Artificial intelligence and the fears of academia

At the beginning of September, during the presentation of the new GPT-6 Astra model, OpenAI announced the emergence of so-called artificial general intelligence. This is supposed to be a level of technological development that would bring language models closer to the mental capabilities of a human being himself.

Although this declaration was also met with skepticism from AI industry experts, further events heightened the excitement around the potential capabilities of artificial intelligence. First, OpenAI announced that their language model had solved one of the "millennium problems" (i.e., a fundamental mathematical challenge for science). Then, a new piece of information spread on social media: mathematicians who were close to solving the same problem and also used GPT models in their work expressed concern that the breakthrough announced by OpenAI was achieved thanks to their efforts.

Let us also add that for some time now, statements from former employees of the company have appeared more and more frequently in the media, claiming that the models created by OpenAI are dangerous and that in the future, as fully autonomous systems without human control, they could pose a threat to humanity.

There has already been a lot written about the dangers associated with the development of artificial intelligence. Progress in this field is linked to a serious environmental footprint, as data centers necessary for training language models require enormous amounts of energy and water. Questions about privacy remain unanswered, since users upload to language models whatever they wish—from family photos to medical test results—and all of this becomes part of a vast data magma used by Big Tech companies.

Finally, there are sometimes less, and sometimes more serious questions about safety and the future—how to implement control systems when such powerful technology is developed by private corporations that are not subject to anyone’s control?

As a sociologist, I am unable to assess these threats. I have no idea whether in 10 years we will wake up in a world straight out of Terminator. I lack the competence to judge whether warnings that uncontrolled AI development will lead to language models taking over the world are based on real processes or are simply expressions of technophobia characteristic of times of scientific-technological breakthroughs. However, I can ask about the nature of these fears, as they tell us something very interesting about the social emotions that technological development provokes.

AI and academia

This is a particularly interesting issue because just a few weeks ago, we had our own Polish controversy around AI—namely the case of the (in)famous report Poles’ Self-Portrait and Prof. Barbara Mroza-Gorgoń. The scale of reactions and the emotional temperature it generated are remarkable. Besides the usual online lynching (I write "usual" because in the era of polarized social media, such strong reactions are nothing new), many critical voices against the former government plenipotentiary for promoting the Polish brand came from academic circles. The spokesperson of the Conference of Rectors of Polish Academic Schools (an organization representing leaders of Polish universities) demanded an apology from Prof. Mroza-Gorgoń and considered her words, that "every second professor in the world uses" artificial intelligence, inappropriate. On social forums, I read many critical voices from academics, some even describing it as a disgrace to Polish science.

Where does such strong criticism come from? To answer this question, we need to realize what artificial intelligence truly is from the perspective of science as a knowledge-producing domain—and from the perspective of scientists as people specialized in producing that knowledge.

AI and the prestige game

Enthusiasts will say that it is a tool that facilitates work. I should admit here that I myself use language models—mainly for literature review, which is a tedious but necessary process if one wants to publish in reputable, English-language scientific journals. Language models significantly accelerate this process; their paid versions almost no longer hallucinate and can select scientific articles relevant to my topic, saving me from reading dozens of texts to find a few that fit my research. Artificial intelligence also allows me to translate into English what I have written, which is a significant advantage for someone representing semi-peripheral science who must compete with English-speaking scientists from core countries.

There are other, less controversial applications of AI in science, such as automating data collection (e.g., from websites when analyzing discourse) or preliminary or in-depth data analysis (e.g., creating initial categorization keys and coding qualitative data). Those who submit grant applications also use language models—AI works excellently as an assistant that suggests how to write a proposal to increase chances of success.

Critics of AI emphasize that overusing language models leads to a significant inflation of scientific output. According to one analysis, scientists using AI tools publish on average three times more scientific texts and can expect nearly four times more citations. Another study shows a significant increase in the number of grant applications submitted in the prestigious Marie Sklodowska-Curie Actions program between 2022 and 2025. All this makes it easier for scientists to write texts or grant proposals, but increasing competition makes it harder to publish those texts and to secure grants.

The best scientific journals and the most prestigious grant agencies are flooded with well-written proposals. This increased supply of scientific knowledge is not matched by increased demand. The top journals publish the same number of articles, and agencies have the same budgets, so the success rate (the ratio of accepted articles or grant applications to rejected ones) steadily decreases.

Peer review and its challenges

This also raises questions about peer review, a key activity in publishing research. Good research must undergo verification and be checked by other specialists in the field before it can be presented to a broader audience in a scientific article. Double-blind peer review, where reviews are anonymous and performed by other scientists, is the standard in most scientific disciplines. However, the inflation of scientific publications makes it increasingly difficult for scientists to find time to carefully read, thoroughly evaluate, and write reviews (step-by-step justifying their assessment).

Meanwhile, there is evidence that language models are often used not only to write scientific texts but also to evaluate them. We are thus dealing with a situation where artificial intelligence writes articles for scientists and then also assesses those materials on their behalf.

