Sensing Psychosis: Toward Robust Computational Phenotypes in Severe Mental Illness

Justin Baker
Harvard Medical School

Until recently, the inconsistency of neurobiological measures taken together with the imprecise nature of clinical observation has severely limited our ability to apply a personalized medicine approach in psychiatry, or to optimize care for all individuals seeking treatment. As a result, psychiatric care has remained largely open-loop and non-quantitative in nature, since providers may resist adding coarse, quantitative methods to their nuanced, albeit subjective, assessments. The widespread adoption of pervasive computing provides the field with unprecedented opportunities to build and test deep, dynamic models of illness by quantifying behavior at the level of individuals over time. If harnessed effectively, these new tools will allow us to move past the false choice between precise and personalized psychiatry that has confounded the field and limited progress. Critically, unobtrusive, quantitative behavioral phenotyping strategies could transform our ability to infer causal relationships between illness fluctuations, contextual factors, and treatment interventions and thereby radically reshape the process of discovery and development of novel therapeutics, making it more closed-loop, personalized, and targeted toward specific neural circuits. Here I discuss recent efforts to bridge these complementary approaches through single-case experimental designs in individuals with severe mental illness including bipolar disorder and schizophrenia.

Presentation (PDF File)

Back to Computational Psychiatry