Artificial intelligence in education and science: imaginary threats or new opportunities?


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

Artificial intelligence in education and science:
imaginary threats or new opportunities?
Yaroslav Shramko

Kryvyi Rih State Pedagogical University, 54 Universytetskyi Ave., Kryvyi Rih, 50086, Ukraine

Abstract. This opinion paper, selected by the editorial board, is a translation of
a blog post originally published in Ukrainian on the Kryvyi Rih State Pedagogical
University website [1]. The author examines the integration of large language models,
particularly ChatGPT, into educational and scientific practices, arguing against
prevalent concerns about academic integrity violations and the erosion of traditional
scholarly methods. Drawing parallels with historical resistance to technological
innovations such as calculators and typewriters, the paper contends that artificial
intelligence represents a tool for enhancing productivity rather than a threat to
intellectual rigour. The author critiques current institutional responses, including AI
detection systems and mandatory disclosure requirements, as misguided attempts to
regulate what should be recognised as an emerging standard tool in academic work.
The paper advocates for the organic integration of AI technologies into education
and science, emphasising that these tools augment rather than replace human
critical thinking and creativity. Ultimately, the author argues that the academic
community should focus on teaching responsible and effective use of AI tools rather
than attempting to restrict or stigmatise their application.

Keywords: artificial intelligence, large language models, ChatGPT, academic in-
tegrity, educational technology, scientific research, technological innovation in
education

The opening of free access by OpenAI at the end of 2022 to the then-latest version
of the large language model (LLM) ChatGPT-3.5, capable of generating texts in a
form maximally approximating ordinary human language, had consequences that
can arguably exert and undoubtedly are already exerting a substantial, and in some
respects revolutionary, influence on the further development of education and science.
Almost immediately, similar models began to be offered by other developers, including
Google (Bard, now Gemini), Anthropic (Claude), Microsoft (Bing Chat), and Meta
(LLaMA). Since then, intensive discussions have unfolded regarding the role and place
of such systems in the educational process, their significance for scientific activity,
and possible challenges in these and other fields arising from their application.

Overall, we must note that the societal reaction to the emergence of cutting-edge
technologies for operating with linguistic structures has proved ambiguous: from
enthusiasm about new possibilities to alarmist predictions about the “end of writing
practices”, as supposedly henceforth all texts will be created automatically without
human participation. Such sentiments, however, are not new. The history of human
culture convincingly demonstrates that the emergence of new technical means has
almost always been accompanied by waves of scepticism. Thus, not so long ago,
the appearance of calculators was similarly discussed in pedagogical circles: might
this not become an “unfair” substitution for a pupil’s own efforts? Fears were voiced
that children would unlearn how to perform calculations independently, that basic
skills would lose their value, and that mathematics itself would cease to be an
instrument for developing logical thinking. In various countries, there were even

� 0000-0003-4843-0328 (Y. Shramko)
# shramko@rocketmail.com (Y. Shramko)
~ https://kdpu.edu.ua/shramko/yse.htm (Y. Shramko)

CTE
Workshop 
Proceedings

© Copyright for this article by its authors, published by the Academy of Cognitive and Natural Sciences.
This is an Open Access article distributed under the terms of the Creative Commons License Attribution
4.0 International (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any
medium, provided the original work is properly cited.

1

https://doi.org/10.55056/cte.1140
https://orcid.org/0000-0003-4843-0328
mailto:shramko@rocketmail.com
https://kdpu.edu.ua/shramko/yse.htm
https://acnsci.org/cte
https://creativecommons.org/licenses/by/4.0/deed.en
https://acnsci.org


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

heated debates about whether to allow the use of calculators in schools, especially at
the initial stages of education. For many educators, the technological novelty seemed
a threat to traditional teaching methods and even an undermining of the foundation
of mathematical education.

Today, however, it is obvious that these fears were unfounded. The calculator did not
eliminate the need for mathematical thinking, nor did it abolish the ability to compose
equations or prove theorems. It merely freed us from routine operations—endless
calculations on paper that distracted from the main point. Ultimately, calculators were
organically integrated into school practice: first, the pupil masters basic operations,
and only then, at more complex levels, uses the instrument to accelerate calculations.
Thus, technology did not replace knowledge but allowed working with it more efficiently,
opening time and space for a deeper understanding of mathematics.1

Similar discussions accompanied other technological innovations. Typewriters were
considered a threat to “genuine writing”, computers—a factor that would supposedly
deprive pupils of the habit of performing tasks by hand and systematically practising
skills, whilst the internet—a source of easy copying that undermines intellectual
honesty. Yet each of these technologies eventually became an integral part of the
educational process, opening new possibilities for teaching and research.

