Incentive Mechanisms for Strategic Classification and Regression Problems
Incentive Mechanisms for Strategic Classification and Regression Problems
复制标题
战略分类与回归问题的激励机制
DOI：
10.1145/3490486.3538300
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Liu, Mingyan
中科院分区：
文献类型：
--
作者：
Jin, Kun;Zhang, Xueru;Khalili, Mohammad Mahdi;Naghizadeh, Parinaz;Liu, Mingyan
在一个策略分类与回归问题中，我们研究了一类有效防止作弊的激励机制设计。传统的战略分类或回归问题被建模为Stackelberg博弈，或者是分类器设计者（委托人）和服从分类器决策的个人（代理人）之间的委托-代理问题，这些人可能来自不同的人口统计群体。前者受益于其决策的准确性，而后者可能有动机让算法做出有利但错误的决策。虽然之前的工作倾向于关注如何设计一种算法，使其对这种战略操作更具鲁棒性，但本研究关注的是另一种选择，即设计激励机制，以塑造代理的效用，并诱导真正提高其技能的努力，从而使Stackelberg博弈中的双方受益。具体来说，委托人和机制提供者（也可以是委托人本身）在第一阶段一起行动，发布并承诺分类器和激励机制。代理是（同时）第二推动者，对已发布的分类器和激励机制做出最佳响应。当智能体的策略行为仅仅改变其可观察特征时，就会损害算法的性能。但是，如果该动作导致agent真实标签的改善，则不仅可以帮助agent获得更好的决策结果，还可以保持算法的性能。我们研究了补贴机制如何诱导改善行动，积极影响一些社会福利指标，如代理人的整体技能水平（效率）和不同人口群体之间的正阳性率或真阳性率差异（公平）。
We study the design of a class of incentive mechanisms that can effectively prevent cheating in a strategic classification and regression problem. A conventional strategic classification or regression problem is modeled as a Stackelberg game, or a principal-agent problem between the designer of a classifier (the principal) and individuals subject to the classifier's decisions (the agents), potentially from different demographic groups. The former benefits from the accuracy of its decisions, whereas the latter may have an incentive to game the algorithm into making favorable but erroneous decisions. While prior works tend to focus on how to design an algorithm to be more robust to such strategic maneuvering, this study focuses on an alternative, which is to design incentive mechanisms to shape the utilities of the agents and induce effort that genuinely improves their skills, which in turn benefits both parties in the Stackelberg game. Specifically, the principal and the mechanism provider (which could also be the principal itself) move together in the first stage, publishing and committing to a classifier and an incentive mechanism. The agents are (simultaneous) second movers and best respond to the published classifier and incentive mechanism. When an agent's strategic action merely changes its observable features, it hurts the performance of the algorithm. However, if the action leads to improvement in the agent's true label, it not only helps the agent achieve better decision outcomes, but also preserves the performance of the algorithm. We study how a subsidy mechanism can induce improvement actions, positively impact a number of social well-being metrics, such as the overall skill levels of the agents (efficiency) and positive or true positive rate differences between different demographic groups (fairness).
登录
查看更多内容
DOI：
--
发表时间：
2020
期刊：
ICML 2020
影响因子：
--
作者：
Miller, John;Milli, Smitha;Hardt, Moritz
通讯作者：
Hardt, Moritz
DOI：
10.1145/2020408.2020495
发表时间：
2011-08
期刊：
--
影响因子：
--
作者：
Michael Brückner;T. Scheffer
通讯作者：
Michael Brückner;T. Scheffer
DOI：
--
发表时间：
2019
期刊：
In Proceedings of ACM FAT*
影响因子：
--
作者：
Milli, Smitha;Miller, John;Dragan, Anca;Hardt, Moritz
通讯作者：
Hardt, Moritz
具有战略机器学习的网络游戏
DOI：
--
发表时间：
2021
期刊：
Decision and Game Theory for Security
影响因子：
--
作者：
Kun Jin;Tongxin Yin;Charles A. Kamhoua;Mingyan D. Liu
通讯作者：
Mingyan D. Liu
DOI：
10.4230/lipics.forc.2020.9
发表时间：
2020-05
期刊：
ArXiv
影响因子：
--
作者：
M. Braverman;Sumegha Garg
通讯作者：
M. Braverman;Sumegha Garg