AI-Assisted Search: Balancing Communication and Recommendation Set Size
A new paper models the interaction between users and AI recommendation systems, focusing on the trade-off between communication cost and search cost. The user sends a costly, noisy message about preferences; the AI Bayesian agent forms a posterior belief and recommends a set of products. The optimal set size balances expected utility against search cost. The model uses mutual information to quantify both costs. Products and preferences exist in d-dimensional space. The paper is available on arXiv.
Key facts
- arXiv:2605.23944
- Models user-AI recommendation interaction
- User sends costly, noisy preference message
- AI acts as Bayesian agent
- Determines optimal recommendation set size
- Accounts for search cost
- Uses mutual information cost functions
- Products and preferences in d-dimensional space
Entities
Institutions
- arXiv