Are we pretend thinking – Why scholarly communities still matter in an AI world
During a coffee break at a recent academic gathering, a colleague mentioned they needed to ‘”double-check a theory with Sara”. Only later did I realize that “Sara” was not a co-author, but a customized AI bot.
Is the real AI revolution happening not only in our labs – but also in our minds? Beyond functioning as a tool, artificial intelligence is becoming an active participant in how we think. This fluid reality directs to something deeper than the usual concerns about AI in academia. The central question is no longer whether AI helps or harms research. The more difficult question is what happens to researchers when we begin to think with AI systems on a daily basis.
Insights from the EERA NW01 Spring School
This question became increasingly present during the EERA NW01 Spring School at West Universty of Timișoara, where discussions moved past technical debates about academic writing and turned instead to researcher identity, responsibility, and professional learning. The event was very productive, bringing together emerging scholars from diverse countries and contexts, at different stages of their academic journeys. The collaborative and collegial atmosphere that emerged allowed for the kind of honest, plural dialogue that questions about AI in research genuinely demand.
A recurring tension surfaced across many of the conversations: AI offers real practical advantages, as it accelerates literature searches, organises information, supports multilingual writing, explains complex concepts, helps structure arguments, and assists with coding and data analysis. These tools can feel close to indispensable and their use in academia is, of course, not new especially for researchers working under constant pressure to publish, teach, apply for funding, and manage international demands.
Yet, a quieter transformation lies beneath this efficiency. The concern is not so much that AI may produce hallucinated references or inaccurate interpretations, also because researchers are already learning to carefully verify outputs.
The deeper challenge is subtler – AI changes the conditions under which thinking itself takes place.
Shared authorship – or pretend thinking?
In this context, having three days for sustained discussion, reflection, and hands-on engagement-including working with AI tools and drafting AI statements proved invaluable for helping emerging scholars develop the confidence to navigate what responsible academic practice now demands.
While the researcher remains responsible for the final work, parts of the thinking process become increasingly externalised and negotiated through interactions with AI systems. In practice, authorship is no longer only about who writes the text, it is about who validates, questions, verifies, and assumes responsibility for the knowledge being produced. Authorship is now, in some sense, shared with a digital “assistant”, one that some already address by a human name, few want to openly acknowledge, and many feel they cannot survive without.
When you hear colleagues say, “I’ll check with Sara,” consider that this Sara might be their AI confidant.
This transformation becomes particularly visible in moments of overreliance. When the difficult work of methodological and conceptual reasoning is handed over to AI, learning risks becoming more surface than substance – what me might call “pretend thinking”.
As discussed during the Spring School, one of the significant risks for early-career researchers is in fact surrendering too much intellectual agency to technology. The apparent fluency of AI-generated outputs can create an illusion of competence before real understanding has fully developed. In this sense, AI is not replacing researchers; it is reshaping what being a researcher means.
So how do we, as emerging educational researchers, navigate our careers in this new landscape?
Epistemic literacy: Questioning what AI leaves out
The challenge, then, is not simply technical literacy. AI literacy in academia must involve epistemic literacy: the capacity to question outputs, recognise limitations, understand how knowledge is being mediated, and keep attentive to questions of power, authorship, and intellectual responsibility. This is especially relevant in educational research and multilingual contexts.
AI tools are predominantly trained in English and shaped by dominant epistemological frameworks, raising uncomfortable questions about voice, linguistic diversity, and the future of knowledge production. When we place uncritical trust in systems shaped by dominant cultural datasets, we may be marginalising local perspectives, minoritised languages, and alternative epistemologies. This may be revealing how fragile and partial our thinking remains.
And this is where Sara becomes more than a curiosity. If we consult her daily, it is worth asking:
- Where does Sara’s knowledge come from?
- Whose frameworks shaped her answers?
- What did she leave out and what did we stop questioning, because she replied so quickly and so confidently?
Recognising that Sara might be biased is not a reason to stop using her. It is, however, a reason to remain alert. Researchers who can articulate both why they rely on AI and why that reliance troubles them are not researchers without a position. In fact, they are researchers whose position is still being built, carefully and with awareness of its own limits. Epistemic ambivalence, in this sense, is not paralysis – it is a sign that reflection is genuinely underway.
Why scholarly communities matter more than ever
There is also a more insidious risk: AI may encourage a false sense of self-sufficiency, suggesting that we no longer need to seek each other out or work in genuine proximity.
While it is often faster to get an answer through AI, efficiency does not equal long-term quality. Perhaps this is the real challenge now facing researchers – not learning how to write with AI but learning how to remain intellectually present while doing so. This is precisely why preserving human engagement through supportive scholarly communities remains essential.
Because some questions, the ones that matter most, still need to be checked with a person from your scholarly community, not a ‘Sara’.
Key Messages
-
AI has changed the conditions under which thinking takes place which reinforces the need for honest, plural dialogue to support intellectual agency.
-
One of the significant risks for early-career researchers is surrendering too much intellectual agency to technology.
-
Epistemic literacy is mandatory in an AI word: the capacity to question outputs, recognise limitations, understand knowledge mediation, and even identify power imbalance and authorship questions.
-
Intellectual responsibility is crucial to support scholarly integrity in the world of AI.
-
Preserving human engagement through supportive scholarly communities remains essential in an AI world.
Emily Dobrich
Western University, Canada
Cristina Domingues
University of Aveiro, Portugal
Cristina Domingues, M.A., is a doctoral researcher in Education at the University of Aveiro and an FCT fellow at the Research Centre on Didactics and Technology in the Education of Trainers (CIDTFF), where she investigates linguistic empowerment and the co-construction of knowledge with adult migrants. Her participatory research, based on Future Workshops, is carried out in partnership with the Municipality of Fundão, a UNESCO Learning City. She is a language teacher and certified teacher trainer with 27 years of experience, and her research interests include plurilingual education, participatory methodologies, intercultural education and critical AI literacy.
ORCID: https://orcid.org/0000-0002-9076-8093
LinkedIn: www.linkedin.com/in/cibdomingues
