Information Technology Reference
In-Depth Information
tional techniques to understand and to explain human behavioral performance: neu-
roimaging and computational modeling. With neuroimaging techniques, such as
functional magnetic resonance imaging (fMRI [8]), positron emission tomogra-
phy (PET [9]), and event-related potentials (ERP [29]), researchers can uncover
the neural substrates that mediate behavioral performance. These neuroimaging
techniques not only allow researchers to localize where cognitive processes re-
side in the brain, but also allow researchers to uncover commonalities and dis-
similarities between cognitive tasks, discover individual differences, and test psy-
chological theories and models in ways that behavioral techniques alone could not
uncover [3].
Computational modeling has also been a powerful technique to simulate and
compose models for how behavior is mediated. Computational models can be clas-
sified into a number of categories, including, e.g., connectionist [19, 30, 39], sym-
bolic [24, 31], and hybrid [4, 27, 59, 51, 48, 53, 55, 54, 56, 52, 60, 58, 57, 61, 62, 63].
With these computational models, researchers are able to validate, test, and up-
date psychological theories in ways that behavioral testing alone could not do
easily.
Here we utilize computational modeling to account for changes in performance
both behaviorally and neurally due to practice and learning in the context of tran-
scription typing and the psychological refractory period (PRP; the slowing of a sec-
ondary task when it is initiated during the response of a primary task). This novel
model unifies many disparate findings together into a single model without needing
to make many changes to model parameters.
We chose to model the practice and learning effects in transcription typing and
PRP due to the following reasons. First, transcription typing involves intricate and
complex interactions of perceptual, cognitive, and motoric processes, and modeling
its learning processes can help us understand the underlining quantitative mecha-
nisms in complex motor skill acquisition. Second, there exist brain imaging data
on typing and typing related behavior [17, 23] that could be modeled. In addition,
human behavioral performance data, such as typing speed and typing variability,
have been obtained via several experimental studies (please see the review of Salt-
house [44]).
We modeled the learning effect in PRP for similar reasons. First, PRP is the
simplest and one of the most basic paradigms to study multitask performance
and has been used extensively as a paradigm to study multitask performance.
The PRP effect has been applied in many real-world settings such as driving
[25] and has been used as a measure of dual-task competency [5, 11]. There-
fore, modeling the learning effects in PRP may allow us to account for the ba-
sic mechanisms in the acquisition of multitasking skills. Second, an experimen-
tal study has been conducted to study the learning effect in PRP [49], which
provides important human performance data for modeling. For these reasons we
found transcription typing and PRP tasks good candidates to model skill learning
behavior.
