Machine Learning for the Social Sciences (ML4SS)

This repository contains the slides used for the short course entitled "Machine Learning for the Social Sciences" (ML4SS), taught to the students of the Doctoral Program in Sociology and Social Research at the University of Trento (Academic Year 2020-2021).

Contents

Lecture 1: Introduction and Main Paradigms

• What is Machine Learning? A short history of Artificial Intelligence and other beautiful worlds

  • The cradle of AI: Philosophy
  • The cradle of AI: Mathematics and Logic
  • The cradle of AI: Neuroscience
  • Contemporary history of AI
  • Strong vs Weak AI
  • Defining Machine Learning

• A Learning Bipolarism: Supervised Learning and Unsupervised Learning

  • Supervised Learning: Classification
  • Supervised Learning: Regression
  • Unsupervised Learning: Main Tasks

• Is Machine Learning just glorified Statistics?

  • ML vs Statistics: Hype and Reality
  • Differences between ML and Statistics
  • Machine Learning: Beyond Forecasting
  • Epistemological Remarks

Lecture 2: Preliminaries and Algorithms

• First Technical Concepts

  • Learning from Data: Train and Test
  • Generalizability in ML
  • Underfitting and Overfitting

• Evaluating a Model

  • Introduction to Model Evaluation
  • Classification: Evaluation Metrics
  • Regression: Evaluation Metrics

• Algorithms: Part I

  • Back to basics: Logistic Regression from the ML Perspective
  • Regularization Methods (Ridge and Lasso)
  • The Bias-Variance Trade-off
  • Decision Trees

Lecture 3: Algorithms and Societal Impact

• Algorithms: Part II
  • Shortcomings of Decision Trees
  • Bagging
  • Random Forests
  • Gradient Boosting Machines
  • Unsupervised Learning: Recap
  • Clustering: Distance and Similarity
  • Internal and External Evaluation in Clustering
  • Families of Clustering Approaches
  • k-Means
  • DBSCAN
  • Hierarchical Agglomerative Clustering

-Societal Impact of AI

• Perspectives on AI and ML Consequences in Society
• Racial Bias in Facial Recognition
• Racial Bias in Predictive Policing
• Algorithmic Bias in Education
• Critical Conclusive Reflections: the Role of Social Scientists