Management Accounting Section Midyear Meeting

(When) Does Human Intervention in Predictive Analytics Judgments Help Or Hurt?

• In Event: The Future and Past of Management Accounting

Abstract

With the ever-increasing use of predictive analytics in practice, understanding the conditions under which predictive analytic models and human judgment can best work together is critical for the design of effective management control systems. This study utilizes proprietary firm data from an auto-parts retailer to examine questions unaddressed by prior research in augmented decision-making where human judgment jointly operates with predictive analytic models. Specifically, we consider how a model’s predictive ability weakness and an organization’s decision right design choices impact the quality of human judgment relative to predictive analytics alone. Based on theory from research on managerial intuition, we predict and find that human judgment is able to take advantage of model predictive ability weakness associated with greater environmental uncertainty and limited historical data. However, consistent with literature on the detrimental effects of certain forms of accountability, we find that model analysts faced with outcome specific accountability tend to comply with views held by their evaluators at the expense of judgment quality. Thus, we provide empirical evidence of the conditions under which manager intuition can be superior to predictive analytic models but decision right control choices around accountability can affect the value of human judgment in augmented decision-making.

Authors

• Jen Choi, Emory University

• Ewelina Forker, Emory University

• Isabella Grabner, WU Vienna University

• Karen L. Sedatole, Emory University