Information Technology Reference
In-Depth Information
the majority of entities of task second went through the first route rather than the
second route. In some ways this is supported by recent neuroimaging work on PRP
by [20]. Those authors found little 33 differences in activations/neural networks
in the PRP task when performance was assessed at long and short SOAs. Such
large activation differences between short and long SOAs would be predicted by
active monitoring theories of the PRP effect. However, Jiang et al. [20] contend
that their data suggest that the PRP effect reflects passive queuing and not active
monitoring. This is yet other evidence supporting the queuing network architecture
and structure of our model as we did not find much difference in performance in
the Hicog server before and after practice and at short and long SOAs. In addition,
routes are chosen passively with Q learning and are not subject to active monitoring
processes.
With the formation of an automatic process during learning, two parallel routes
were formed in the dual-task situation, which partially eliminated the bottleneck at
the Hicog server. The PRP effect is reduced greatly with the decrease in the pro-
cessing time in both the Hicog and the PM server. However, since the majority of
the entities of the two tasks still went through the Hicog server, the effect of the
automatic process on PRP reduction does not exceed the effect of the reduction of
RT 1 on the PRP effect. This is consistent with the result of Van Selst et al. [49] that
the automatic process does grow from weak to strong but only weakly contributes
to PRP reduction.
9.4 Discussion
In the previous sections of this chapter, we described the modeling of brain ac-
tivation patterns as well as the behavioral phenomena in learning of two basic
perceptual-motor tasks (transcription typing and PRP). In modeling the phenom-
ena in typing, reinforcement algorithms guided how the entities traversed through
different routes before and after learning. The brain areas activated both before and
after learning are consistent with neuroimaging findings. In modeling PRP practice
effects, we used the same simulation model to quantify the formation of automatic
processes (reduction of the visual-motor task 2) during the learning processes in Van
Selst et al. [49] study.
There are several questions to be answered by future research utilizing our model.
First, neuroscience evidence has shown that many brain areas have overlapping
functionality which was not captured by the current model, which assumed discrete
brain areas with specific functions. This will increase the difficulty in modeling the
cooperation of information processes in the different brain areas. Second, the travel-
ing of entities from one server to another does not necessarily indicate the activation
of two brain areas. Brain area activation as uncovered with fMRI studies is based on
brain hemodynamics, which is an indirect measure of neural activity and thus has
poor temporal resolution. Therefore, using fMRI data to guide modeling of process-
ing times is somewhat tenuous. Therefore, 35 caution should be taken in comparing
the simulation results of the model with the results of fMRI studies.
We are currently developing a computational model of the human cognitive sys-
tem which is able to account for experimental findings in both neuroscience and
