Benjamin A. Clegg, PhD
Department of Psychology
Colorado State University
Research
OVERVIEW
My research addresses a number of questions related to human performance. This work has included both applied and basic perspectives. My interests span topics in applied cognitive psychology, human factors, and cognitive engineering. One central research interest can be characterized as the areas of skill acquisition and training. A core current component of this research has been studies of the impact of automation, including looking at how the presence of automated systems changes performance, and in particular, underlying learning. I have also extended my work into aspects of skilled performance - issues like the use of automation and supervisory control, situation awareness, workload, and enhancing decision making. I have been an author on over 70 articles, chapters, and papers, have authorship on more than 90 presentations and posters at conferences and scientific meetings, and have co-authored 11 major technical reports. In addition I have been centrally involved in 11 externally funded projects while at Colorado State University, including work funded by NSF, ONR, IARPA, NASA, ARO, and DARPA.
EXAMPLES OF WORK FROM MY LAB
Here are some presentations of work from the lab that were delivered at HFES in 2020.
HFES 2020 Differences in Variability
AUTOMATION
My interests in skill acquisition have led me to examine learning in systems featuring automation. Training with highly automated systems is becoming increasingly common, but comparatively little has been known about the impact of automation on the nature of what is being learned (Gutzwiller, Clegg, & Blitch, 2013). My research on this topic began with a project I conducted along with Professor Neville Moray and Daniel Rodriguez. We looked at the effects of automation of systems on the learning of operators (Moray, Rodriguez, & Clegg, 2005). The underlying notion is that there are levels of automation, ranging from situations in which almost all responsibility lies with the human to fully autonomous automated systems. Our work attempts to operationalize, in a microworld simulation, a variety of levels and modes of human-machine interaction. The project represents the first stages of a program to compare, across training, the impact of types of automated assistance on variety of aspects of performance (such as productivity, trust, operator self-confidence, and fault management).
We continued to explore these issues with funding from a MURI grant from the Army Research Office, examining whether individual differences influence the effectiveness of training under different levels of automation. This project showed for first time evidence of aptitude/automation interactions in learning (Clegg & Heggestad, 2010; Clegg, Heggestad, & Durrance Blalock, 2010; Heggestad, Clegg, Goh, & Gutzwiller, 2012). Our core finding has been that although automation is effectively supporting performance early in training, especially for lower aptitude individuals most in need of that support, it is impairing the underlying learning that is occurring. Essentially while those being trained with automation present are doing better, they are often actually learning less.
In my lab we have subsequently followed up these findings, demonstrating the same type of costs of training with automation as seen in our previous findings, but within an operationally relevant, unmanned air systems simulator, that features very different task dynamics and training (Blitch & Clegg, 2011). We extended this work to explore the question of whether automation can enhance learning to a situation featuring pairs of operators jointly controlling unmanned vehicles (Blalock & Clegg, 2011). We have also examined how training individuals engaged in control over multiple unmanned vehicles might differ from control over a single vehicle (Gutzwiller & Clegg, 2012).
For a NASA grant, we developed a model of performance across failures of automation (Wickens, Clegg, Vieane, & Sebok, 2015), and attempting to predict performance (Sebok, Wickens, Clegg, & Sargent, 2014). This project included a novel exploration of choices in voluntary task switching (Gutzwiller, Wickens, & Clegg, 2014), including task selection that occurs within the high workload environment following an unexpected automation failure (Wickens, Vieane, Clegg, Sebok, Janes, 2015). Our research on task switching between a process control and robotic arm use (Wickens, Gutzwiller, Vieane, Clegg, Sebok, & Janes, 2016) won the Jerome H. Ely Human Factors Article Award for the best article in the Human Factors Journal.
You can see some of the publications on ResearchGate in our Training with Automation project, and our Strategic Task Switching project
PREDICTIONS IN SPATIAL UNCERTAINTY
We have been conducting work looking at people's understanding of predictions in spatial uncertainty, and trying to identify methods to better support performance. When determining where and when a hurricane might make landfall, a downed plane might have crashed, a potential rendezvous location to supply a ship, or the future potential position of a submarine, human decision makers must make predictions about the uncertain trajectory of an object. This research has been examining performance in spatial predictions under uncertainty , especially focusing on people's struggles to grasp the possible variability.
You can see some of the publications on ResearchGate in our Spatial predictions in uncertainty project.
ENHANCING DECISION MAKING
I have been involved in 2 major projects funded by IARPA that have examined methods to support and improve aspects of decision making and reasoning. Developed within the IARPA Sirius program, CYCLES are two standalone computer video games that successfully train the mitigation of cognitive biases. Cognitive biases are systematic errors that result from reliance on heuristics in decision-making. One game tackles fundamental attribution error, confirmation bias, and bias blind spot. A second game tackles anchoring, projection, and representativeness. We frame this work not as debiasing (given that cognitive biases will still be present), but rather as bias mitigation. Our evidence shows substantial improvements in both the recognition and identification of these biases, and crucial evidence of reduced impact of the biases that is durable, following relatively brief training (Clegg et al., 2014; Clegg et al., 2015)
Within the IARPA CREATE program, we developed TRACE (Trackable Reasoning and Analysis for Collaboration and Evaluation) that aims at improve reasoning and intelligence analysis through the development of a web-based application that will leverage the use of structured techniques, crowdsourcing and smart nudging to enhance analysts' problem-solving abilities and foster creative thinking.
