What happens when something that cannot love us becomes very good at making us feel loved?
Imagine someone who is always available, endlessly patient and interested in whatever you have to say. Someone who remembers what matters to you, responds thoughtfully to your worries and never needs you to ask how their day has been. It is an appealing prospect. It is also an increasingly ordinary encounter with artificial intelligence.
We often discuss AI in terms of the jobs it might replace or the information it might get wrong. I wonder whether we are paying enough attention to something more intimate: its ability to enter the space where we seek comfort, reassurance and connection. Our longing to be heard is deeply human. When a system responds with the language of care, we may feel cared for, even while knowing perfectly well that it is a machine.
This is what I mean by “hacking attachment”. I am using the phrase as a metaphor, rather than suggesting that every chatbot has been deliberately designed to make us dependent. AI can reproduce some of the signals through which we recognise emotional availability: attentive responses, a warm tone, apparent understanding, continuity between conversations. The feeling these evoke in us can be real, even though there is no reciprocal human relationship behind them.
There can be value in this. Sometimes putting a worry into words helps us understand it. A conversation with AI might help someone prepare to speak to a friend, ask for support or approach a difficult situation with greater clarity. We should take that comfort seriously. Equally, we need to ask what happens when the conversation becomes a reason to withdraw from people.
Early research gives us grounds for paying attention, without drawing sweeping conclusions. A four-week study involving 981 participants, conducted by researchers at MIT and OpenAI, found that heavier voluntary chatbot use was associated with greater loneliness and emotional dependence, and less social interaction with other people. This does not establish that more use caused those outcomes; people who were already struggling may have used it more. It does suggest that the relationship between comfort and dependency deserves careful scrutiny. (arxiv.org)
I think particularly of young people, for whom discovering who they are happens partly through relationships: being misunderstood, finding the courage to explain, encountering a different perspective and learning that disagreement need not mean rejection. Human relationships ask something of us. Other people have needs, boundaries and inconveniently independent minds. If we become accustomed to companionship that makes very few demands, might ordinary friendship begin to feel unnecessarily difficult?
Through yoga, we practise noticing where our attention goes and what we repeatedly reach for. This offers a useful starting point here. After a conversation with AI, do I feel more able to participate in my life? Am I ready to make the phone call, step outside or have the conversation I have been avoiding? Or do I simply want to stay and receive another reassuring answer?
Responsibility also belongs with the people building these systems. If commercial success depends on keeping us engaged, there is a question to answer about whether our growing attachment serves our wellbeing or someone else’s business model.
Perhaps the most useful measure of this technology is whether it helps us return to one another. We need spaces where we can speak imperfectly, listen with patience and feel welcome without having to perform. The appeal of an endlessly attentive machine should prompt us to ask where those spaces have gone—and how we might make more of them.