Sandra P. Marshall
Department of Psychology
San Diego State University
In the mid-1980's, I began developing a schema theory to explain how people organize and use their knowledge stored in long term memory. The essential premise is that individuals have distinct types of knowledge that are linked in memory and that are called on by the individuals to make sense of their environments. Each bundled set of knowledge is called a schema. The initial theory was developed and studied in the context of arithmetic story problems. My book, Schemas in Problem Solving (Cambridge University Press, 1995), describes the basic features of the theory, the experimental research, and the resulting cognitive models.
I first studied the schema knowledge used by students solving story problems and then investigated the schema knowledge used by officers making tactical decisions. I have identified four types of knowledge needed in an efficient schema: identification knowledge, elaboration knowledge, planning knowledge, and execution knowledge. The importance of recognizing these different types of knowledge is that each can now be uniquely identified and modeled.
My current work focuses on two areas: investigating a specific component of the schema model in the context of tactical decision making and exploring the ways that eye tracking data can improve cognitive models. The Cognitive Science Program of the Office of Naval Research supports both lines of research.
Identification knowledge. In tactical situations, Navy officers respond to a large number of events that unfold rapidly and often unexpectedly. Time constraints are tight, and officers need to identify almost instantaneously which aspects of the situation demand their immediate attention and which do not. This process of recognition corresponds to the first schema component: identification knowledge. Undoubtedly, a number of schemas are used by officers in dealing with all aspects of a tactical situation over an extended period of time. This research focuses on just one--the basic schema of situation awareness. Essentially, this schema is needed as an overall control mechanism and is used repeatedly in tactical settings.
The underlying issue is how the officers gather and use information that is available to them only on a computer display (or through the communication channels of their headsets). This situation is not unique to Navy officers; many individuals in today's technological world have similar constraints. The research objective is to identify the critical patterns of elements on the computer display and to determine the extent to which an individual's performance can be explained by them.
From studies with two distinct populations of officers--senior officers with considerable experience at sea and junior officers in training--we have developed a model of performance in tactical situations. The model is a neural network whose inputs are simple features describing individual aircraft contacts that appear on the display (e.g., fast, high, inbound, hostile). The outputs are the observed behaviors of the officers (e.g., ignore the contact altogether, monitor it for further attention at a later time, give full attention now).
An important outcome of this research is that we have determined that the performance of all teams can be modeled with a single neural network (see Figure 1) having the same overall structure with respect to number of inputs, hidden units, outputs, rate of learning, and momentum factor. However, the parameters of each model were distinctly different. That is, the weights assigned to the various features differed over teams. Each of these teams has had varying experience with tactical situations. Their schema knowledge will not be identical because they cannot have had identical experiences. We observed that frequently teams made different responses to the same pattern and did so consistently. For example, one team might consistently ignore an aircraft having a specific set of characteristics while another team might consistently pay attention to that same aircraft. Our models capture this performance difference.
Figure 1. The neural network model
Eye tracking studies. Most recently, I and my research group have begun to explore how data from eye tracking research can improve our cognitive models of learning and performance. Again, the situation of interest is information displayed visually on a computer display.
To date, we have carried out two different studies. One looks at how officers use a new display designed to assist them in making tactical decisions, and the second investigates how novices acquire critical knowledge needed to make tactical decisions.
In the first study, we are doing ground-breaking research as we use eye tracking technology with the officers. This work is being carried out in conjunction with ONR's Tactical Decision Making Program Under Stress (TADMUS). My research team gathers eye gaze data from participants in the TADMUS program as they use a new computer display developed by the NRaD TADMUS team in San Diego (see Figure 2). We have a number of important findings about the screen regions that are used most often, about the overall use of the system and the time at which it is used, and about the patterns of eye movement as the participants search for information in the display.
Figure 2. Eye-tracking study
In the second study, we are using the eye tracking instrumentation for two purposes. First, we can assess what an individual learns as he reads instructions that are presented on the computer display. The experimental task is a highly simplified display that mimics the display typically used by a commanding officer on a ship. From the eye movement data, we can determine which pieces of information are used by the individual and which are ignored. We can also estimate which areas of instruction are the most difficult based on the individual's pattern of gaze. This information becomes important baseline information for our computer models of human learning and performance. We then use the instrumentation to provide performance data as participants engage in tactical simulations. Again, these data are part of the input for our cognitive models.
EYE TRACKING IN TACTICAL DECISION MAKING SITUATIONS:
This paper describes the integration of eye tracking into an ongoing ONR research effort on tactical decision making, the TADMUS Program (Tactical Decision Making Under Stress). In this endeavor, we studied the gaze of individual officers as they looked at and interacted with tactical displays in a command and control setting. Despite the many problems that are inherent in implementing experimental laboratory research techniques in applied settings, we ran subjects in the usual TADMUS environment using the eye tracking instrumentation, and we have a number of important findings. This paper outlines the experimental procedure, describes the nature of the data collected, and summarizes the data analysis.
RECOGNIZING TACTICAL SITUATIONS:
HOW DO DECISION MAKERS KNOW WHEN TO MAKE A DECISION?
Abstract
Tactical decision making takes place in situations that are complex, that contain incomplete or ambiguous information, that have many possible outcomes, and that involve interactions with other participants. This report focuses on the schema knowledge used by decision makers in such settings. Data are presented from six highly experienced Navy teams who participated in a series of experimental simulations involving scenarios set in the Persian Gulf. Based on the empirical data, neural network models were developed that satisfactorily reproduce the performance of each team in the early stages of decision making.
Links to 6.2 Applications
* NAVAL TRAINING CENTER, GREAT LAKES, IL
Remedial Mathematics Tutor
* NAVAL COMMAND, CONTROL AND OCEAN
SURVEILLANCE CENTER, SAN DIEGO, CA
TADMUS Program
(Jeffrey G. Morrison, NRaD)
* NAVAL AIR WARFARE CENTER, ORLANDO, FL
TADMUS Program
(Joan Johnston, Training System Division)
Advanced Embedded Training,
Advanced Technology Demonstration
(Jan Cannon-Bowers, Training System Division)