Incentivizing Exploration with Selective Data Disclosure

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Immorlica, Nicole, Mao, Jieming, Slivkins, Aleksandrs, Wu, Zhiwei Steven
Format: Preprint
Published: 2018
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915903499665408
author Immorlica, Nicole
Mao, Jieming
Slivkins, Aleksandrs
Wu, Zhiwei Steven
author_facet Immorlica, Nicole
Mao, Jieming
Slivkins, Aleksandrs
Wu, Zhiwei Steven
contents We propose and design recommendation systems that incentivize efficient exploration. Agents arrive sequentially, choose actions and receive rewards, drawn from fixed but unknown action-specific distributions. The recommendation system presents each agent with actions and rewards from a subsequence of past agents, chosen ex ante. Thus, the agents engage in sequential social learning, moderated by these subsequences. We asymptotically attain optimal regret rate for exploration, using a flexible frequentist behavioral model and mitigating rationality and commitment assumptions inherent in prior work. We suggest three components of effective recommendation systems: independent focus groups, group aggregators, and interlaced information structures.
format Preprint
id arxiv_https___arxiv_org_abs_1811_06026
institution arXiv
publishDate 2018
record_format arxiv
spellingShingle Incentivizing Exploration with Selective Data Disclosure
Immorlica, Nicole
Mao, Jieming
Slivkins, Aleksandrs
Wu, Zhiwei Steven
Computer Science and Game Theory
Data Structures and Algorithms
Machine Learning
We propose and design recommendation systems that incentivize efficient exploration. Agents arrive sequentially, choose actions and receive rewards, drawn from fixed but unknown action-specific distributions. The recommendation system presents each agent with actions and rewards from a subsequence of past agents, chosen ex ante. Thus, the agents engage in sequential social learning, moderated by these subsequences. We asymptotically attain optimal regret rate for exploration, using a flexible frequentist behavioral model and mitigating rationality and commitment assumptions inherent in prior work. We suggest three components of effective recommendation systems: independent focus groups, group aggregators, and interlaced information structures.
title Incentivizing Exploration with Selective Data Disclosure
topic Computer Science and Game Theory
Data Structures and Algorithms
Machine Learning
url https://arxiv.org/abs/1811.06026