Modeling Effects of Age in Complex Tasks: A Case Study in Driving Dario D. Salvucci Alex K. Chavez (salvucci@cs.drexel.edu) (achavez@drexel.edu) Frank J. Lee (fjl@cs.drexel.edu) Department of Computer Science, Drexel University 3141 Chestnut St., Philadelphia, PA 19104 Abstract While computational cognitive modeling has made great strides in addressing complex dynamic tasks, the modeling of individual differences in complex tasks remains a largely unexplored area of research. In this paper we present a straightforward approach to modeling individual differences, specifically age-related cognitive differences, in complex tasks, and illustrate the application of this approach in the domain of driving. We borrow ideas from rigorous work in the EPIC cognitive architecture (Meyer et al., 2001) and extend them to the ACT-R architecture (Anderson et al., in press) and a recently-developed ACT-R driver model (Salvucci, Boer, & Liu, 2001) to model the effects of age on driver behavior. We describe two validation studies that demonstrate how this approach accounts for two important age-related effects on driver performance, namely effects on lateral stability and brake response during both normal driving and driving while performing a secondary task. Introduction Computational architectures and cognitive modeling have in recent years begun to account for increasingly complex and dynamic tasks, in domains such as piloting combat aircrafts (Jones et al., 1999) and controlling air traffic (Lee & Anderson, 2001). While such models have captured many aspects of human cognition and performance in these tasks, one aspect of complex tasks, namely individual differences, remains a largely unexplored area of research. The modeling community has seen several rigorous studies of individual differences in the context of cognitive architectures, perhaps most notably the work of Meyer et al. (2001) in the EPIC cognitive architecture (Meyer & Kieras, 1997) and that of Lovett, Daily, and Reder (2000) in the ACT-R architecture (Anderson et al., in press). However, due to their emphasis on specific sources of individual differences, these studies focused on relatively short laboratory tasks in controlled environments rather than more complex continuous tasks in dynamic environments. Our goal in this paper is to generalize ideas from existing work on individual differences in simpler tasks to account for individual differences in complex dynamic tasks. We illustrate our approach in the domain of driving, a complex task that people perform on daily basis. There now exist several so-called “integrated driver models” (e.g., Aasman, 1995; Levison & Cramer, 1995) that attempt to combine the lower-level aspects of driving (e.g., steering control) with the higher-level aspects of the task (e.g., decision making, navigational planning). In particular, Salvucci, Boer, and Liu (2001) have developed and refined an ACT-R driver model that predicts many aspects of driver control, situational awareness, and decision making during common highway driving. However, to date, no integrated models of driving, including the ACT-R driver model, have accounted rigorously for any individual differences in driver behavior and performance. This paper builds on previous work by presenting an account of individual differences, specifically age-related differences, in the complex task of driving. Not surprisingly, age plays a significant role in driver differences, often couched in broad terms as differences between younger drivers (roughly 20-30 years of age) and older drivers (roughly 60-70 years of age). Our approach borrows recent results of Meyer et al. (2001), who explored models of age-related individual differences in the context of the EPIC cognitive architecture. Age effects on driving offer a particularly interesting challenge to computational cognitive modeling: on the one hand, some studies have found that older drivers exhibit performance equal to that of younger drivers for certain combinations of driving and/or secondary tasks; on the other hand, other studies have found that older drivers sometimes experience extremely reduced performance, particularly in the presence of secondary tasks (e.g., using a cell phone). Thus, the effects of age are far from trivial and must be taken in the fuller context of both the complex behavior necessary for driving and also the complex interaction between the driver and the “artifact” (i.e., vehicle, road, etc.) through which the driver’s behavior is externalized. In the next section of the paper, we describe our basic approach and its instantiation in the ACT-R cognitive architecture. We then present two modeling studies that validate our approach for complementary tasks and aspects of behavior, namely drivers’ ability to maintain lateral stability on the road and drivers’ ability to respond (i.e., brake) to sudden external stimuli. While our work in this paper emphasizes driver behavior and the ACT-R cognitive architecture, the fundamental ideas generalize well to other complex task domains and other modeling frameworks. Thus, our ultimate goal is to explore the interaction between basic individual differences and their downstream effects on performance in complex dynamic task environments.
[1]
J. Cerella.
Information processing rates in the elderly.
,
1985,
Psychological bulletin.
[2]
Dario D. Salvucci.
Interfaces on the Road : Rapid Evaluation of In-Vehicle Devices
,
2004
.
[3]
Matthew P. Reed,et al.
Comparison of driving performance on-road and in a low-cost simulator using a concurrent telephone dialling task
,
1999
.
[4]
Marsha C. Lovett,et al.
A source activation theory of working memory: cross-task prediction of performance in ACT-R
,
2000,
Cognitive Systems Research.
[5]
P A Hancock,et al.
The distraction effects of phone use during a crucial driving maneuver.
,
2003,
Accident; analysis and prevention.
[6]
Frank J. Lee,et al.
Does Learning a Complex Task Have to Be Complex?: A Study in Learning Decomposition
,
2001,
Cognitive Psychology.
[7]
T. Salthouse,et al.
Divided attention abilities in young and old adults.
,
1982,
Journal of experimental psychology. Human perception and performance.
[8]
William H Levison,et al.
DESCRIPTION OF THE INTEGRATED DRIVER MODEL
,
1995
.
[9]
W. Surwillo,et al.
The relation of simple response time to brain-wave frequency and the effects of age.
,
1963,
Electroencephalography and clinical neurophysiology.
[10]
Shane T. Mueller,et al.
Executive-process interactive control: A unified computational theory for answering 20 questions (and more) about cognitive ageing
,
2001
.
[11]
Jannes Aasman,et al.
Modelling driver behaviour in soar
,
1995
.
[12]
Frank J. Lee,et al.
Simple cognitive modeling in a complex cognitive architecture
,
2003,
CHI '03.
[13]
D. Woodruff.
Relationships among EEG alpha frequency, reaction time, and age: a biofeedback study.
,
1975,
Psychophysiology.
[14]
John R Anderson,et al.
An integrated theory of the mind.
,
2004,
Psychological review.
[15]
A. Newell.
Unified Theories of Cognition
,
1990
.
[16]
Erwin R. Boer,et al.
Toward an Integrated Model of Driver Behavior in Cognitive Architecture
,
2001
.
[17]
Dario D. Salvucci.
Predicting the effects of in-car interface use on driver performance: an integrated model approach
,
2001,
Int. J. Hum. Comput. Stud..
[18]
Daniel V. McGehee,et al.
Collision Warning Timing, Driver Distraction, and Driver Response to Imminent Rear-End Collisions in a High-Fidelity Driving Simulator
,
2002,
Hum. Factors.
[19]
C. Lebiere,et al.
The Atomic Components of Thought
,
1998
.
[20]
Randolph M. Jones,et al.
Automated Intelligent Pilots for Combat Flight Simulation
,
1998,
AI Mag..
[21]
D E Kieras,et al.
A computational theory of executive cognitive processes and multiple-task performance: Part 1. Basic mechanisms.
,
1997,
Psychological review.