The situation is similar with the application of large language models like ChatGPT,
which almost immediately acquired the name “artificial intelligence” (not least thanks
to the very name of the company OpenAI, which opened the path to life for these
systems). The name “artificial intelligence” sounds appealing, but it does not quite
correspond to the essence of these systems. For the average user, it creates the
impression that we are dealing with genuine thinking or even machine self-awareness.
In reality, large language models possess neither intentions nor understanding in
the human sense. They work based on statistical patterns identified in data arrays
and perform tasks by selecting the most probable linguistic continuations. In other
words, we are dealing not with intelligence as such, but with a sophisticated imitation
of intellectual acts. Yet this very imitation is often so convincing that it creates the
illusion of genuine thought, especially when the user observes how such models
imitate human communication style, maintain dialogue, and respond to queries in a
way that gives the impression of genuine thinking.2

In the case of artificial intelligence, we are dealing not with a thinking being but
with a computer programme that can be used as a certain tool. The very term “intelli-
gence” in this name is metaphorical and can be misleading, creating the impression of
autonomous rationality. What we are essentially dealing with are software systems
capable of generating texts, generalising and classifying data, and imitating dialogue.
This is more an algorithmic mechanism than “intelligence” in the human understand-
ing. One can draw a parallel with the metaphor “computer memory”: it is not memory
in the psychological or biological sense, but merely a technical method of storing
information. The question is for which tasks this tool is truly effective and how this
tool can and should be used in educational and scientific activity.

Such systems manifest themselves best in cases involving standard text process-
ing. They are capable of quickly creating drafts, selecting examples, forming brief
summaries, carrying out translation, or explaining basic concepts. Moreover, artificial
intelligence demonstrates significant potential in editing: it helps clean up linguistic
flaws, give text greater stylistic fluidity, and adapt it for a specific audience. Such
functions free the user from technical routine, allowing them to focus on content,

1We are grateful to S.O. Semeríkov for the calculator example.
2Since disputes about terminology are usually not productive, and the expression “artificial intelligence”
itself has already firmly taken root in wide circulation, we shall also use it henceforth—with the
appropriate caveat.

2

https://doi.org/10.55056/cte.1140


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

argumentation, and the creative component of work.
In this context, the question arises as to how justified is the assertion that using

artificial intelligence tools to perform such tasks can be considered a violation of the
principles of academic integrity. Let us consider a hypothetical situation from legal
practice. Suppose we face the task of composing a statement of claim for court. For
this, we engage ChatGPT, providing it with all necessary data regarding our specific
case, and which, based on this information, helps organise the structure of the
claim, select appropriate formulations, eliminate technical flaws, and prepare the final
document in compliance with all legal requirements. Would this mean that we have
acted “dishonestly”? The absurdity of such logic becomes obvious if we imagine that
the court, having learnt about the use of a language model, refuses to consider the
case, as if the document has lost its legal force merely because its initial version was
not formed by hand. In such a case, any use of text editors or templates could also be
qualified as a violation. Obviously, the technology itself cannot be grounds for doubt
about the legitimacy or integrity of a document if its content meets legal requirements.
At the same time, responsibility for the submitted document—its legal correctness,
completeness, and reliability—in any case lies entirely with us.

Let us consider another, more “physical” example—using an excavator to dig a
trench. Let us imagine that we seriously demand to consider the use of construction
equipment “dishonest” and insist that true integrity is achieved only when a person
digs a trench with a spade. Such an approach, again, appears absurd, as the use of an
excavator is not a whim or convenience for comfort’s sake, but a necessary condition
of technical progress. It is precisely thanks to modern technology that we are able to
perform work much faster, achieve higher precision, and ensure quality that cannot
be guaranteed with manual execution. In other words, the tool does not diminish the
significance of the result but makes it possible at a qualitatively new (higher) level.
Construction equipment is precisely a tool that a person uses to achieve a goal more
effectively, in this case—construction. Similarly, a computer programme or language
model is merely a technical means that helps a person cope with work, significantly
increase its productivity and effectiveness, but in no way removes responsibility from
the person for the final result.

At the same time, artificial intelligence, despite its stunning capabilities in the
sphere of text processing, is not capable of performing genuinely creative scientific
tasks. If we turn to it with an “abstract” request to write a full-fledged scientific article
for publication in a leading professional journal (and even provide a possible topic for
such an article), it will undoubtedly create a structured and formally correct text. It
will have all the necessary external attributes of a scientific publication: introduction,
literature review, methodological part, results, and conclusions. However, upon
closer reading, it will become obvious that such material lacks scientific novelty
and original ideas. Artificial intelligence operates with already existing knowledge,
generalising, combining, and rephrasing it. It does not create new concepts, does not
form hypotheses, and does not conduct empirical research.

