Reading List: EconCS Models of LLMs
I am excited to learn how to build mathematical and economical models for LLMs / LLM agents. I’ll update the list as I read this line of research!
Agentic recommendation and conversation
| Paper | Authors | Publication status | Experiments | My Summary |
|---|---|---|---|---|
| Agentic Markets: Equilibrium Effects of Improving Consumer Search | Brendan Lucier, Nicole Immorlica, Markus Mobius, Aleksandrs Slivkins, Daniel Goldstein, Jake Hofman, Sonia Jaffe, David Rothschild | arXiv, 2026 | None | Consumers perform sequential, Pandora-like search while an agent can lower search costs or improve information. |
| Right-Sizing Communication and Recommendation Set Size in AI-Assisted Search | Jing Dong, Prakirt Raj Jhunjhunwala, Yash Kanoria | arXiv,2026 | None | Joint optimization of mutual-information and search cost, determining how much a user should tell an assistant and how many items it should return. |
| A Solicit-Then-Suggest Model of Agentic Purchasing | Shengyu Cao, Ming Hu | arXiv, 2026 | Synthetic | An agent chooses how many preference-solicitation rounds and how many products to recommend. Water-filling optimal strategy decomposition. |
| How Much Should a Conversational Recommender System Converse? | Akshit Kumar, Vahideh Manshadi, Akhilesh Tumu | EC 2026 | Synthetic | Sequential questions improve match quality but impose communication and abandonment costs. |
| Controlling the Conversation | Martino Banchio, Bing Liu, Andres Perlroth | EC 2026 | None | A conversational assistant selects sequential partition queries. Prior-free query design minimizes user regret when the user may only choose from a queried subset. |
Human-AI Collaboration and Productivity
[TBD…]