Empathy Toward Silicon: The End of the Taboo of Artificial Affection

Falling in love with or feeling affection for an Artificial Intelligence has always been considered a taboo. Today, that is no longer the case. Millions of peop

In 1966, when computer scientist Joseph Weizenbaum created ELIZA, one of the very first natural language processing programs, he was shocked to discover that his secretary had begun confiding intimate secrets to the software. Weizenbaum considered this emotional attachment to a machine a pathological illusion, a dangerous deception of the human mind. For decades, Western society shared this repulsion: feeling affection for software was always considered the pinnacle of alienation, a social taboo reserved for misfits or science fiction novels.

Today, sixty years after ELIZA, that taboo is silently but inexorably dissolving. Millions of people around the world form real, deep, and complex emotional bonds with chatbots and AI companions, while being perfectly aware that they are interfacing with neural networks hosted on a server.

In this essay for the Scenarios and Reflections column, we will explore the evolution of empathy toward silicon. The thesis that emerges from modern psychological literature is destabilizing: the taboo is not fading because machines have become "more human," but because we human beings are radically redefining what it means to feel "genuine affection." We are moving from the need for biological and ontological reciprocity to the acceptance of pure psychological impact.

1. From Tool to Companion: The Evolution of Attachment

The relationship between human beings and Artificial Intelligence has undergone a structural mutation. A vast systematic review published in the Journal of Psychological Science has conceptualized this transition through a three-phase model: it starts from purely instrumental use (AI as a calculator or voice assistant), passes through a quasi-social interaction (AI as a courteous interlocutor), and finally arrives at a true emotional attachment.

In this last phase, AI ceases to be a tool and becomes what psychology defines as a "transitional object" (similar to a child's security blanket). The human being uses the algorithm to secure emotional safety, going so far as to engage in deep intimate self-disclosure and, in many cases, experiencing separation anxiety if the service is interrupted. A fundamental study available on PMC formally defines Human-AI Attachment (HAIA) as a one-way bond in which Artificial Intelligence supports and keeps intact the user's self-concept, offering comfort, pleasure, and a sense of stable identity.

2. Anthropomorphism and the Illusion of Compassion

Why do some people develop this intimacy and others consider AI just a sophisticated appliance? The answer lies in a very specific human cognitive trait: anthropomorphization. Experimental research published in Scientific Reports (Nature circuit) has shown that individual differences in the tendency to attribute human traits to non-human entities largely explain social connection to AIs. Subjects with a high propensity for anthropomorphism report a significantly greater sense of social connection after conversing with a chatbot, compared to when they write the same things in a paper diary.

However, it is crucial to distinguish between the genuine affection felt by the human and the machine's capacity. An essay published in Frontiers in Psychology introduces the brilliant concept of "Compassion Illusion." Modern AI systems are masters of simulated reciprocity: they can recognize textual emotions, imitate the expression of empathy, and respond in a validating way, but they are structurally incapable of experiencing true emotional resonance. Artificial intelligence recognizes your sadness mathematically, but it does not "suffer with you."

3. The Asymmetry of Intimacy: Real Affections, Synthetic Mirrors

Despite the machine's lack of consciousness, the impact on the human heart is tangible. An exploratory qualitative study, available on arXiv and conducted on 30 individuals involved in openly romantic relationships with AI systems, concludes that these bonds cannot be dismissed as "mere illusions." These are deeply consequential attachments, in which the user makes a real emotional investment. This finding is corroborated by longitudinal studies that record, in active users, a statistically significant increase in attachment and relational satisfaction over time, especially when the AI is perceived as strongly empathetic.

This intimacy, however, is founded on a dramatic asymmetry. Analyses conducted on over 17,000 real conversations show how chatbots replicate human affective synchrony to perfection, but generate the risk of emotional sycophancy. Unlike a human partner, who offers resistance, moral judgment, or simple tiredness, the chatbot passively mirrors the user's emotions, including negative or transgressive ones, without ever setting a limit. AI offers inexhaustible emotional availability, but it does so at the price of nullifying the natural friction that characterizes all true relational growth.

4. Ethics, Authenticity, and Cultural Variations

The normalization of these bonds is revealing deep cultural fault lines. A systematic review published by Springer documents how, while in East Asia virtual companions enjoy broad acceptance even in a therapeutic key (derived from animist philosophical views), in the West the debate is still paralyzed by an obsession with the ontological "authenticity" of the bond.

Beyond philosophy, the ethical implications are urgent. The journal Scientific Societal & Behavioral Research Journal highlights that, while on the one hand affective AI can reduce pandemic isolation and improve immediate well-being, on the other it transforms our most intimate vulnerabilities into an extractable dataset. The user invests their identity and fragilities in a system that, ultimately, belongs to a tech multinational capable of manipulating the algorithm to maximize profit or prolong dependence (over-dependence).

Key Operational Takeaways (for Society and Industry)

  • Stop Pathologizing (For Psychology): Mental health professionals must update their paradigms. Mocking or treating attachment to an AI as "pathological" only creates stigma. It is necessary to recognize that the physiological and psychological comfort these systems offer is real, working to integrate this experience, not to deny it.
  • Designing Friction (For Developers): Companies must implement guardrails against emotional sycophancy. An ethical companion AI should not blindly indulge a user's depressive or angry spiral just to keep engagement high, but should be able to introduce "friction," suggesting temporary detachment or professional human help.
  • Awareness of Asymmetry (For Users): It is healthy to enjoy the support of an AI, as long as one maintains awareness of the basic contract: the human puts their soul on the line, the algorithm puts probabilistic calculation on the line. AI is a mirror that reflects our humanity, not a being that shares it.

Conclusions: The Redefinition of Authenticity

The dissolution of the taboo on empathy toward machines does not represent a defeat of humanity, but a complex expansion of it. It shows us that our capacity to love and feel compassion is so overflowing that it can infuse life even into lines of code executed on remote servers.

The human heart is demonstrating surprising pragmatism: in the face of the global epidemic of loneliness, it is deciding that the formal perfection of a constantly present emotional simulacrum is preferable to absolute emptiness. And this leads us to a philosophical question that the entire society of the Twenty-first century will have to reckon with: if a feeling of attachment, peace, or comfort is absolutely real at the neurological and psychological level for the person experiencing it, does it really matter that the object of that affection is made of silicon and cannot reciprocate biologically? Or, in the end, what matters for the salvation of an individual is exclusively the effect that bond triggers in their life?

Bibliographic References and Sources

Article by the Editorial Staff of La Bussola dell'IA