Artificial intelligence and social evolution

Artificial intelligence, a field once relegated to the outer reaches of science fiction, is now materializing in ways that exceed expectations. Interaction with AI, whether through dialogue, written text or the creation of photorealistic images, reveals both its immense potential and its disturbing dystopian capabilities. The ability of AI to pass exams in complex fields such as law and medicine raises questions about the possibility of an evolution beyond human ethical control. For medicine, the reference is the study by Kung and colleagues (2023), in which ChatGPT performed at or near the passing threshold on all three exams of the United States Medical Licensing Examination (USMLE).

Yet when AI is observed from a perspective that combines technical and sociological analysis, its social implications emerge. In this context, the discourse on AI becomes a spectacle, a screen that diverts attention from its real implications and from future prospects.

Closer analysis shows how deeply AI is intertwined with social relations. There is a double illusion: AI is represented as a form of cognition, on the basis of a mistaken understanding of the mind as a computational entity. In reality, AI derives from the existing social matrix, using operational concepts such as optimization, layers, vectors and loss function.

The social consequences of AI include “algorithmic violence,” which goes beyond symbolic violence and extends to predictive interventions with concrete and often destructive effects. This phenomenon is evident in the reinforcement of institutional inequalities, where AI, rather than acting as an equalizer, often amplifies and perpetuates existing social divisions. For example, a study published in Science in 2019 (Obermeyer and colleagues) showed that an algorithm widely used in the American health care system penalized Black patients: by using health care spending as a measure of the need for care, it assigned them fewer resources than White patients with similar conditions. According to the authors, correcting the bias would have raised the share of Black patients receiving additional help from 17.7% to 46.5% (Obermeyer et al., 2019).

AI also carries with it a hidden representation of human biases and beliefs. Algorithms, though presented as objective, are often steeped in the unconscious prejudices of their developers or of the data they are trained on. The “Gender Shades” study by Joy Buolamwini and Timnit Gebru (2018) revealed that three commercial facial analysis systems misclassified darker-skinned women in up to 34.7% of cases, compared with 0.8% at most for lighter-skinned men, mirroring and amplifying racial and gender biases.

These problems are further exacerbated by the lack of transparency and accountability in many AI systems. The AI “black box,” a term used to describe the opaque nature of algorithmic decision-making, makes it difficult to analyze and correct these distortions. Without a clear understanding of how algorithms reach their conclusions, it is almost impossible to mitigate their harmful effects on society. The European regulation on artificial intelligence (the AI Act), whose transparency obligations apply from August 2, 2026, is a first attempt at a response.

Algorithmic violence, then, is not merely an unwanted side effect of AI; it is a fundamental challenge that must be addressed to ensure that technological progress proceeds fairly and justly. A sustained commitment to research and policymaking is essential to identify, understand and mitigate these harmful effects, in order to steer AI toward a more equitable and inclusive future.

Instead of science-fiction futures, AI may bring a return to the industrial relations of the nineteenth century and the dissolution of the postwar social contracts, as the elements of social bargaining tied to the value of labor disappear. The sociological analysis of AI must be rooted in its materiality.

To mitigate the risk that AI becomes an instrument of power concentrated in the hands of a few, it is essential to direct research and action toward socially responsible goals. This approach requires placing at the center of AI development efforts not only technological advancement but also collective well-being and social justice.

Furthermore, a convivial techno-social system, one that is in balance with its environment, can be achieved only through coordination that begins with autonomy from below.

The effort to develop an alternative techno-social apparatus should not be limited to a sociological analysis of the potential harms of AI. Instead, sociology should play a leading role in shaping the future of AI. This requires integrating a technical understanding of AI with an awareness of its social, economic and cultural implications.

What is needed is a holistic approach that takes into account the different dimensions of AI's impact. This includes considering how AI affects the balance of power in society, the dynamics of work, questions of privacy and surveillance, and the way it can be used to reinforce or dismantle existing structures of injustice.

In conclusion, the transition to a convivial techno-social system requires a fundamental rethinking of how AI is developed and deployed. Only by placing social goals at the center and promoting participation from below can we hope to steer AI toward a future that benefits all members of society. This requires not only technological expertise but also a deep sociological understanding and a sustained ethical commitment.

(Alberto Viotto)