The Algorithmic Unconscious

by Luca M Possati

This book applies the concepts and methods of psychoanalysis to the study of artificial intelligence (AI) and human–AI interaction. It develops a new, more fruitful approach for applying psychoanalysis to AI and machine behavior. It appeals to a broad range of philosophers working on psychoanalysis, technology, AI ethics, and cognitive sciences, psychoanalysts, psychologists, and computer scientists. The book is divided into four parts. The first part (Chapter 1) analyzes the concept of "machine behavior." The second part (Chapter 2) develops a reinterpretation of some fundamental Freudian and Lacanian concepts through Bruno Latour’s actor-network theory. The third part (Chapters 3 and 4) focuses on the nature and structure of the algorithmic unconscious. The author claims that the unconscious roots of AI lie in a form of projective identification, i.e., an emotional and imaginative exchange between humans and machines. In the fourth part of the book (Chapter 5), the author advances the thesis that neuropsychoanalysis and the affective neurosciences can provide a new paradigm for research on artificial general intelligence. The Algorithmic Unconscious explores a completely new approach to AI, which can also be defined as a form of "therapy." Analyzing the projective identification processes that take place in groups of professional programmers and designers, as well as the "hidden" features of AI (errors, noise information, biases, etc.), represents an important tool to enable a healthy and positive relationship between humans and AI. Psychoanalysis is used as a critical space for reflection, innovation, and progress.

144 pages, Hardcover

Published February 25, 2021

Reviews

[Ricardo Acuña · Rating 4 out of 5]

Luca M. Possati's "The Algorithmic Unconscious:" is an interdisciplinary exploration into the intersection of psychoanalysis and artificial intelligence, proposing that psychoanalytic concepts can illuminate the often opaque behaviors of AI systems. Possati reinterprets foundational Freudian and Lacanian theories through the lens of Bruno Latour's actor-network theory, suggesting a symbiotic relationship between human psychology and technological artifacts.

Central to Possati's thesis is the notion of the "algorithmic unconscious." He posits that AI systems embody an unconscious layer, shaped by the developers projections and emotional exchanges—a process he terms "projective identification." This perspective implies that our interactions with AI are not purely rational but are imbued with deep-seated desires and fears, influencing both AI development and deployment.

His conclusions emphasize a) The need to integrate psychoanalytic concepts into AI research to address the subconscious and emotional dimensions of artificial intelligence; b) Future studies that could investigate how AI systems develop "unconscious" patterns of behavior based on hidden biases in data, training processes, and interactions with users; c) A new area of research that can explore how AI’s "unconscious" reflects collective human biases and whether psychoanalytic tools can help make these implicit structures more transparent.

Possati address the importance of forming AI development teams that include psychoanalysts, cognitive scientists, and ethicists alongside engineers and data scientists. These teams could: analyze how AI systems interact with users on a psychological level; design AI models that take into account affective and unconscious human responses; develop new ways to interpret and visualize AI decision processes through a psychoanalytic lens.

It is indeed a provocative and disruptive theoretical framework to analyze the AI systems from a new perspective that includes not only the technical dimension, but to include a human-to-machine psychoanalytical dimension, the dynamics and how they affect and influence each other. Given that psychoanalysis is a controversial field, considering the hypothesis are based on phenomenological investigation based on subjective non-scientific methods, it is questionable the central hypothesis of the book: "that the concepts and methods of psychoanalysis can be applied to the study of AI and human–AI interaction". However, I am a psychoanalysis advocate, so it deserves to consider the author´s proposals because they could result useful for the AI research field.