Beyond the Black Box: Why Algorithms Cannot Replace the Unconscious or the Psychodynamic Therapist

Aner Govrin^*

• The Program for Hermenutics and Cultural Studies, Bar-Ilan University, Ramat Gan, Israel

This paper critically examines the limitations of artificial intelligence in replicating human psychological processes, specifically challenging its ability to capture the complex structures of the unconscious and the nuanced dynamics of psychotherapeutic relationships. Drawing on psychoanalytic theory, particularly Matte-Blanco's analysis of unconscious logic and Winnicott's concept of therapeutic holding, the research demonstrates that AI fundamentally fails to engage with the non-linear, contradictory, and embodied nature of human psychological experience. To substantiate these theoretical claims, the paper presents a clinical vignette that illustrates AI's profound therapeutic shortcomings, specifically its inability to address complex psychological issues like separation anxiety and projective identification. The case study highlights critical therapeutic elements AI cannot replicate, such as meaningful silence, nuanced countertransference, and embodied emotional containment. While algorithmic systems may superficially mimic pattern recognition, they cannot replicate the profound intersubjective, temporal, and affective dimensions of human psychological understanding. The study warns of a more insidious risk: patients potentially modifying their psychological self-presentation to conform to computational logic, thereby sanitizing and distorting their complex inner experiences. Ultimately, the paper argues that AI's limitations are structural rather than technical, emphasizing the irreplaceable role of embodied, relational human connection in psychological care and understanding, while acknowledging AI's potential supplementary functions in mental healthcare.