Doctoral Dissertation Research in DRMS: Integrating Managerial Insight and Optimal Algorithms

DRMS 博士论文研究：整合管理洞察力和最优算法

• 批准号: 1729837
• 负责人: Elena Katok
• 金额: $ 2.33万
• 依托单位: University of Texas at Dallas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-05-15 至 2018-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1729837&HistoricalAwards=false
• 关键词: Doctoral; Dissertation; Research; DRMS; Integrating

An extensive body of experimental work shows that human decision-makers routinely make sub-optimal choices when faced with uncertainty in the laboratory. Since these deviations from rational behavior cut into firm profits, one might conclude that employees who routinely encounter such choices should be replaced with software that is programmed to automatically implement optimal strategy. On the other hand, if human managers have ?managerial insight? that can resolve some of the uncertainty about future demand, then they should be considered valuable assets rather than liabilities. Humans have a broad contextual awareness of the world, but their computational abilities are limited and error-prone. Computers are relatively unaware, but they are unerring and have nearly unlimited processing capability. This Doctoral Dissertation Research Improvement Grant, the co-PI develops a model to examine how a business might best take advantage of human managers? insights while scaffolding their decision-making with technology. The project will explore whether it is better to have human managers adjust recommendations made by an automated system or whether managerial insight should be translated into a machine-usable format for automated processing with the software holding ultimate decision rights. Implications of these findings to issues like supply chain management will be explored.The project involves an experimental methodology where participants in a ?newsvendor problem? setting are endowed with managerial insight in the form of a noisy signal about upcoming demand. With this additional piece of information, a manager computing the optimal insight-informed ordering policy should outperform the optimal policy absent this information. However, in the lab, human subjects fail to improve profitability when making these decisions because their superior demand information is undercut by their inferior computational ability. The co-Pi will explore whether, by disaggregating the forecasting task (where managerial insight can improve operations) from the order quantity selection (which can be mechanically computed once demand information is known) can lead to truly optimal responses.

大量实验工作表明，人类决策者在实验室面临不确定性时通常会做出次优选择。由于这些偏离理性行为的行为会影响公司的利润，人们可能会得出这样的结论：经常遇到此类选择的员工应该被编程为自动实施最佳策略的软件所取代。另一方面，人类管理者是否具有“管理洞察力”？如果能够解决未来需求的一些不确定性，那么它们应该被视为有价值的资产而不是负债。人类对世界有广泛的情境意识，但他们的计算能力有限且容易出错。计算机相对来说没有感知能力，但它们不会出错，并且具有几乎无限的处理能力。在这项博士论文研究改进补助金中，联合首席研究员开发了一个模型来研究企业如何最好地利用人力管理者？洞察力，同时利用技术支撑他们的决策。 该项目将探讨是否让人类管理者调整自动化系统提出的建议更好，或者是否应将管理见解转化为机器可用的格式，以便由拥有最终决策权的软件进行自动化处理。将探讨这些发现对供应链管理等问题的影响。该项目涉及一种实验方法，其中参与者面临“报童问题”。环境被赋予了管理洞察力，其形式是关于即将到来的需求的嘈杂信号。有了这条额外的信息，经理计算最佳的洞察通知订购策略应该优于没有此信息的最佳策略。 然而，在实验室中，人类受试者在做出这些决策时无法提高盈利能力，因为他们优越的需求信息被他们较差的计算能力削弱了。 co-Pi 将探讨通过将预测任务（管理洞察力可以改进运营）与订单数量选择（一旦了解需求信息就可以机械计算）分开，是否可以带来真正的最佳响应。