Generalization
Problem: Learn from evaluations to increase satisfaction with future interactions
Strategy:
- Learn biases toward configurations/prices
- Learn biases toward agents to interact with
- Learn biases other agents have toward configurations/prices or other agents...
Example: A query planner is underutilized, learns what kinds of QPAs the users are favoring, and adjusts its advertisements/prices/capabilities to compete better
Status:
- Currently implements price adjustments
- Working toward excluding cheaters in long run
- A basis for recommendation services