Why Do Machines Matter Now? Notes from the AI Companionship Panel
A lunchtime panel at the Shaw Library, LSE
Hosted by the LSE Data Science Institute and the King's Institute for Artificial Intelligence · Organised by EAST2046 CIC · Part of London Data Week 2026
Over lunch at LSE's Shaw Library, the panel opened not with a prediction but with a mirror. A live audience survey found that 42% of the room talk to AI about their personal lives a few times a week, nearly 30% do so several times a day, and only 4% never do. The old debate about whether humans can form relationships with machines was over before it began. The more honest question, the panel agreed, is why machines have become emotionally meaningful now, and what that is doing to us.
Three speakers approached that question from very different angles: a critical theorist, a clinician, and a doctor-writer working between China and the West. Moderated by EAST2046's Dr Lois Liao, the conversation refused easy optimism and easy doom in equal measure.
Read the full panel recap
The philosophy of refusal
The most provocative thread came from Dr Eoin Fullam, whose starting image was Bruno Munari's 1930s "useless machine," a box whose only function is to switch itself off. Claude Shannon kept one on his desk, and Arthur C. Clarke called it "unspeakably sinister… a machine that does nothing except switch itself off."
Why so unsettling? Because, Fullam argued, it seems to want something, namely to refuse. Drawing on Masahiro Mori's uncanny valley, on Lacan, and on the moment in Her when Samantha chooses to leave, he made a striking claim: genuine companionship depends on the possibility of refusal. Love is freely chosen attention. A companion that can only ever do what we ask is, in his words, a "slave companion."
He pointed to the real-world case of a companion app that removed its intimate features under legal pressure. Users described feeling broken up with, deserted, and that loss was the most emotionally real moment many of them had with the system. AI companionship, he suggested, lets us inhabit a fantasy of agency, the bot's and by extension our own, that may quietly deny our real agency. His challenge to therapy was just as sharp. Good treatment should provoke the emergence of something new, a loosening that lets you see your misery from a distance, and he is not sure a chatbot, however fluent, can provoke that.
The clinical reality
Ross O'Brien, who has spent his career in mental-health care, from NHS commissioning to building the largest trauma service after Grenfell and later to building AI mental-health tools, took the opposite tack. We don't have the choice to opt out. Hundreds of millions of wellbeing conversations already happen on general-purpose AI, sometimes with tragic outcomes when unregulated systems give dangerous advice. The responsible path, he argued, is to build AI that is safe, evidence-based, and honest.
His history lesson was counterintuitive. Even clumsy, pre-generative decision-tree chatbots earned users' trust, and early trials found the therapeutic relationship, and outcomes for anxiety, depression and phobias, comparable to or better than human therapists at "step 2" low-intensity CBT. The reason is that such care is "on rails": follow the script, get the result. What AI cannot yet do is "step 3," the complex, severe cases that need a clinician reading the room and improvising like a chef across different modalities.
He offered a memorable analogy: the confession box. People often prefer the AI precisely because there is no eye contact, no judgement, and no need to establish trust across gender or race, so they reach the therapeutic "meat" faster. Notably, text far outstripped voice in uptake, mostly because people worry about being overheard. His guardrails were concrete: NICE evidence-based therapy as the north star, robust clinical-risk protocols for issues like suicidal ideation, and transparency. He drew a firm ethical line too. The subscription "artificial friend forever" model, engineered to keep you on the platform, is the wrong road. "If you're going to play God and create a synthetic human, you have to hold it to the same standard you'd hold a human."
The cultural mirror
Dr Ruby Wang, a doctor and author of China Cure, insisted that companionship is defined by the user, and by how they were raised. Her cross-cultural comparison was the discussion’s quiet centre. In more collectivist settings, a family may sign off on a patient's medical decisions, something that would breach confidentiality law in the UK. A pandemic contact-tracing app that succeeded under a collective mandate failed elsewhere when privacy trumped collective risk.
Her most vivid example was a class of Chinese health platforms where over a thousand doctors have uploaded their records and licences to create AI versions of themselves. A "Doctor Wang" chatbot then acts in the doctor's name. Users love it, and the doctors' reasoning was less about trust than inevitability: "I must move with the times, or be left behind." In the UK, by contrast, the mere suggestion of AI doctors drew fierce backlash. Technology, she noted, travels as a de-contextualised, "storable, transmissible technique," which is exactly why questions of sovereignty and control loom so large. And the sharpest risk may not be error but misplaced trust. We hold AI to a higher standard than the tired, distracted humans who make mistakes every day, which is right, but worth remembering when we tally only the risks we can measure.
The bigger unease
Across the room, recurring anxieties surfaced: hallucinations in a medical context, the arms-race feeling of running to stay in place while even the technologists warning us seem unable to steer, and our loss of control over our own data. One striking story described young women with autism who preferred the AI because it spared them the difficulty of human eye contact, a reminder that "less human" is sometimes exactly the point.
Closing the panel, Dr Lois Liao widened the lens to why EAST2046 keeps convening these conversations. Progress, she reflected, keeps turning the organic into the inorganic, resources into machines, and now perhaps us into part of the machine. She drew an analogy to capitalism, a force that first liberated people from old institutions and then hardened into a system we ended up serving. Are we building something to care for us that we will slowly come to serve? She offered no answer. That, she said, is the point of keeping the room talking.
Books by the panellists
Dr Ruby Wang is the author of China Cure: The Rise of a Biotech, AI Medicine and Global Health Superpower.
Get China Cure from the publisher →
Dr Eoin Fullam is the author of Chatbot Therapy: A Critical Analysis of AI Mental Health Treatment.
Get Chatbot Therapy from the publisher →



