Learning Personalized Decision Support Policies
Learning Personalized Decision Support Policies
复制标题
DOI：
10.48550/arxiv.2304.06701
复制
发表时间：
2023-04
期刊：
影响因子：
--
通讯作者：
Umang Bhatt;Valerie Chen;Katherine M. Collins;Parameswaran Kamalaruban;Emma Kallina;Adrian Weller;Ameet Talwalkar
中科院分区：
文献类型：
--
作者：
Umang Bhatt;Valerie Chen;Katherine M. Collins;Parameswaran Kamalaruban;Emma Kallina;Adrian Weller;Ameet Talwalkar
个人决策者可能会受益于不同形式的支持，以改善决策结果，但当每种形式的支持将产生更好的结果？在这项工作中，我们认为，个性化访问决策支持工具可以是一种有效的机制，用于实例化人工智能辅助的适当使用。具体来说，我们提出了学习决策支持政策的一般问题，对于给定的输入，选择哪种形式的支持提供给决策者，我们最初没有先验信息。我们开发了$\texttt{Modiste}$，一个交互式的工具来学习个性化的决策支持策略。$\texttt{Modiste}$利用随机上下文Bandit技术为每个决策者个性化决策支持策略，并支持对多目标设置的扩展，以考虑支持成本等辅助目标。我们发现，个性化的政策优于离线的政策，并在成本意识的设置，降低所产生的成本，最小的性能退化。我们的实验包括各种现实形式的支持（例如，专家共识和来自大型语言模型的预测）。我们的人类受试者实验验证了我们的计算实验，表明个性化可以在实践中为与$\texttt{Modiste}$交互的真实的用户带来好处。
Individual human decision-makers may benefit from different forms of support to improve decision outcomes, but when each form of support will yield better outcomes? In this work, we posit that personalizing access to decision support tools can be an effective mechanism for instantiating the appropriate use of AI assistance. Specifically, we propose the general problem of learning a decision support policy that, for a given input, chooses which form of support to provide to decision-makers for whom we initially have no prior information. We develop $\texttt{Modiste}$, an interactive tool to learn personalized decision support policies. $\texttt{Modiste}$ leverages stochastic contextual bandit techniques to personalize a decision support policy for each decision-maker and supports extensions to the multi-objective setting to account for auxiliary objectives like the cost of support. We find that personalized policies outperform offline policies, and, in the cost-aware setting, reduce the incurred cost with minimal degradation to performance. Our experiments include various realistic forms of support (e.g., expert consensus and predictions from a large language model) on vision and language tasks. Our human subject experiments validate our computational experiments, demonstrating that personalization can yield benefits in practice for real users, who interact with $\texttt{Modiste}$.