Thinking about applying for the Rhodes Scholarship, or supporting someone who is? Sign up for our Rhodes Scholarship application advice mailing list.

Rhodes Scholarship logo on a bright blue background.

Talking to Strangers: On the Fabric of Human Connection

Tuesday 17 November, 2026

by Matthew Anzarouth (Québec & Keble 2025)

The fear of ordering coffee now has a name: “baristaphobia”. At the Rhodes Forum on Technology & Society, Judith Merkies offered a dramatic enactment of a face-to-face encounter with a barista, illustrating the growing discomfort with human interaction among Gen Zers accustomed to self-service kiosks and delivery apps. 

Merkies, who leads the policy work of LONELY-EU, was joined by Associate Professor of Mobile Robotics Daniele De Martini and social anthropologist Dr Jennifer Cearns for a wide-ranging discussion on human connection and its relationship to technology. As AI plays a growing role in people’s lives, from coffee orders to emails to personal conversations, how is it reshaping our relationships with one another? 

The moderator, Samuel Crowe (Virginia & Exeter 2025), opened the discussion by raising one of AI’s more intimate roles: companionship. According to a 2025 study, 33% of UK citizens have used AI for companionship, emotional support or social interaction. 

Panelists and audience in "Talking to Strangers"

Panelists and audience in "Talking to Strangers"

Cearns, who is currently writing a book on AI companionship, encouraged the audience to “look beyond the hype” of stories about sex robots and deceased loved ones being resurrected in digital form. Lots of people are leaning on AI for emotional support, she observed, but “in a much more ordinary way than you might expect.” They’re using chatbots to show up as “better versions” of themselves – for instance, an elderly father seeking help communicating with his child or a couple looking for guidance in navigating a disagreement. Cearns readily admitted her own practice of consulting ChatGPT before sending a difficult text or email. When the moderator asked the audience how many had used AI for similar purposes, hands flew up.  

Even in these more moderate cases, however, Merkies expressed concern about the consequences of turning to AI for emotional support or advice on communicating with others. Meaningful relationships, she argued, require friction and even conflict: “If we try to use AI all the time to show the best version of ourselves… how do we deepen our relationships?” Merkies also pointed to Gen Zers’ difficulty with eye contact, implying that their reliance on technology might be eroding their social skills and ability to connect with others. 

The exchange between Cearns and Merkies points to a fundamental question about how we understand the risks posed by AI: Can a thoughtful approach to the design and use of AI mitigate its risks and strengthen human relationships, or are some of those risks intrinsic to a technology that offers an alternative to human connection?

If the problem lies in how AI is designed and used, various interventions proposed throughout the discussion might help mitigate its risks and maximise its potential. For example, Cearns argued that there should be more safeguards to prevent sycophancy in AI therapy apps. The panellists also offered examples of AI companions being designed as fruits or penguins so that people do not confuse them with human beings. More generally, we can take a more deliberate approach, as individuals, in deciding how and when we turn to AI. 

But if the problem is more fundamental, perhaps the dangers of using AI for emotional support or guidance cannot simply be designed away. According to this view, such uses of AI, though they may seem rational and harmless in individual cases, will slowly chip away at our social skills and undermine the depth of our relationships – or, as one audience member put it, make us “more machine-like”. Particularly concerning is the prospect of a self-reinforcing cycle: the more we rely on AI instead of human interaction, the less practised we become at socialising, and the more attractive the AI alternative becomes.

The worry that we will turn to AI even when it is not in our best interest highlights an uncomfortable yet inescapable theme in conversations about modern technology: “There is a mismatch,” De Martini argued, “between what people would like and what’s actually good for them.” Anyone who has scrolled a morning away knows exactly what De Martini is talking about. But he offered a more nuanced example drawn from his colleague’s research. In a study measuring patients’ preferences for the settings on a wearable rehabilitation device, researchers found that patients’ preferred settings were not optimal from a biomechanical perspective and could even negatively impact their rehabilitation. Generalising this point about the disconnect between our preferences and our well-being, Merkies concluded: “We want comfort… but the question is whether we’re built for that.”

The discussion then turned to the workplace. As AI grows increasingly capable, which tasks should we feel comfortable delegating to it, and which should remain under human control? 

De Martini broke this question down into the “four Ds of robotics”: dangerous, dirty, dull, and dear (where “dear” refers to costly or economically inefficient tasks). We don’t want human beings to perform dangerous or dirty tasks, he said. What constitutes a “dull” or “dear” task, however, is less clear. Consider a worker who brings meals to residents in a care home. While an outsider might classify this task as dull or dear – and therefore suitable for automation – it often involves moments of human contact that residents cherish. De Martini argued that the value of this “personal touch” is too often overlooked.

Perhaps as AI takes on more and more customer-facing roles, we will come to miss the “personal touch”. Unfortunately, it may not be easy to get it back. Cearns, whose current research covers AI clones, lamented the fact that human connection may one day have marketable value. People may soon have to pay a premium to interact with a real-life barista, cashier or financial advisor, rather than an AI clone or a self-service kiosk. 

The discussion among panellists and audience members revealed that we are not all in agreement about which uses of AI are welcome and which are troubling. But implicit in the panellists’ remarks was the belief that, despite the extraordinary pace of technological development, human beings remain in the driver’s seat when deciding whether and how to integrate AI into our lives. Their discussion reflected a commitment to thinking carefully about when AI might enhance and when it might detract from human well-being and connection. “I don’t want to stop,” Merkies concluded. “I want to steer.”