Kostas Triantis, John Lawrence Professor at Virginia Tech, Granted $2 Million by NSF for Conducting Research on the Relationship of Cognitive Biases and Automation

Kostas Triantis, a professor in the Grado Department of Industrial and Systems Engineering, has received a four-year, $2 million National Science Foundation grant to explore how cognitive biases influence trust in automation and decisions to delegate tasks to automated technologies.

A brief description of the project is provided:

Every effective railway operation has a strong, well-managed control center. Also known as the “nerve centers” of the rail network, control centers serve as hubs of the newest rail technology and help ensure the smooth operation of train services, provide passengers with information, and resolve safety incidents. 

At the heart of every control center sits the interface between human operators and technology. The general public may believe that technology alone improves efficiency, quality, safety, and cost. However, few people consider that these same technologies may also introduce errors and adverse events.

In the earliest years of the railway industry, train drivers relied heavily on their own perceptions and judgment to perform their jobs safely. Over time, those methods were augmented with technology designed to work with a variety of modern driver systems. Today’s train drivers encounter more automated technologies within the cab than ever before.

We still don’t fully understand the impacts of technology and how decisions are best made between human operators and increasingly autonomous technologies to carry out safety-critical tasks, but Kostas Triantis, a professor in the Grado Department of Industrial and Systems Engineering in the College of Engineering, will use a National Science Foundation (NSF) award to find out. For the project, researchers from multiple universities spanning the disciplines of decision theory, organizational theory, systems engineering, and human factors engineering will collaborate with a network of European (INFRABEL, Pro Rail, Network Rail, European Agency for Railways) and U.S. infrastructure (Union Pacific) providers and federal research centers (VOLPE Center).

“This research should deliver highly relevant and scientifically based insights and tools that will further improve safety levels, staff wellbeing, and the smart use of automation in control rooms,” said Bart Roets, principal engineer at Infrabel. “It will also help in preparing (and anticipating) for a future where automation plays an even more important role than it does today. I also expect new and innovative types of control room datasets and metrics to emerge from the interactions between the researchers, our data engineers, and other project collaborators.”

Dr. Roets who received his Ph.D. from Ghent University in Belgium serves as the intermediary between the research group and INFRABEL. He organizes the provision of new data that the research team analyzes, represents INFRABEL at the bi-weekly research meetings, serves on Ph.D. research committees as a committee member, contributes to various research journal papers that the research team works on, and is the intermediary between INFRABEL and other European infrastructure providers (ProRail, Network Rail, and others). He is a critical and central figure of the research team.

With this four-year, $2 million NSF grant funded under the Leading Engineering for America’s Prosperity, Health, and Infrastructure (LEAP-HI) program, Triantis will work with a transdisciplinary team of engineers and social scientists to explore how cognitive biases influence trust in automation and decisions to delegate tasks to automated technologies.

A railway controller works at INFRABEL in Belgium. Photo courtesy of Bart Roets for INFRABEL

In the airline industry, air traffic flow management and collaborative decision-making drive safe aviation. A number of decisions have been related to the ability of pilots to interact effectively with automated systems. When circumstances dictate, the pilots and operators in general have to make decisions as to when to intervene and use systems manually. When they are using automated systems, their monitoring workload becomes larger, which then increases the probability of making an error. For example, Boeing was under intense scrutiny after its Max jet was involved in two fatal accidents in 2018 and 2019. Pilots were not trained to handle the new automated system, known as MCAS. Because the pilots were not trained on the new technology, they could not override the system, resulting in a crash and loss of life.

Based on interviews with engineers, operators, and managers, we know that during peak times of activity, controllers often turn off automated systems because regulating the transportation network becomes too complicated. Even though controllers are trained to deal with a variety of scenarios, different controllers respond to scenarios in a variety of ways. Younger controllers tend to use automated systems more than their senior colleagues. The companies (transportation agencies), in order to achieve cost efficiency, would like to use technology more. However, decisions must balance multiple factors, including economic feasibility, safety, and workload demands placed on operators.

Similar situations abound in the healthcare sector where managing human-technology interaction is critical. For example, in operating rooms, doctors, nurses, and other stakeholders need situation awareness and an acceptable level of mental workload to effectively manage automated systems. Such systems have the potential to prevent many potentially harmful incidents, but poorly designed automated systems may cause more incidents instead of preventing them, especially when automation reduces the human’s role to being primarily supervisory. Building off the LEAP-HI grant, Triantis has reached out to secure funds for a seed grant to further explore these issues in collaboration with MEDSTAR in Northern Virginia and Carilion in Roanoke.

For the past four decades since Triantis’ dissertation, the researcher has sought to marry different disciplines for a more holistic approach to societal problems. 

“Our focus has been more and more multidisciplinary,” said Triantis. “Not just with evaluating the impacts of technological investments, but also with the well-known challenge of the integration of social networks and their behaviors with technological change. We are trying to see the relevance of specific theoretical paradigms and different modeling approaches to address complex societal problems. As a Ph.D. student, there was something that I couldn’t really articulate at the time, this idea of doing multidisciplinary research, and how different approaches pertain to different disciplines. How some of the theory actually relates to reality.”

The project will support outreach activities to stimulate interest in the sociotechnical domain area for students going into STEM-related fields. We will integrate the research findings into the curricula for our undergraduate and graduate students. Included are outreach activities to stimulate interest in LEAP HI-related disciplines for students going into STEM-related fields. The overall goal is to reach out to a diverse group of students on the human-automation interface and the modeling of socio-technical systems. The principal elements of the plan include the following items: (1) Course development and implementation: We design or extend courses based on our research results. We have already designed and are offering a course in Socio-Technical Systems. (2) Recruitment and Retention Plan-Reaching out to Women and Minority Students: We employ practices developed at VT, Georgetown, and BYU to recruit and retain students from underrepresented minorities including women. (3) Mentoring Plan for Professional Development: CEED has implemented several successful mentoring plans that target under-represented students (African American, Hispanic/Latino, and women), and will guide us in implementing these in our project. (4) Outreach Programs: We will increase public-at-large awareness and K-12 student interest in LEAP HI-related disciplines and STEM through VT CEED pre-college activities. In particular, we believe students in grades 7-12 will be interested in the simulator decision-making challenges, the influence of Distributed Situational Awareness, and fatigue on decision-making. Fatigue and decision-making are critically important topics to outreach for grades 7-12 students. (5) International Research Internship and Cohort Experience: We will send a group consisting of the postdoctoral researcher and two graduate students as part of an internship program 2 to do summer research work at INFRABEL.

Plans for Disseminating the Research: Connection to Policymakers: The direct involvement of our collaborators will create a “core” of infrastructure providers that will benefit directly from the results of the research. We will expand this “core” to include other collaborators such as US DOT-FTA/FRA, NIOSH, and the International Union of Railways and will identify key technology-related workshops, conferences, and trade shows where we would present research findings and most importantly implementation strategies. The insights obtained in this work will inform the design and policy choices of both infrastructure providers (such as railroads, airlines, trucking, maritime, and pipeline controllers) and regulators with authority over infrastructure systems (such as the FRA, FAA, US Coast Guard and FTA in the US). Our framework provides critical input for the engineering community, specifically for safety regulators working with automation-driven infrastructure systems.