LtCol Scott Humr is a Ground Supply Officer and Information Systems Officer. He holds a B.A. in Management Information Systems with a minor in Economics from the University of New Mexico, a M.S. in Information Technology Management from the Naval Postgraduate School, and a M.S. in Military Science from Marine Corps University. His military education includes Expeditionary Warfare School, Marine Corps Command and Staff, and Naval Command and Staff at NPS. His operational tours include Headquarters Co, 7th Marine Regiment (HqCo and Reg Supply Officer); 3d Light Armored Reconnaissance Battalion (Bn Supply Officer); I Marine Expeditionary Brigade-Afghanistan (Info Mgmt Officer); I Marine Expeditionary Force (Info Systems Officer); Combined Task Force 51/5th Marine Expeditionary Brigade (Deputy Director of Logistics). He also served at Manpower & Reserve Affairs, Officer Assignments Branch as the Graduate Education Manager.
Humr is a Ph.D. candidate in Information Sciences. His dissertation is on the temporal evolution of trust in Artificial Intelligence - supported decision-making. LtCol Humr is scheduled to graduate in September 2023. He is orginally from Mantua, Ohio. He is married to the former Teresa Jensen and they have three children.
NPS gave me exposure to foundational knowledge in information systems, information security, and knowledge management. This included a broad array of classes ranging from networks, software program management, DOD acquisitions and budgeting process, to business process reengineering. The diverse curriculum prepared me to enter and succeed in a number of new situations for improving my command’s information technology and knowledge management strategies to support the commander’s decision-making cycles.
Most impactful for me is the continuous interactions with NPS faculty, researchers outside of NPS, and colleagues to challenge and refine each other’s ideas. These collegial interactions have been a vital aspect of my personal growth and understanding how science is done within a community.
Cloud computing was a very hot topic when I was here for master’s degree. Since then, it has become ubiquitous. I’ve always tried to follow the latest trends to help understand how these new technologies might provide us a warfighting advantage. As a student at the Marine Corp’s Command and Staff College from 2016-2017, AI and ML were on the ascent once again. I availed myself of the opportunity to write a monograph on the topic and how it may affect the Marine Corps in the near future. This led me to subsequently publish a number of articles on these topics and enroll in some online introductory classes on machine learning. Since I’ve been back at NPS, I wanted to steer my research trajectory toward understanding how we can implement such technologies within a number of different levels of the DOD.
It’s clear throughout the recent history of the DOD that we are not good at adopting new technologies, nor understanding their second and third order effects. This led me to begin looking at research that addressed why organizations have difficulty adopting new technologies and research on a number of famous accidents involving technologically-mediated environments. I found socio-technical systems had a rich history for developing our understanding on how technologies are often overlayed onto existing organizations that have resulted in a number of suboptimal outcomes or billion-dollar failures. These outcomes were the result of not taking into consideration the human and social aspects of these technologies. My sense was the hype behind AI technologies will have suboptimal outcomes if we are not considering how people interact with these technologies. This led me specifically to trust in AI and decision-making in particular because of the number of poor outcomes reported in the research literature on over-trust and under-trust in automated technologies. Furthermore, there have been a number of calls from senior military and civilian DOD leaders on the need for understanding trust in AI, which continues even to today.
There are many to be sure, but several that always come to my mind is AI brittleness and data provenance. It is clear from the popular media and the research literature that most of the AI wonders we see today are from narrow applications for specific contexts. Introducing noise or obscuring images can easily fool AI classification schemes that would not fool most humans to include small children. It is impossible to train machine learning algorithms on every single perturbation that could exist as it would be computationally intractable. To overcome this, we have to develop a number of ensemble methods that can do a better job than single models alone.
Data provenance, or data lineage is important for training machine learning models. Data provenance is important for a number of reasons. First, it is difficult to make decisions or train machines if you do not know how the data was gathered, was it modified in any way, what questions were used to initially elicit the data if it came from end users. This is the classic garbage in, garbage out concept. Second, with ever-increasing content being generated by AI, researchers are raising concerns that machine learning models will be trained on machine learning-generated content which will cause a self-reinforcing loop of circular training leading to suboptimal performance. Programs such as ChatGPT and Midjourney AI produced content could easily affect performance of other AI/ML algorithms training on such content. Therefore, without understanding where our data comes from, decision-making could easily be adversely impacted if we are not careful.
We need to proceed with caution with eyes wide open to the potential problems that can arise with any of these advanced technologies. There is of course no panacea to address these challenges, except for having educated individuals who understand the benefits as well as the pitfalls of these technologies. Having an educated and adaptable DOD workforce who are able to appropriately integrate these technologies into their organizations in a meaningful manner is a foundational step to ensuring we lead off on the right foot. In my estimation, NPS can play a major part in this through its interdisciplinary approach and DOD focused research.
The Naval Postgraduate School is a prime location where mid-career military officers and civilians can come together to share their many challenges and collaborate with a number of research faculty for addressing human-machine teaming (HMT) strategies. For instance, the NPS Joint Field Experimentation Exercises (JIFX) along with other NPS entities such as Center for Autonomous Vehicle Research (CAVR) and Consortium for Robotics and Unmanned Systems Education and Research (CRUSER) can play an outsized roll in leading HMT research and informing strategies. NPS along with its Cooperative Research and Development Agreement (CRADA) partners can further accelerate research in a number of different areas of HMT. Additionally, NPS provides opportunities to nest its latest research with actual fleet experimentation such as Project OVERMATCH to help inform strategies and future research projects.
I do find the use of digital twins (DTs) to be an interesting concept that may go beyond its canonical use for improving maintenance. DTs can be applied to modeling HMTs and human processes of oversight of autonomous vehicles. The concept of DTs can be further expanded to a number of areas to include simulating an entire combat operations center, airfield operations, and decision-making processes. I believe we’ve only begun to scratch the surface on how this concept may be applied.
Conferences such as Sea, Air, & Space and the Armed Forces Communications & Electronics Association provide a venue for seeing some of the latest technologies and building understanding of what is in the art of the possible. These events showcase the industries that are the ultimate producers of the technologies the DOD will leverage to equip the current crop of warfighters with potential to provide feedback to them. The conversations, speaker events, and demonstrations can help generate new ideas and ways of accomplishing our missions we’ve not yet conceived. Events like these have the potential for all stakeholders to understand each other better and inform our military industrial base on what is required on the tactical edge to the garrison back office.
It is easy for many to become so focused on the problems at hand with the technologies we currently possess. This is both natural and necessary; anything else is dreaming that will not accomplish the mission at hand. However, these events provide an opportunity to get above the fray of the immediate for a few days to think how we can better accomplish our mission in the near and long-term futures. We cannot leave the thinking about future technologies and capabilities to others. Decisions we make today affect that future. We often have to make our bets years upfront. Sometimes this works, and other times we get it wrong. To paraphrase former U.S. Secretary of Defence Robert Gates, who said it the best, our record for getting the next war is impeccable; we’ve gotten it wrong 100 percent of the time. Nevertheless, we have a solemn obligation to not only our current Service, but to the men and women of the future. We need to do our best to anticipate the technologies that will best help them accomplish their missions. For these reasons, I believe these events are part and parcel of our professional development as warfighters and are needed to inform our thinking for war fought in any number of domains.