arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2304.09362cs.LG
收藏

未知动态下的长期公平性

Long-Term Fairness with Unknown Dynamics

University of Michigan(密歇根大学);University of California, Santa Cruz(加利福尼亚大学圣克鲁兹分校)

Tongxin Yin, Reilly Raab, Mingyan Liu, Yang Liu

AI总结:

本文在在线强化学习框架下形式化长期公平性,提出一种通过牺牲短期激励来适应未知动态的算法,并证明其在累积损失和公平性违反上具有概率保证,实验结合演化博弈论与真实数据验证了方法的有效性。

摘要:

虽然机器学习可能短视地强化社会不平等,但它也可以被用来动态地寻求公平结果。在本文中,我们在在线强化学习的背景下形式化长期公平性。这种形式化能够容纳动态控制目标,例如推动人口状态中固有的公平性,而这些目标无法纳入静态公平性表述中。我们证明,这一框架允许算法通过牺牲短期激励来适应未知动态,从而将分类器-人口系统驱动到更理想的均衡状态。针对所提出的设置,我们开发了一种算法,该算法借鉴了在线学习领域的最新研究成果。我们证明该算法能够在累积损失和累积公平性违反(作为人口群体之间的统计规律)上同时实现概率上界。我们将所提出的算法与短视分类器的反复重训练作为基线进行比较,并与一个缺乏安全保证的深度强化学习算法进行比较。我们的实验根据演化博弈论对人类群体进行建模,并整合了真实世界数据集。

英文摘要:

While machine learning can myopically reinforce social inequalities, it may also be used to dynamically seek equitable outcomes. In this paper, we formalize long-term fairness in the context of online reinforcement learning. This formulation can accommodate dynamical control objectives, such as driving equity inherent in the state of a population, that cannot be incorporated into static formulations of fairness. We demonstrate that this framing allows an algorithm to adapt to unknown dynamics by sacrificing short-term incentives to drive a classifier-population system towards more desirable equilibria. For the proposed setting, we develop an algorithm that adapts recent work in online learning. We prove that this algorithm achieves simultaneous probabilistic bounds on cumulative loss and cumulative violations of fairness (as statistical regularities between demographic groups). We compare our proposed algorithm to the repeated retraining of myopic classifiers, as a baseline, and to a deep reinforcement learning algorithm that lacks safety guarantees. Our experiments model human populations according to evolutionary game theory and integrate real-world datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