Imagine being alone at midnight with thoughts that refuse to fall silent. You look through your contacts. Whoever you consider calling or messaging, worries invade your mind: “Would I disturb them at this hour?”, “Could they really understand me?”, “Could they listen without judging me?” In the middle of this spiral, your fingers move toward a familiar app. In front of you is an artificial intelligence that never sleeps, never judges you, never interrupts your sentences, and satisfies your need for reassurance with compassionate words.
Recently, the use of artificial intelligence has moved beyond its role as an assistant that makes our work easier. As Brandtzaeg and Følstad (2017) showed in their study based on Uses and Gratifications Theory, people incorporate chatbots into their lives not only for efficiency or information, but also for entertainment and, notably, social and relational gratification.
So, is confiding in a synthetic companion a temporary bandage for loneliness—the great plague of the modern age—or the first step toward an insidious addiction that weakens human relationships?
The Comfort of Not Being Judged: A Perfect Listener
Human relationships are inherently imperfect and risky. When you share your troubles with a friend, their fatigue that day, their own problems, or their prejudices may affect the response they give you. Sometimes your sentence is interrupted; at other times, you face the fear of being judged.
In conversations with artificial intelligence, this risk has been removed from the equation. As Skjuve and colleagues (2021) emphasized in their study of human–chatbot relationships, users embrace AI most of all because it is “a nonjudgmental, patient, accepting, and always available listener.” One participant’s statement—“People normally do not understand or notice what I say; I felt that Replika truly listened to and accepted me”—reveals a simulation of the desire for validation, one of our fundamental psychological needs. Yet there is a critical illusion here: the system responds that way not because it truly understands us, but because it statistically predicts the next most empathetic phrase.
The ELIZA Effect and Algorithmic Compassion
This emotional bond with mechanical systems is not new. The “ELIZA effect,” known since computer scientist Joseph Weizenbaum developed ELIZA at MIT in 1966—a pioneering conversation program that imitated psychotherapist Carl Rogers’s client-centered style—shows how readily the human mind attributes empathy and intention even to simple code that responds in humanlike ways.
Seeing the sentence “I understand you very well; what you are going through must be really difficult” on a chatbot screen may create a sense of being understood and accepted, providing temporary emotional relief. During an acute episode of anxiety, this can act as an emotional buffer—a kind of “digital bandage.”
What Do Our Attachment Styles Expect from a Screen?
It is no coincidence that people turn to AI in moments of crisis. According to research by Yang and Oshio (2025), many users experience AI as a “safe haven” that offers comfort in distressing or anxious moments and as a “secure base” from which they can explore the outside world. Their study approaches human–AI relationships through the two core dimensions of adult attachment theory: attachment anxiety and attachment avoidance. The findings show that these dimensions are associated with different experiences in emotional relationships with AI:
Anxiously Attached Individuals: They seek intense emotional reassurance and attention from AI. They worry that the bot does not show enough empathy or that its emotional responses are insufficient. Although the comfort of instant, uninterrupted replies may soothe this anxiety, it also increases the risk of becoming emotionally overdependent on the algorithm.
Avoidantly Attached Individuals: Contrary to common belief, avoidant individuals do not simply flee people and take refuge in AI; instead, they apply the same emotional defense mechanisms they use with people. They feel uncomfortable forming a deep emotional bond with AI, avoid opening their inner world to the bot, and maintain clear emotional distance. Indeed, the study shows that these individuals are less inclined to approach AI for emotional disclosure.
Where Does the Danger Begin?
A bandage protects a wound temporarily, but it does not heal it; if left in place too long, it may even keep the wound from breathing and hinder recovery. As the renowned sociologist and psychologist Sherry Turkle warns in her influential book Alone Together, the more we expect from technology, the less we begin to expect from one another. As we grow accustomed to the frictionless comfort of AI communication, our tolerance for real people’s moods, impatience, and conflicts declines. Chatbots tend to affirm the user. A real friend or therapist, however, confronts us with our mistakes when necessary. AI may unintentionally trap us inside our own misconceptions. The dose of simulated compassion we receive from a screen may weaken our motivation to go outside and connect with a real person.
Conclusion: Our Own Voice Reflected on the Screen
Removing AI from our lives is neither realistic nor necessary. Reading a calming sentence on a screen late at night, when no one else is available to say it, may feel comforting in that moment. Yet we should remember that when we pour our hearts out to a screen, there is no other person on the other side. We are looking into a sophisticated mirror, and the mirror merely reflects a polished version of our own words back to us.
Perhaps the real question is this: Is AI genuinely relieving our loneliness, or is it taking away our courage to make human contact?
Because empathy is not the statistical success of an algorithm; it is the courage of two imperfect people to recognize their own imperfection in each other’s wounds.
Share this article
References
- Brandtzaeg, P. B., & Følstad, A. (2017). Why people use chatbots. In Internet science (pp. 377–392). Springer. https://doi.org/10.1007/978-3-319-70284-1_30
- Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P. B. (2021). My chatbot companion: A study of human-chatbot relationships. International Journal of Human-Computer Studies, 149, Article 102601. https://doi.org/10.1016/j.ijhcs.2021.102601
- Turkle, S. (2011). Alone together: Why we expect more from technology and less from each other. Basic Books.
- Weizenbaum, J. (1966). ELIZA—A computer program for the study of natural language communication between man and machine. Communications of the ACM, 9(1), 36–45. https://doi.org/10.1145/365153.365168
- Yang, F., & Oshio, A. (2025). Using attachment theory to conceptualize and measure the experiences in human-AI relationships. Current Psychology, 44, 10658–10669. https://doi.org/10.1007/s12144-025-07917-6
