Learning With and From People - Collaborative Consumption Systems and User Engagement

Explore a comprehensive lecture on integrating human participation in computational systems and machine learning frameworks. Delve into collaborative consumption systems like shared mobility and Airbnb, focusing on user engagement strategies to improve system effectiveness. Learn about novel multi-armed bandit algorithms and online learning methods for optimizing incentives and understanding users' switching costs. Discover the application of hemimetrics as a structural constraint to tackle data sparsity challenges. Examine real-world applications, including incentivizing users to explore unreviewed Airbnb hosts and addressing imbalance in bike-sharing systems. Gain insights from a smartphone app deployment in Mainz, Germany, showcasing the practical implementation of these algorithms. The lecture also touches on future directions in teaching for biodiversity monitoring and air quality sensing.

Intro
New Opportunities: Smart Mobility
Rethinking Data Science / ML
Research Overview
Imbalance Problem
Incentivizing Users for Switching Choices
Architecture of our System
Simulations: Boston Hubway • Simulator using Boston Hubway data
Deployment: Mainz, Germany • 30 days pilot deployment in Mainz, Germany
Learning Pairwise Switching Costs Large Number of Learning Instances
Structural Constraints: Alternate View Independent
Teaching for Bio-Diversity Monitoring Identifying Endangered Woodpeckers
Teaching: Future Directions
Sensing the Air We Breathe