In most contexts categorizing an object involves learning to identify what counts as a member of the category. Figuring out which strategies people use in learning categories has been one of the central focuses of the researchers in the field. In many models of human attentional learning a key feature is that it learns to shift its attention in response an error signal provided from objective feedback. In studies conducted in our lab we have observed that in many contexts human do not in fact conform to this error driven assumption. On the contrary shifts of attention have been observed to occur equally when the participant makes a correct response and when they make an error.
The investigations in this project seek to make clear that current models of attention such as RASHNL rely heavily on the error driven assumption. Through a variety of simulation methods we are illuminating the error driven nature of these models. By comparing the results from these simulations with human eyetracking data recorded in our lab we are showing that the error driven assumption is untenable.