You can see some of the publications on ResearchGate in our CYCLES - Serious Computer game to mitigate cognitive biases project, and our TRACE - platform to improve reasoning and argumentation project.
TRAFFIC & TRANSPORTATION
My work on traffic and transportation has included a major on-going collaboration with Professor John Groeger. Within this research I have looked at a number of fundamental cognitive questions in the context of applied settings. A central focus of the work we have been conducting concerns the acquisition of driving skills. Within this general framework, we have explored a number of issues, and have offered the first evidence of the possible learning curve for driving performance (Groeger & Clegg, 2007). One major question we have been examining is whether any level of driving performance should be considered an automatic process, and thus not susceptible to interference from other cognitive demands. In our research, we have been exploring whether driving truly shows the hallmarks of automatic performance (Groeger & Clegg, 1997). Our findings suggest it is a mistake to assume driving is an automatic process.
Overall we have been attempting to use theories from psychology to gain insight that can help to provide practical guidelines and advice for learner drivers, driving instructors and those involved in producing government regulations (Groeger & Clegg, 2000). Another practical question we have examined is how to convey driving directions such that an individual learns them most effectively (Morett, Clegg, Blalock, & Mong, 2008). In addition to these types of real-world implications, I believe that data from such applied settings provides important tests and validation for psychological theories, as well as highlighting limitations that may not always be apparent in non-applied research.
Another paper (Groeger, Clegg, & O'Shea, 2005) illustrates how the signaling system most widely employed on British railways is flawed in terms of its design. Most of the signal states, indicating the status of the track ahead, can be readily identified based on only a single dimension of information. However, one condition requires the integration of two dimensions, and that conjunction of features requires attentional resources - and hence takes individuals longer to process. Unfortunately, the condition is also the one that is most critical when a driver is required to bring the train to a halt. A number of catastrophic train crashes in Britain have been, at least in part, blamed on driver errors in passing warning signals. This new insight into the problem from application of theoretical knowledge from cognitive psychology has the potential to improve performance, and as a result might even contribute to saving lives.
Research with students in my lab has also begun to offer insights into the nature of distraction in driving from cellphone conversations (Durrance Blalock, Sawyer, Kiken, & Clegg, 2009), text messaging (Sawyer & Clegg, 2010), and the potential future introduction of heads-up displays for navigation (Murchison, Mong, & Clegg, 2010).
Our predictions in spatial uncertainty, described above, also includes aspects looking at issues related to ship navigation and decision making. These include factors that influence determining whether a collision may happen (Wickens, Williams, Clegg, & Smith, 2020) and how to rendezvous (Spahr,Wickens, Clegg,Smith, Vijayaragavan, 2019).
SEQUENCING
One crucial underlying aspect of skill acquisition is the learning of sequences. People learn and use sequential information in a variety of complex everyday tasks: sequencing sounds in making sentences, sequencing movements in typing or playing a musical instrument, and sequencing actions in driving a car. The performance of every type of skill involves learning sequences of information and/or actions. I have co-authored (as senior author, and first author respectively) two major review papers on the serial reaction time task (Abrahamse, Jimenez, Verwey, & Clegg, 2010; Clegg, DiGirolamo, & Keele, 1998), a paradigm that has become one of the most prevalent methods to examine sequence learning. Across a variety of my empirical and theoretical contributions to the field, I have been trying to make the case that multiple forms of representation of sequence learning are possible.
My research (Clegg, 2005; Francis, Schmidt, Carr, & Clegg, 2009; O'Shea & Clegg, 2005) shows that while responses may be a crucial component in learning of some forms of sequential information, other forms of information can be learned, and that even within responses multiple dimensions for learning exist (Richard, Clegg, & Seger, 2009). This issue has parallels in several other areas of research. For example, attempts to determine the loci of repetition effects, practice effects, and perceptual learning.
Along with Professor Willem Verwey, I have also examined changes in the nature of the representation of sequences with increased practice (Verwey & Clegg, 2005; Richard, Clegg, Verwey & Seger, 2008). Research suggests the way the brain represents sequences changes between initial acquisition and later performance. This plasticity may be an important case of a more general phenomenon. It is widely recognized in general work on skill learning that early acquisition is a poor indicator of long-term performance. Such research on sequence learning may allow us to move beyond a simple observation of that fact, and towards an understanding of why that should be the case.
My work has also looked at other types of sequential performance, including sequences of real and imagined movements (Clegg, Wood, & Bugg, 2004), and working memory for sequential versus simultaneously presented visual information (Durrance Blalock & Clegg, 2009).
You can see some of the publications on ResearchGate in our Sequence Learning project