Prestige in science

All these negative phenomena are certainly what those who joined the almost unanimous chorus criticizing Prof. Mroza-Gorgoń had in mind. I do not want to question their good intentions here. AI can indeed be a major threat to science, and its tools must be used with extreme caution. The criticism directed at the co-author of the report is largely justified. Nevertheless, the radicality of this reaction is surprising and perhaps conceals an unconscious fear—not so much of the direct consequences of technological development for specific activities, but of losing status.

My hypothesis is as follows: the emotions related to the case of the former government spokesperson for Poland’s promotion abroad are a manifestation of fear of a radical change in the rules of the academic game based on individual prestige.

From an outsider’s perspective, scientists do what they do to make life better for all of us. Scientific knowledge serves the interests of all people (e.g., to help us live longer, healthier, and more comfortably). In the classic view of sociologist Robert K. Merton, elements of the academic ethos include, among others, communism and disinterestedness—that is, the belief that knowledge is a common heritage of all humanity and that scientific activity should be driven by the pursuit of truth, not particular interests. However, this is a largely idealized picture, as many critics of Merton have pointed out.

The changing game of prestige

Of course, I do not deny that scientists conduct research for the good of humanity. I rather mean that individual prestige remains an extremely important motivation for them. It may be the most valuable currency in the world of science. As another sociologist, Pierre Bourdieu, argued, in science—as in other areas of social activity—power is at stake. Power in science depends on the amount of prestige, which includes, among other things, publications in prestigious journals, projects funded by renowned grant agencies, or the number of citations.

This accumulated prestige can later be exchanged for other resources (e.g., a highly prestigious scientist can earn more) or used as capital to succeed in non-academic contexts (e.g., a scientist with high prestige can become a well-known media authority or engage in politics). At the same time, accumulating this currency—prestige—can be very difficult and require many sacrifices, including personal ones. Conversely, established scientists—those who have already gained a lot of prestige—find it easier and reach their life goals more quickly.

For such a form of prestige, it is crucial to attribute a specific scientific achievement to a particular person. Prestige can always be associated with a specific scientist in the field of science. Prestige is someone’s; otherwise, as a currency, it makes no sense. That is why scientists attach great importance to linking achievements to specific names—history, after all, records many disputes over the priority of a publication, discovery, or patent. Another good example of this mechanism is the discussion about the order of authors in scientific publications—whoever is listed first always has the most prestige, the second a bit less, and so on. I have personally witnessed at least a few heated debates about who should be credited first.

A good example is also the aforementioned dispute over the priority of solving the millennium problem. One might ask—who really cares who made the breakthrough first, since ultimately what matters is the increase in knowledge? Unfortunately, in science, it’s not just about that. Scientists are largely driven by the desire to do something significant that will ensure their name is recorded in history. We can evaluate this differently, but this is often how it looks.

Changing the rules of the prestige game

And it used to be like that, because the widespread implementation of language models in science is redefining the rules of the academic game. AI allows for much faster accumulation of prestige in the form of articles, grants, or citations. As I pointed out earlier, there is already a sharp increase in the number of texts submitted to top scientific journals. Moreover, the scientific publishing market has adapted well to these circumstances—more and more, "mega-journals" are gaining popularity, publishing dozens of texts daily, with rapid review processes and high citability. Publishers like MDPI or Frontiers are becoming a natural ecosystem for AI-supported science.

The effectiveness of using AI in scientific work means that the path to achieving the status of a recognized scientist is becoming easier (I am ignoring here the quality of such knowledge; that’s a topic for a completely different discussion). In the traditional model, at least in most cases, accumulating prestige was rather slow. In the AI era, everything accelerates. A scientist can quickly gain relatively large resources of prestige and use them across various fields, competing with those who "built" their careers under the old rules.

As you can easily guess, such a situation inevitably leads to conflict between those who know how to use language models and thus build their position in the "shortcut" way, and those whose efforts and time were required to reach their current status. I believe that this is what we saw in the reaction of the Polish academic community to Prof. Mroza-Gorgoń’s case.

Of course, the Poles’ Self-Portrait report itself is an excellent example of how not to use AI. At the same time, it is also a preview of how the academic game for prestige might look in the future and how those who are now part of the academic elite may be unable to adapt to new rules. Prof. Mroza-Gorgoń’s case exemplifies what is inevitably coming—an upcoming future that appears not only as a significant change but as something overturning existing hierarchies.

AI and the sense of alienation

AI also makes it extremely problematic to answer the question "who is the author of the research/patent/discovery/article," striking at the key— as I mentioned above—dimension of academic prestige, which is authorship. In the traditional model, it was quite straightforward—one could easily attribute discoveries to individuals. And although, as the brilliant Ludwik Fleck showed nearly a hundred years ago, groups and collectives are the real creators of scientific discoveries, according to the rules of the academic field, claiming "I am the author of this discovery" is a crucial move in the game for prestige.

