If this email forwarded to you, sign up to receive weekly letters for the examined intellectual life.

All previous editions of The Letter are available on our website.

— Written by San Choudhury

Seventy per cent of junior researchers and 64% of senior researchers in a 2020 Wellcome study on research culture believed that a career in research can be isolating and lonely. Across all 4,267 respondents, 84% of whom worked in academia, the corresponding figure was 69%. Within academia more specifically, a study published online in 2024, drawing on responses from 4,563 UK doctoral students and academic staff during the first COVID-19 lockdown, found that doctoral students reported greater loneliness than staff. For many currently in academia, particularly those pursuing doctoral studies, these figures possibly only confirm something they already know through experience: social isolation and an unsatisfied yearning for meaningful connection during the “emotional and multi-faceted journey of becoming a scholar.”

Existing research has documented several conditions that contribute to loneliness among doctoral students. First, the organisation of research work itself shapes opportunities for connection. While research conducted within established teams provides occasions to collaborate, exchange ideas, and develop relationships through the work, individually organised projects – one candidate, one project – understandably offer fewer such opportunities. Research has consequently drawn attention to isolation in the arts, humanities, and social sciences, contrasting these arrangements with research teams found in some lab-based disciplines. Still, there are variations within and between disciplines: working in a lab does not guarantee companionship, just as research in the arts, humanities, or social sciences does not inevitably mean working in isolation. 

Alongside the organisation of research are the institutional treatment of the doctoral student and the institutional conditions of academic research. In some disciplinary settings, doctoral students are treated as learners rather than full members of a research group, making recognition as a colleague less guaranteed. Competitive academic environments, a key feature of our neoliberal universities, make sharing difficulties or unfinished ideas feel risky. Separation becomes a strategy for protecting oneself, even as it makes connection harder. There are some bizarre accounts documented in Janta and colleagues’ article describing students’ attempts to socialise being met with suspicion or further exclusion. 

Funding, which determines both the number of doctoral places a department can support and the resources available to their researchers, likely only compounds these difficulties. Where funding is limited, small cohorts may be accompanied by fewer opportunities for students to meet and organise activities together. At my university, the Business School, with nearly thirty students in a single annual intake, had an established doctoral society that regularly organised events, conferences, and social gatherings. Students I spoke to in departments such as politics, English literature, and history described fewer opportunities of this kind; some barely knew other doctoral students. One English literature PhD student told me that only one other person had started alongside her in the department. Finding people who understand the experience can therefore depend significantly upon which department a student happens to join and what resources it has.

Against the backdrop of the systemic isolation and loneliness amongst doctoral students, AI chatbots become strangely appealing as a way for coping with and meeting their unmet needs for intellectual companionship and peer support. 

In contemporary discourse about ethical AI use in academic research, the chatbot typically appears as a “thinking partner”, “research companion”, and other such collegial roles. On Reddit, there is a thread titled “using AI more as a thinking partner than a generator”. Contributors to the thread describe using it to talk through their research to make sense of it, checking their understanding of what they are studying and how they are going about it. They talk about “pushing back on responses, refining the idea, testing it against real situations”, instead of simply prompting for answers. This “backing-and-forthing”, as one Redditor called it, sounds similar to how we might describe a conversation with another researcher. In a recent conversation, someone described it as a “rubber duck” process, where you talk about your research to a rubber duck – except that, with generative AI, it’s a rubber duck that talks back. The same metaphor was used in the Reddit thread. Meanwhile, a study of international doctoral students in Malaysia reported that students described using ChatGPT for emotional support: to vent, feel calmer, and work through academic problems without feeling judged. 

Isolated and lonely doctoral students want someone to talk to about their research. They want meaningful interactions about the journey, research life, and becoming a scholar, but they struggle to find them. They have limited access to peers sharing the same experience: someone who understands the journey, someone beyond their supervisor to turn to, someone with whom to talk through problems and ask questions without feeling judged. When something resembling that conversation becomes available, it makes sense that they turn to it. 

