Introduction (week 1)
Machine learning in human contexts
Recommendation platforms as ecosystems
Recommendation systems: from classic methods to state-of-the-art
Decision making under uncertainty and the paradox of choice
The problem of prediction vs. policy
[Unit #1] Human choice: modeling and prediction (weeks 2, 3)
Choice prediction as a learning problem
Decision theory and rational models of choice
Prospect theory and other behavioral theories of choice
The role of uncertainty in decision making
Learning with behavioral choice data
[Unit #2] Sets of alternatives: choice behavior and recommendation (weeks 4, 5)
What it means to recommend
Discrete choice: choosing from sets of alternatives
Behavioral violation of choice axioms
Prediction in discrete choice: approaches and challenges
Accuracy vs. value, satisfaction, and regret
[Unit #3] Recommendation as action: outcomes and effects (weeks 6, 7)
What makes for a good recommendation?
Content-based methods vs. collaborative filtering approaches
How choice data affects recommendation
How recommendation affects choice data
[Unit #4] Dynamics: temporal aspects of recommendation systems (weeks 8, 9)
Recommendation as a (very challenging) policy problem
The pros and cons of prediction-based recommendation
Retraining dynamics and recommendation trajectories
Temporal phenomena: hyper-popularization, polarization, echo chambers, and filter bubbles
Challenges and prospective solutions
[Unit #5] Recommendation ecosystems: stakeholders and incentives (weeks 10, 11)
Recommendation systems as an ecosystem
Multiple stakeholders: users (consumers), content creators (suppliers), and the system
Incentives and strategic behavior
The system as a social planner: misaligned incentives, and what to do about them
Projects: initiation and demonstration (week 12)