Hyperlearning: A Connectionist Model of Psychosis in Schizophrenia (2009)
Abnormal brain processes that underlie schizophrenia are incompletely understood. Diagnosis of this disorder relies in large part on psychotic symptoms that are observed through conversational language. In this paper, two such symptoms, delusions (fixed false beliefs) and derailments (inability to follow a coherent discourse plan) are modeled using DISCERN, a connectionist model of human story processing. Simulations of alternative pathologies thought to underlie schizophrenia are applied to DISCERN, and the resulting language abnormalities are evaluated for symptoms of schizophrenia. "Hyperlearning", a simulation of excessive dopamine release, is shown to produce a compelling model for both delusional and derailed language. Applied to different locations in the model, hyperlearning led to different symptoms, suggesting how clinical subtypes of schizophrenia could arise from a common underlying process.
In Proceedings of the 31st Annual Meeting of the Cognitive Science Society, N. A. Taatgen and H. van Rijn (Eds.), Amsterdam, The Netherlands 2009.

Uli Grasemann Postdoctoral Alumni uli [at] cs utexas edu
Ralph E. Hoffman Formerly affiliated Collaborator ralph hoffman [at] yale edu
Risto Miikkulainen Faculty risto [at] cs utexas edu