It is of little surprise that they seek “thinking partners”, “research companions”, and emotional support partners, and find some version of these in AI chatbots. When, as a result of institutional conditions, they don’t feel like they belong, aren’t accepted as colleagues, or keep themselves separate for fear of judgement, the appeal is hard to dismiss. They yearn for this kind of interaction, and here is something offering a version of it. What other options do they have when there is no one else available? A Redditor commented that human thinking partners would be great alongside AI, but:

“...then there's also the fact that people have availability issues. No only you need to find someone who's knowledgeable about the specific thing you want to discuss, but you also need to actually find the person (personally or via the internet), surpass possible language barriers, make sure the person is available and willing, make sure they're patient with you, make sure they're financially accessible (in case they're charging to discuss with you)... and so many other things.”

Indeed, in a study of university students more broadly, researchers raise the possibility that students might seek support from AI instead of people such as librarians or professors. Its sample was not specifically doctoral researchers, but this possibility is interesting to consider alongside an observation I’ve made about the type of social media content currently (not) trending in the academic niche. There was a time when academic content on Instagram was full of advice, tips, and explanations of how to do research. It became annoying, and to me it seemed to reduce learning how to research to an endless collection of instructions, leaving little room for learning through doing. Now, it seems to have become considerably less common. Its absence has made me miss it, but more than that, it's interesting to consider why it seems to have disappeared. I wonder whether part of the explanation is that this content no longer performs as well because people can get those answers from chatbots. Why wait for someone to explain how to develop a research question when you can ask a tool immediately? This is of course an observation, not evidence of a wider trend; content and platforms change for many reasons. But regardless, it leaves me wondering about the occasions for contact that might disappear alongside it. A research question answered by a chatbot is potentially one fewer question taken to another researcher; sometimes that’s a question that might have been the beginning of a relationship.

For students already vulnerable to loneliness and isolation, there is a risk that using AI chatbots could leave them even lonelier. This is because ultimately, while using AI may be useful, or perceived to be useful, it does not resolve the absence of companionship, meaningful relationships, and peer connection. At the end of the day, being able to ask endless questions is not the same as being known by someone, contributing to their thinking, and finding a place within a scholarly community. 

What’s particularly worrying in all of this is that in academic cultures where suffering is already treated as something doctoral students must endure, a tool that makes an unsupported environment easier to tolerate could also make the need for change easier to overlook. The student has somewhere to take the question, a place to vent, and something that responds. However, this does not change the fact that the department they ‘belong’ to may still offer few opportunities for meaningful exchange, and the student may still spend their days without anyone who knows how they are doing. There is a potentially grave institutional danger here: that chatbots help people cope with conditions that ought to change. After all, loneliness does not need to disappear altogether to become less visible. If researchers find private ways of managing it, the pressure to provide meaningful human support is only likely to weaken. Increasingly, an old problem is acquiring a technological workaround while the arrangements producing it remain intact.

Before researching this essay, I knew loneliness in academia through experience. Even so, I found the extent to which it has been documented – and its persistence – alarming. We have known about this for a long time, and I fear chatbots are only going to exacerbate institutional loneliness in academia precisely by making it a privately and invisibly managed condition. 

P.S. I wanted to end this essay by suggesting that the next time we have a question or want to talk something through, it is worth considering whether there is another person we could turn to before opening a chatbot. But given that I have just spent a considerable number of words developing an institutional account of loneliness in academia, I fear that this call falls flat. The reality is that sometimes there really is no one available, and that should not become another failure for the lonely researcher to carry. It should prompt a harder question about the communities in which we expect people to become scholars: why is a chatbot the most accessible conversation they can find?

If this essay stirred something, be it a question, a reflection, a story of your own, we want to hear it. Even if it’s something negative, it’s okay. The Scholarly Letter exists because we believe that intellectual life begins in conversation and communing. Respond to this email we want to hear what you have to say and share.