Modeling sequential information integration with parallel constraint satisfaction

An important aspect of human cognition is the sequential integration of observations while striving for a consistent mental representation. Recent research consistently stresses the importance of fast automatic processes for integrating information available at a certain point in time. However, it is not clear, how such processes allow for maintaining a consistent and up to date mental representation in the light of new information. We compare variants of two methods of modeling sequential information integration with parallel constraint satisfaction models: carrying over results from the previous integration step or decaying input strength of older observations. Results of these models for consistent and inconsistent sets of observations are compared to human data from a diagnostic reasoning task.