The Quanta Podcast Game Theory, Algorithms and High Prices
28 snips
Nov 25, 2025 Ben Brubaker, a computer science writer for Quanta Magazine, explores the intriguing intersection of algorithms and pricing in the retail world. He discusses how competition can falter when automated algorithms inadvertently lead to rising prices, complicating traditional views of collusion. Brubaker also explains learning algorithms, revealing how they adapt to competitors' pricing. The conversation dives into research simulations showing emergent collusive behaviors and the contentious definitions surrounding this issue, ultimately shedding light on the complexities of modern pricing strategies.
AI Snips
Chapters
Transcript
Episode notes
Algorithms Can Raise Prices Without Intent
- Algorithmic pricing can produce high prices without human intent or secret deals.
- Computer scientists study these emergent effects by analyzing learning algorithms' interactions.
Learning Algorithms Are Decision Recipes
- Learning algorithms continually adjust decisions based on incoming data like competitor prices.
- Their mathematical behavior can be studied independent of whether a human or software runs them.
No-Swap Regret Prevents Threat Signaling
- No-swap-regret algorithms avoid overreacting and prevent certain kinds of threats.
- In many games they guarantee good equilibria and low-price outcomes when used by all players.