This ambitious project is to create a dynamic field theory model of category learning which allocates attention in ways that mimic human overt attention.
DFT models utilize populations of interacting neural units as their basic representational structure. As the model learns and processes information presented to it in a task, the pattern of neural activations change dynamically and adapt to features of the environment. One thing that is particularly interesting about these models is that the processing occurs in time and we can observe the model changing its activations as a continuous processes.
Our lab’s interest in these models stems from our desire to model eye movements that have been recorded in a number of categorization tasks. One of our findings is that the participants in our studies were shifting their attention in a highly dynamic manner, in that the location of their fixations were dependent on which stage of the trial they were at.
We are currently preparing a manuscript based off the thesis work of Walshe (2011) and Barnes (2012). Some of the results that have been captured by our DFT model include within-trial fixation orderings, experiment level attentional optimization, fixation duration regularities and the possibility of a Hebbian based eye-movement perseveration finding.
To get more information about this project, contact Jordan Barnes or Mark Blair.