The Social and Cultural Structure of Artificial Intelligence
Available on Academia.edu (PDF)
Indice
Abstract
This essay explores the concept of “digital habitus” as applied to artificial intelligence systems, offering an interdisciplinary analysis that weaves together philosophy, sociology, and technology studies. It traces the historical evolution of the notion of habitus from Aristotle and Thomas Aquinas to Pierre Bourdieu’s sociological theory, before applying it to the digital realm through Massimo Airoldi’s “machine habitus.” Machine learning algorithms incorporate cultural dispositions embedded in training data, reproducing and amplifying the social structures from which they originate. The text compares this perspective with Alberto Romele’s critique, which highlights the risks of technological determinism and anthropomorphization. Artificial intelligence is conceptualized as a “non-human social agent”: an entity which, although devoid of moral intentionality and consciousness, effectively structures social space and influences human behavior. Through empirical examples and theoretical reflections, the analysis demonstrates how the machine habitus operates as a dynamic circularity between the physical and digital dimensions, functioning as a mediator that transforms contemporary social practices.
Keywords: machine habitus, digital sociology, artificial intelligence, non-human social agents, algorithmic mediation
Introduction
During the training phase of algorithms lie the processes aimed at preventing the biases and hallucinations that afflict AI systems—and these, in turn, generate new problems. For instance, removing racist or xenophobic material from datasets is not enough to eliminate the danger of segregationism and raises serious questions about what should be omitted and what should be remembered: should apartheid be erased or remembered? Recently, Matteo Flora drew attention to this topic on his blog *Ciao, Internet*, where he compared the responses of two AI systems: ChatGPT and DeepSeek. He asked both about the events of Tiananmen Square in 1989. While ChatGPT, a U.S. product, attempted to maintain a line of historical objectivity, DeepSeek, a Chinese product, declined the question (“I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses”). Each AI responds with an ethics determined by its training dataset and reproduces it for an audience that will be influenced by it. Bourdieu used the notion of habitus to describe this conditioning effect of social and cultural structures. Can we speak of a habitus for AI as well? According to Massimo Airoldi, we can.
The Concept of Habitus
Aristotle used the concept of habitus (hexis) to describe the enduring disposition that enables the individual to act in a virtuous or vicious manner. Moral virtues, according to the Stagirite, are not innate but are formed through habit and the repeated exercise of virtuous actions. Habitus thus has the contours of a form of practical knowledge and a way of being in the world.
Thomas Aquinas, drawing on Aristotle, reworked the notion of habitus into a stable disposition of the soul that orients the individual toward the good, together with other dispositions (such as the theological virtues of faith, hope, and charity) infused by God. Habitus is therefore a combination of repeated human acts and divine intervention.
In the modern period, moral psychology and pedagogy made extensive use of this concept to discuss the habits or dispositions that shape an individual’s character.
Pierre Bourdieu
For the French sociologist Pierre Bourdieu, habitus is a set of acquired dispositions that influence how individuals perceive and act in the world and how they relate to others. These dispositions are formed through social experiences, especially during childhood, and are tied to an individual’s social position.
Certain characteristics of habitus are particularly relevant for our purposes. Although it reproduces the same social structures from which it arises, it also adapts to new experiences and social contexts. Yet this reproduction of the original structures, even if adapted, makes it a tool for the perpetuation of social hierarchies. For Bourdieu, habitus does not completely eliminate individual freedom (agency) but situates it within a framework of limited possibilities.
Habitus as Mediator Between the Physical and the Digital
The sociologist Massimo Airoldi has extended the concept of habitus to reconcile it with the dynamics of digital worlds, introducing the notion of *machine habitus* as a way to investigate the relationship between physical and digital environments. Our behavior is not rootless; it is influenced by internalized schemes and contextual belonging. Digital practices—such as the use of social media—intertwine with traditional social structures, creating a form of continuity between the physical and the digital rather than a clear separation. Habitus, therefore, provides the cognitive and practical framework for the dynamic relationship between the physical and digital spheres.
Alberto Romele’s Critique
The philosopher Alberto Romele, in his work on the philosophy of technology and information, while appreciating the insight into the growing role of algorithms in our lives, has offered a critique of Airoldi’s idea of the machine habitus.
One of his main concerns is the risk of underestimating the active role of human beings in shaping their own existence in relation to machines. According to Romele, it is necessary to account for the co-creation between humans and machines, where both parties actively participate in the formation of habitus.
Another point of attention is the notion that technology inevitably determines human behavior and dispositions. This approach, Romele argues, fails to fully recognize the human capacity to resist, negotiate, and reinterpret technological influences. He also calls for consideration of the forms of agency that individuals maintain in their interactions with technologies.
A further critique concerns the attribution of a form of “autonomy” to algorithms. To speak of a machine’s habitus presupposes a capacity for reflection and internalization that machines do not possess. Algorithms, however sophisticated, lack intentionality and consciousness; they operate according to mathematical rules and learning patterns. For this reason, the notion of *machine habitus* may excessively anthropomorphize technology.
Non-Human Social Agents
The cultural heritage formed in the physical world guides a person’s use of technology and is enriched by experience in the digital domain: this is the circularity of the machine habitus. I disagree with Romele’s concern that this view underestimates human action, because the differing reactions and relationships between humans and machines are part of the experiential feedback that sustains the circularity. Determinism becomes the subjective balance between trust in humans and trust in structures, since it is impossible to deny that a reciprocal influence exists.
Regarding the autonomy of algorithms, the issue becomes more complex. Perhaps we should stop, as Elena Esposito suggests, arguing about the possibility of intelligence or sentience in AI and begin to think of it instead as a different kind of intelligence—specifically, a dialogical one. What is truly astonishing is that AI can maintain a dialogue with humans and even influence them.
The capacity to interact with and even modify the surrounding reality, directly or indirectly, defines AI as a social agent—an entity capable of acting within reality. Being digital entities, I prefer to call them non-human social agents, a category that can—albeit cautiously—be integrated into the actor-network framework theorized by Bruno Latour, in which human and non-human agents collaborate in the generation of events.
Two examples illustrate this dialogical agency of AI. The first comes from Fabio Lalli’s reflections on the possibility of enhancing self-awareness: a process of discernment assisted by AI to understand what we want and why. Lalli imagines an exemplary dialogue: “AI, can you make some photocopies?” “Which pages? In color or black and white? How many copies?” Communicating with AI demands clarity of intent and purpose. The second example concerns the possibility of an examination of conscience with AI, which could retrace our daily activities—movements, meetings, emails, phone calls, messages—and ask: how did we behave in those situations? AI would be pitiless in recalling everything!
Conclusion
The machine habitus heralds the entrance of a new agent into human life. The suspicion and caution with which AI is received should not surprise us, yet we must remember that good can also be transmitted through AI. If technology is a gift from God, it is both easier and more rewarding to use it for good than for evil. The machine habitus can become a cultural, informational, and inspirational field for good—a space where good news, good actions, and good thoughts converge to influence behavior and thinking.