Let’s take again the example of solving the millennium problem. Two mathematicians, who suggested that OpenAI might have "stolen" their breakthrough, based their work on language models. These models belonged to OpenAI, so perhaps the scientists themselves contributed to their own failure.

Who is the author of the breakthrough? The mathematicians who probably couldn’t have done it alone and used AI? Or the owners of OpenAI, who may have used prompts prepared by those mathematicians to improve their models? Or perhaps artificial intelligence itself? But AI is based on data previously created by humans, on the entire scientific legacy, and prompts are also formulated by humans.

AI, of course, relies on knowledge uploaded into the language model by humans, so ultimately, a living person is the author of the ideas that the LLM suggests to us. However, there is a problem here related to the multitude of references generated by AI and the speed at which they are assimilated. In traditional academic work, understanding and using others’ ideas takes time and, in a sense, transforms the person who benefits from others’ scientific achievements. To quote a text, I must first read it, understand it, and be inspired by it, then creatively use it, which in turn changes my way of thinking. When using AI, this process accelerates greatly, and the mechanism of absorption (adopting others’ ideas to say something of my own) essentially does not occur. There is no time for others’ ideas to become my own.

In my opinion, this creates a sense of alienation from the product of one’s own intellectual work. If I write an article and AI makes corrections or suggests extensions I wouldn’t have thought of, who is ultimately the author of that article? If AI starts juggling concepts and theories I didn’t know and only assimilated from summaries, do I use them consciously (i.e., in accordance with my own, authorial intentions), or do I rather let myself be led by someone else in my thinking? As I mentioned earlier, the ideas circulating in LLMs probably have their authors. Nonetheless, the experience of using AI in academic work can lead to an uncomfortable feeling of a lack of subjectivity among scientists.

In the times before AI, I, as a scientist, thought, discovered, presented, and negotiated the prestige I gained during my career. In the era of AI, this is no longer so obvious. This, in turn, can lead to existential questions. If it’s no longer just me thinking, just me being the author, then perhaps the scientist is no longer needed by the world? Maybe the development of AI in the future—when the capabilities of language models will be much greater than now—will cause an irresistible awareness of being superfluous. This could be unbearable for intellectuals for whom subjectivity, and consequently individual authorship as an essential element of prestige, is a crucial part of their professional identity.

Returning to the case of the former government plenipotentiary, Prof. Mroza-Gorgoń not only broke certain rules of the prestige game. Although she played the role of a trained expert herself, her example shows that intellectual work of such a specialist does not need to be done by her and can be delegated to a relatively cheap and easy-to-use device.

The exaggerated reaction to Prof. Mroza-Gorgoń’s embarrassment perhaps expresses a fear of a future where individual authorship and the subjectivity of scientists become relics. A future where discoveries are attributed not only to scientists but also to corporations and AI models. A future where scientific careers (if they even exist) will not be built on the current principles. A future where thinking, knowledge, and intellectual skills will cease to be important indicators of social status.

We are all Platos

To better explain where scientists (in Poland and worldwide) currently stand, it is worth referring to the key cultural transition from orality to literacy. The invention and dissemination of writing was a technological revolution that changed many aspects of social life—from trade and science to politics. However, not everyone was an enthusiast of this technological novelty. Among its fierce critics was Plato, who believed that writing damages memory and exposes philosophers to the risk of misunderstanding (the speaker can defend what they say; writing, once circulated, lives its own life beyond the creator’s control). Plato’s stance can be read as more than simple technophobia: it is the attitude of an elite representative who sees that the rules of existing social hierarchies are being brutally changed, and their own competencies will no longer be as valued as before.

The fear underlying some of the intense reactions to Prof. Mroza-Gorgoń’s case is an emotion similar to Plato’s. Of course, there are many AI enthusiasts in science, and I do not deny that. Yet, in the academic world, there are also many opponents who, besides valid objections, fear losing both their prestige and important elements of their identity.

On one hand, AI probably damages some skills, fosters laziness, and perhaps leads us into a dead end. Already now, it is pointed out that the data fed into AI are generated by itself, causing the entire system to operate in a kind of idle mode. On the other hand, the rapid development is accompanied by a fear-driven negation, which is both a manifestation of fear and an attempt to halt potentially irreversible (perhaps) changes that turn the existing world upside down.

**
Michał Wróblewski – sociologist, philosopher, works at the Institute of Sociology at UMK in Toruń, interested in the sociology of medicine and environmental sociology. Currently conducts research on vaccine skepticism and activism related to air quality. Author of, among others, Whose fears? Whose science? Structures of knowledge in the face of scientific-social controversies (together with A. Nowak and K. Abriszewski, Poznań 2016), Sociology of epidemics. Emerging infectious diseases from a social sciences perspective (together with Ł. Afeltowicz; Toruń 2021).

The post Artificial intelligence and the fears of academia first appeared on Krytyka Polityczna.