For example, if we set the task of preparing an article about the impact of digital-
isation on the education system, the language model will quite skilfully reproduce
general considerations that have already been voiced more than once in scientific
and journalistic literature: the convenience of online tools, possibilities of distance
learning, the need for new skills for teachers. However, if we try to obtain from it a
fundamentally new approach or original theoretical concept, the result will be, essen-
tially, a repetition of what is already known. The computer programme is not capable
of posing a new research question, seeing unexpected connections, or formulating an
innovative hypothesis. This is also clearly visible in the example of natural sciences. A
language model can create a quite competent text about the properties of a certain

3

https://doi.org/10.55056/cte.1140


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

chemical compound or even describe a possible experiment. But it conducts no real
experiment, obtains no new data, and does not analyse them. Its “research” is a
simulation based on already published knowledge. That is why scientific creativity
and genuine breakthrough in knowledge remain the prerogative of the researcher—a
person capable of critical thinking, intuitive insights, and the search for fundamen-
tally new solutions. Therefore, artificial intelligence should be considered not as a
replacement for the scientist, but as an auxiliary tool that can support research work
but can never substitute its essence. For genuine scientific activity, the emergence
of large language models not only poses no threat but, on the contrary, opens new
possibilities and actively promotes its development.

At the same time, to obtain the desired and truly useful result, it is very important to
apply language models according to their purpose and real capabilities. After all, one
should not be surprised when an attempt to use an excavator for laying an asphalted
road or creating a small flowerbed in the yard gives a result far from expected. The
same applies to artificial intelligence: if it is entrusted with tasks for which it is not
designed, the system will produce answers that will inevitably appear inappropriate or
erroneous. But the problem lies not in the tool, but in the incorrect determination of
the limits of its application.

A similar mismatch between task and tool is well illustrated by a case recently
described on social media. An editor asked a language model to find Russianisms
in her text, deliberately providing for analysis an excerpt in which there were none.
In response, the system “detected” Russianisms where no deviations from the norm
were observed: the list included perfectly literary Ukrainian words like “settlement”,
“threshold”, “signboard”, “foreigner”, as well as common expressions like “came to
mind” or “casting glances”. As a result, a long list of “errors” appeared, which were
not actually errors. The poster used this result to accuse ChatGPT of “ignorance”
and cast doubt on the possibility of its application for editing texts, whilst, of course,
emphasising her own indispensability in this role. However, this example demon-
strated not so much the shortcomings of the system itself as the incorrectness of the
formulated task. After all, this case perfectly shows: artificial intelligence performs
exactly the task that is set for it. If the system receives a request to “find Russianisms”,
it “finds” them, even if it has to invent them out of thin air. The problem here lies
not in the “evil intent” or “incompetence” of the model, but in the incorrectness of
the task set. The language model does not have its own linguistic intuition or critical
thinking—it merely combines data within given conditions. If the query is formulated
too generally or with a biased assumption, then the result will prove absurd. Instead,
the correct formulation of the task should sound different, for example: “Analyse
whether there are Russianisms in this text, and if so—carefully substantiate each
case with reference to the norms of modern Ukrainian literary language”. Such a form
of query reduces the risk of random invented “findings” and at the same time gives
the user the ability to critically assess the quality of explanations. The cited example
convincingly demonstrates: the effectiveness of the language model’s work directly
depends on the accuracy and correctness of task formulation. And responsibility for
this result—as always—lies with the person.

In the sphere of education and science, discussions around artificial intelligence
often take the form of concern about possible violation of academic integrity norms,
which is also nothing new. Almost every innovative technology that enters the ed-
ucational space is perceived as a potential threat capable of stimulating dishonest
behaviour or reducing the level of individual effort. However, the relevance of such
fears regarding language models is rather dubious. They rather indicate a misunder-
standing of the essence of the tool than a real threat to academic principles—especially
when public discussions feature initiatives to prohibit students from using ChatGPT

4

https://doi.org/10.55056/cte.1140


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

or to introduce into educational and scientific practice systems for determining “AI
texts” using so-called “detectors”. However, the practical and conceptual value of such
initiatives appears rather problematic. No algorithm can reliably distinguish machine-
generated text from human text, and prohibiting the use of tools that are already
integrated into everyday practices only creates artificial isolation of the educational
process from the modern world and poses unacceptable moral dilemmas to those we
teach.

