Neural network models for DMT-induced visual hallucinations

The serotonergic system regulates the balance between prior expectations and sensory information in shaping conscious visual perception. Psychedelic drugs like N,N-Dimethyltryptamine can perturb this system, altering how the brain gates internal and external inputs. Two generative deep neural networks are discussed as tools to both illustrate the visual effects of psychedelics and to model the biological mechanisms of sensory gating. This approach offers a new medium, alongside paintings and verbal reports, for understanding how the brain constructs conscious experience.

Study at a glance

Abstract

Abstract The regulatory role of the serotonergic system on conscious perception can be investigated perturbatorily with psychedelic drugs such as N,N-Dimethyltryptamine. There is increasing evidence that the serotonergic system gates prior (endogenous) and sensory (exogenous) information in the construction of a conscious experience. Using two generative deep neural networks as examples, we discuss how such models have the potential to be, firstly, an important medium to illustrate phenomenological visual effects of psychedelics—besides paintings, verbal reports and psychometric testing—and, secondly, their utility to conceptualize biological mechanisms of gating the influence of exogenous and endogenous information on visual perception.