Why the race to apply AI in psychiatry is so dangerous, and how to understand the new tech-driven psychiatric paradigm.
AI psychiatrists promise to detect mental disorders with superhuman accuracy, provide affordable therapy for those who can’t afford or can’t access treatment, and even invent new psychiatric drugs. But the hype obscures an unnerving reality.In The Silicon Shrink, Daniel Oberhaus tells the inside story of how the quest to use AI in psychiatry has created the conditions to turn the world into an asylum. Most of these systems, he writes, have vanishingly little evidence that they improve patient outcomes, but the risks they pose have less to do with technological shortcomings than the application of deeply flawed psychiatric models of mental disorder at unprecedented scale. Oberhaus became interested in the subject of mental health after tragically losing his sister to suicide. In the book, he argues that these new, ostensibly therapeutic technologies already pose significant risks to vulnerable people, and they won’t stop there. These new breeds of AI systems are creating a psychiatric surveillance economy in which the emotions, behavior, and cognition of everyday people are subtly manipulated by psychologically savvy algorithms that have escaped the clinic. Oberhaus also introduces readers to the concept of “swipe psychology,” which is quickly establishing itself as the dominant mode of diagnosing and treating mental disorders.
It is not too late to change course, but to do so means we must reckon with the nature of mental illness, the limits of technology, and what it means to be human.
Daniel Oberhaus is a science and technology communicator based in New York City.
He is represented by Robert Guinsler at Sterling Lord Literistic and the author of two books for MIT Press.
His most recent book, The Silicon Shrink (2025), is about the promise and peril of AI in psychiatry. His first book, Extraterrestrial Languages (2019), is about the art, science, and philosophy of interstellar communication.
Daniel is the CEO and founder of HAUS, a strategic communications agency for deep tech startups. Previously, he was a staff writer at WIRED Magazine covering space exploration and the future of energy.
A thought provoking examination of the history of psychiatric care and the early intersection with the development of machine learning systems. The book can be a little technical, but its basic premise is not hard to understand. Advances in psychiatric treatment over the years have not materially improved patient outcomes or stemmed the rising number of people afflicted with mental disorders, largely due to the vague and difficult quantitative nature of mental health. Therefore it is alarming that the increasing use of AI programs to treat mental health issues proceeds at a rapid pace despite underlying problems of patient confidentiality, excessive surveillance and ultimately treatment effectiveness.