Artificial intelligence is primarily a productivity-enhancing technology that can
remove part of the routine burden from the researcher or student. Its use does not
eliminate the need for critical thinking, hypothesis formulation, or result evaluation.
On the contrary, these competencies come to the fore, as working with a language
model requires the ability to clearly formulate tasks, analyse the received response,
and be responsible for the final result. Therefore, the very posing of the question
about academic integrity in connection with the use of artificial intelligence systems is
irrelevant. The use of a tool in itself is not a violation: everything is determined by how
exactly it is applied. The violation of integrity lies not in the presence of technology,
but in the dishonest attitude towards educational and scientific tasks—a phenomenon
that existed long before the appearance of language models.

The correct strategy lies not in rejecting the new tool, but in organically integrating
it into educational and scientific practice. The use of large language models should
become an integral competency of the modern person. This is not merely about
the technical ability to open a programme and obtain a ready answer, but primarily
about the ability to set a correct task, critically evaluate the result, and combine
algorithmic assistance with one’s own intellectual and creative work. After all, any
tool is useful to the extent that the user knows how to apply it. If a student formulates
a task superficially, they will receive a superficial answer. If the task is clear and
thought through, the result can substantially help save time and focus on the main
thing: analysis, argumentation, formation of new ideas. In this sense, working with a
language model requires greater responsibility from the user than traditional search
systems or reference books. Of course, one can always simply “press the button” and
mechanically (thoughtlessly) use the generated text. However, the result in such a
case will prove superficial and meaningless, and this insufficiency will become obvious
to anyone who seeks to familiarise themselves with the text at the level of its real
content, rather than merely glancing through it. That is precisely why for an author
working with artificial intelligence technologies, the (hermeneutic) ability to interpret
and evaluate the text being produced is especially important—to transform it from a
formal framework into a meaningful and valuable result.

The history of science and education knows such examples well. Once, students
were taught to work with library catalogues, as without this it was impossible to
navigate the flow of literature. Subsequently, the key skill became the ability to use
electronic databases and computer programmes for information processing. Today,
another competency is added to this list: the ability to work with artificial intelligence
tools. And to neglect it means to remain aside from how modern science and education
function.

Let us briefly dwell on the requirement to obligatorily indicate that a certain text
was created using artificial intelligence technologies, which recently can be found
even in the guidelines of official academic (educational and scientific) institutions,
editorial boards, etc. In our view, such a requirement appears at least inadvisable.
We do not accompany scientific articles with a disclaimer about which computer
they were written on or which operating system was used. Such clarifications have
long lost meaning, as they have become an everyday technical norm. The same will
soon happen with artificial intelligence: the use of language models in the process of

5

https://doi.org/10.55056/cte.1140


CTE Workshop Proceedings, 2025 https://doi.org/10.55056/cte.1140

text preparation will become so standard that special indication of this will appear
anachronistic. To emphasise the fact of using a tool that is gradually becoming
universal means not so much protecting academic principles as demonstrating a
misunderstanding of elementary techniques of scientific and educational activity.

Therefore, the task of universities and scientific communities lies not in limiting
access to new technologies, but on the contrary—in teaching future specialists to
use them competently and responsibly. Artificial intelligence does not replace the
human researcher but gives them the ability to work faster, more precisely, and more
ambitiously. And this is where its true value lies. In conclusion, let us emphasise
once more: large language models do not pose a threat to education and science, but
on the contrary—should become and are already becoming an effective tool for their
further development. The teacher’s task at the present stage consists primarily in
mastering these technologies as quickly as possible and being able to teach students
their responsible and effective use. Like almost every new and truly revolutionary tool,
artificial intelligence naturally causes understandable, sometimes semi-conscious
resistance at the beginning, but it will eventually prove that it not only does not
destroy the fundamental foundations of education and science, but, on the contrary,
opens new and exciting perspectives.3

References
[1] Shramko, Y., 2025. Shtuchnyi intelekt v osviti ta nautsi: uiavni zahrozy chy

novi mozhlyvosti? Available from: https://kdpu.edu.ua/blogs/2025/09/07/
shtuchnyj-intelekt-v-osviti-ta-nautsi-uyavni-zagrozy-chy-novi-mozhlyvosti/.

3And yes, for complete “transparency”: this text was created using the fifth model of ChatGPT, typed
in the LibreOffice text editor on a MacBook Air 2 computer. One could also add the processor serial
number or keyboard colour—but this would hardly make more sense than obligatory marks about the
use of artificial intelligence.

6

https://doi.org/10.55056/cte.1140
https://kdpu.edu.ua/blogs/2025/09/07/shtuchnyj-intelekt-v-osviti-ta-nautsi-uyavni-zagrozy-chy-novi-mozhlyvosti/
https://kdpu.edu.ua/blogs/2025/09/07/shtuchnyj-intelekt-v-osviti-ta-nautsi-uyavni-zagrozy-chy-novi-mozhlyvosti/