Machine Learning in Air Force Human Resource Management

Research Questions

  1. What ML techniques are applicable to HRM functions?
  2. What HRM objectives could ML systems satisfy?
  3. What criteria must ML systems meet to be considered technically feasible?
  4. What challenges will the DAF face in building a portfolio of ML projects for HRM?

The Department of the Air Force (DAF) is working to develop and field artificial intelligence and machine learning (ML) systems for mission areas and support functions, including human resource management (HRM). Recent developments have improved the access that organizations have to data and analytic tools, opening a wide range of possible ML projects that they could pursue.

Given resource limitations, decisionmakers must choose which projects to pursue among many promising options. The DAF needs a framework to evaluate the business value, feasibility, and complexity of proposed projects.

To understand how the DAF can form a balanced portfolio of ML projects for HRM, the authors reviewed how private-sector organizations evaluate and select such projects. From the review, they arrived at a five-step framework. Broadly, the framework involves evaluating the business value, technical feasibility, and implementation complexity of possible ML projects and forming a portfolio from these evaluations. Each of these steps draws on multiple predefined criteria, which may be assessed using qualitative or quantitative methods. The authors demonstrate steps of the framework using 19 use cases for applying ML throughout the DAF HRM life cycle.

Notably, this approach does not purport to find the best approach to a business problem. It finds a potentially useful ML approach to addressing a business problem but does not provide a full analysis of alternative solutions, such as non-ML approaches, to address the problem.

Key Findings

  • ML techniques such as supervised and unsupervised learning, optimization, language processing, and reinforcement learning are applicable to many HRM functions.
  • ML systems can satisfy four HRM objectives: process improvement, performance improvement, enhancing service member opportunities, and enhancing service member motivation. The use cases considered overwhelmingly support process improvement.
  • To be technically feasible, ML systems must have measurable outcomes, relevant inputs, sufficient data, and a suitable algorithmic approach. In the use cases considered here, data sufficiency was the most common bounding constraint. Additionally, use cases that involve process improvement are more technically feasible.
  • The DAF must balance multiple objectives as it builds a portfolio of ML projects for HRM. This includes the relative weight of effort supporting core business functions versus more-transformational initiatives. Another consideration is the types of implementation complexity that different initiatives entail.


  • To maximize return on investment, the DAF must use a systematic process to evaluate ML projects for HRM and to build a balanced portfolio.
  • The DAF should shape an innovation portfolio that includes low-risk/low-reward projects along with higher-risk but potentially transformative ones.
  • The DAF should develop a common ML ecosystem to enable rapid creation, comparison, and reuse of ML pipelines, models, and U.S. Department of Defense datasets.
  • To enable applications of ML to HRM, the DAF must continue to invest in data infrastructure and outcome definitions.
  • As the DAF evaluates projects, it must consider the types of technical and nontechnical complexity they entail.

Research conducted by

This research was prepared for the Department of the Air Force and conducted within the Workforce, Development, and Health Program of RAND Project AIR FORCE.

This report is part of the RAND research report series. RAND reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for research quality and objectivity.

This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited; linking directly to this product page is encouraged. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial purposes. For information on reprint and reuse permissions, please visit

RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors.


Download eBook for Free

Format File Size Notes
PDF file 1.7 MB

Use Adobe Acrobat Reader version 10 or higher for the best experience.

© 版权声明
点赞14 分享