogram
Accuracy
Generalized Workflow
Deliberation
Repetition
Feature Discovery
Feature Analysis
Three Perspectives
Technical Perspectives
Systems Perspectives