Targeted Automation and Sustaining Human-Algorithm Learning

42 Pages Posted: 7 Dec 2022 Last revised: 16 Jan 2024

See all articles by Christina Imdahl

Christina Imdahl

Eindhoven University of Technology

William Schmidt

Emory University - Information Systems and Operations Management

Kai Hoberg

Kuehne Logistics University

Date Written: December 3, 2022

Abstract

In many decision processes, a planner must review and optionally adjust recommendations (decision instances) from a decision support system (DSS). When the DSS is well-tuned to its task, adjustments by a planner can be rare and may even degrade the DSS's performance. Targeted automation, i.e. automation per decision-instance, could address these inefficiencies by predicting when a planner will improve an individual \emph{decision instance} and automating those instances that they will not improve. As more instances are automated, however, there are fewer instances that receive planner input, which starves the prediction model of the observations it needs for retraining. To maintain predictive performance, we must overcome the loss in the model's ability to learn from a planner's decisions over time.
Using four years of procurement ordering data from our research partner, a large materials handling equipment manufacturer, we develop and train an elastic net classifier that predicts individual instances in which a planner will improve a DSS-generated procurement order quantity decision.
We demonstrate how to mitigate the performance erosion that automation engenders by structuring the selection of the ML model's classification threshold as a newsvendor problem that balances the costs and benefits of human reviews, including the value of learning.
Depending on the value of an improvement, our approach automates around 84\% of all DSS recommendations while retaining three times more improvements than random automation.
Our research contributes to a broader debate on the allocation of decision authority between humans and algorithms, and creates a framework for targeted automation in an operational setting that balances the benefits of algorithmic learning and automation.

Keywords: Automation, machine learning, inventory management

Suggested Citation

Imdahl, Christina and Schmidt, William and Hoberg, Kai, Targeted Automation and Sustaining Human-Algorithm Learning (December 3, 2022). Available at SSRN: https://ssrn.com/abstract=4292438 or http://dx.doi.org/10.2139/ssrn.4292438

Christina Imdahl

Eindhoven University of Technology ( email )

PO Box 513
Eindhoven, 5600 MB
Netherlands

William Schmidt (Contact Author)

Emory University - Information Systems and Operations Management ( email )

1300 Clifton Road
Atlanta, GA 30322
United States

Kai Hoberg

Kuehne Logistics University ( email )

Grosser Grasbrook 17
Hamburg, 20457
Germany

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
115
Abstract Views
377
Rank
446,449
PlumX Metrics