GenAI helps to plan trips, text and search for information and, increasingly, to interpret health symptoms. Doctors are already noticing this: patients come not only with their ailments, but with a ready interpretation generated by the chatbot – a so-called ‘diagnosisAI’. In business, the change is even more pronounced. Companies are using AI tools to create presentations, analyse data, generate content and code, thus taking over areas that for decades were the domain of human creativity and expertise. The scientific community is particularly affected by this process. The work of the researcher is based on analysis, interpretation, the construction of meaning and the creation of new connections between phenomena – precisely the areas where generative systems are developing most rapidly today. What is at stake is no longer just new tools, but the very definition of the role of the researcher.
Two worlds: delegatable and non-delegatable tasks
An interesting picture emerges from a survey conducted as part of the AIResearchers project, funded by European funds from the Erasmus+ programme. It shows that among academics – university managers, research teams and lecturers in Poland, Spain and Portugal – no simple division can be made between ‘enthusiasts’ and ‘opponents’ of AI.
The most important conclusion is simple: scientists do not reject AI. On the contrary – they understand its value very well. However, they see a clear line between what can be given to a machine and what must not be given.
On the ‘pro’ side are primarily repetitive, time-consuming and procedural tasks. AI is readily regarded as a high-quality administrative assistant: it helps to prepare grant proposals, project reports, abstracts, slides, teaching materials, literature reviews or preliminary analyses of large data sets. In the area of creating impact, text editing, preparing abstracts, graphics, presentations and educational materials were particularly highly rated.
The second area is levelling the playing field in global learning. For researchers for whom English is not their first language, generative AI can reduce the ‘native speaker’ barrier. It helps to refine the style, adapt the text to the requirements of the journal and enter the international publication circuit on a more equal footing.
The third area concerns the conceptual phase. Many researchers treat AI as an intellectual interlocutor – a tool to help organise thoughts, suggest alternative takes on a problem, test working hypotheses or look at an issue from a different perspective.
However, usefulness does not mean trust.
AI helps, but does not reassure
Researchers point to a whole package of concerns. The first concerns cognitive uncertainty. Generative models can produce answers that are smooth, elegant and convincing, but not necessarily true. Hallucinations, ‘nicely packaged falsehoods’, response instability and limited repeatability are particularly problematic in an environment whose foundation is verifiability.
The second concern relates to professional identity. If AI can generate a text, an analysis or a draft of an argument, the question arises: what actually remains as evidence of the researcher’s expert work? Does the technology enhance competence, or does it blur the line between genuine intellectual effort and a well-served tool?
The third barrier is more practical. Many researchers do not know which tools to use, how to secure them, what is allowed to be uploaded to the system and what is not. Added to this is institutional ambiguity. A lack of rules does not always mean freedom. More often it means anxiety – especially for junior researchers dependent on promoters, supervisors and promotion procedures.
Red line: ethics, relationships, responsibility
The sharpest line appears where ethics, evaluation, relationships and accountability come into play. Researchers do not want to delegate conflict resolution, mentoring, student assessment, decisions about research integrity, data protection, intellectual property or privacy to AI. In these areas, AI is not seen as a support, but as a risk of degrading the scientific process and diluting accountability. This sends a very important message. Scientists are not saying: ‘AI is bad’. Rather, they are saying: “AI can perform the process, but it cannot take over the meaning.”
It can inspire and organise, but it should not decide. It can help analyse the data, but the final interpretation, assessment of relevance and responsibility for conclusions must remain with the human. The ability of AI to delegate issues of ethics and transparency in the research process was particularly low on the list. In this sense, generative AI does not replace the researcher. Rather, it forces him or her to define more precisely where uncontrolled automation ends and responsibility begins.
Pressure that cannot be waited for
The problem is that higher education and science today operates under enormous efficiency pressures. Publish faster, analyse more, write better, win grants more often, prepare materials more efficiently. If AI realistically speeds up some of these processes, refusing to use it may start to look not like an ethical choice, but like a lack of productivity.
The winners, then, may be those who are quickest to integrate GenAI into their daily work as a permanent part of the workshop. Teams using the best paid models and institutional subscriptions may gain an even greater advantage. This creates another inequality: between those who have access to an exclusive ‘upgrade’ of their own capabilities and those who remain with free, inferior tools.
The paradox is thus obvious. AI democratises access to intellectual production, but at the same time it can create a new hierarchy – based on the quality of access to tools. And when everyone starts writing faster, analysing more efficiently and presenting better, the bar will rise again. As in the perverse quote from The Incredibles: “when everyone is special, no one is”.
So in the science of the GenAI era, the question is no longer, will researchers use artificial intelligence? The question is: what will remain a truly human contribution when technological enhancement becomes standard?
Author: Dr Gregory Banerski
