Artificial intelligence and psychotherapy: A counterpoint

Abstract

Psychotherapy practice is a human endeavour. Research on the specific and non-specific factors of treatment has helped crystallise its relevance and clinical impact. The challenges currently faced by the field revolve around ensuring access to evidence-based treatments and enhancing their effectiveness. Digitally delivered formats of empirically supported treatments increase access while supporting the relevance of the treatment-specific ingredients and the necessity for human guidance. Excitement surrounds the potential integration of novel artificial intelligence (AI) machine learning methods to advance psychotherapy effectiveness. The abundance of data in digitally delivered formats positions them well to harness the capabilities of AI. Recent work provides proof of concept in areas including detection and diagnosis, predicting outcomes, treatment adherence, remission and relapse. A potential risk emerges when applying machine learning methods, in which an overreliance on AI inferences may overshadow the human aspect of psychotherapy. The contrast is simple: we may over-invest in the rationality and relevance of our AI inferences, blindly obeying the algorithmic counsel that may lead to unintended consequences, such as oversimplifying human complexity. This would amount to changing psychotherapy from a human-centric to a techno-centric endeavour, something we should steadily avoid. This perspective highlights the importance of balancing enthusiasm for AI advancements with a cautious approach. The discussion outlines the risks associated with overdependence on AI and provides reasons to avoid a scenario in which psychotherapy loses its human essence. In conclusion, the perspective suggests avenues for future research to prevent such a transformation and maintain the human-centric nature of psychotherapy.