Humanization or Dehumanization of Contemporary Education Using AI Tools - „Chatbots“ and „Humbots“

Embedded media: https://open.spotify.com/embed/episode/7enD6vsC1wTZr39sciXcCk?utm_source=generator

Introduction

As a result of the global transformation of society, "modern man" at the beginning of the third millennium faces many opportunities, challenges and threats. Unprecedented rise has been recorded by technological progress, and as M. Krejčí (2019, p. 21) writes, "we are moving from a culture that uses technology to a culture that has let itself be consumed by it; technology and media control our lives as never before, we increasingly feel an information avalanche." Technological inventions have fundamentally changed the way people interact with each other. As T. Reddy (2017) writes, the internet enabled communication between people through writing (without direct, so-called "face to face" communication).

Gradually, as P. Šantavy (2023, p. 31) states, "the arrival of the Third Industrial Revolution (also known as the digital one), which occurred in the mid-20th century, also marked the dawn of artificial intelligence and its systematic research." Subsequently, "chatbots," as conversational artificial intelligence technologies, allowed society to communicate with machine algorithms instead of communicating with real people (Reddy, T., 2017).

In many countries, AI has become ubiquitous in everyday life – from smartphone personal assistants to customer support chatbots, from entertainment recommendations to crime prediction, from facial recognition to medical diagnoses (Fengchun, M. et al., 2021). As P. Šantavý (2023, p. 36) states, "the modern development of AI systems and the reach of their use in various areas of life has brought a large number of questions concerning the coexistence of AI and humans, or society, thus these are psychological, sociological and ethical questions."

Transformation of the School System and Educational Policy

The school has been the subject of sharp criticism in recent decades; its functioning and educational goals have aroused the interest of experts calling for its comprehensive reform (Štech, S., 2016). In European history, there is no period in which it was not said that education must completely change (Feřtek, T., 2015).

The form of today's school has been deeply affected by socio-historical determination including economic, social and political transformation processes of society. As M. Beňo, Ľ. Šimčáková, J. Herich (2006, p. 35) state: "the school is the embodiment of society's interests, it is a mirror of society." Similarly, Š. Porubský, B. Kosová, K. Vančíková and S. Babiaková (2013, p. 12) state that "the institution of the school is in some way dependent on its environment, while also shaping it." According to R. Šíp (2020, p. 328), "the school is primarily a political institution, it is not and has never been ideologically neutral, because it silently reproduces the current state of society." Similarly, M. Beňo, Ľ. Šimčáková, J. Herich (2006) also write that the historical development of the school points to its dynamization and adaptation to many advances in society. The school integrates changes in society and translates them into its goals, teaching content, methods and resources, teacher qualifications and the management of the entire education system, and loses its acquired exclusivity – superiority over the family.

Despite long-term efforts to improve the quality of the education system, today's school must face many challenges that determine the requirements for the goals and content of education. The last phase of the transformation of education, as B. Kosová and Š. Porubský (2011) write, is open to the future. It remains a question whether a genuine curricular reform will be successfully implemented, thereby naturally stabilizing systemic changes. Contemporary education is particularly subject to democratization and individuality. According to S. Štech (2016), the effort to change the principles of school education founders on the vague ideas of experts who do not reflect theoretical or empirical evidence of the positive impact of the planned reform. These changes should be strongly argued and require a culturally sensitive approach so that the healthy development of children in institutional education is not endangered.

Ultimately, as described by U. Ordon, A. K. Gębora (2017), changes in the modern school presuppose a departure from the old educational system. The school faces the task of preparing young people for social, professional and cultural life in a different way than has been done in recent decades. As P. Staněk, P. Ivanová (2017, p. 69) state: "the foundation of education should be the ability to learn; overall education should enable an individual to function in a complex information society based on robotic systems, information technologies and a system based on logic in general."

As is evident, S. Štech (2016) similarly writes that in every new generation there are innovations in the content and forms of education; the school is not an enemy of change, even though it often appears that way in society. R. Šíp adds (2020, p. 440) that "the crisis of the school institution is a challenge for us to finally, with a century-long delay, adopt thinking and action suitable for life in a globalized and universal world."

Futurology of education and other pedagogical forecasts create concepts of "schools of the future," in which we anticipate changes in the manner and forms of teaching, such as team teaching of multiple teachers in one class, implementation of new didactic technologies, home-based learning (Průcha, J., 2017). According to P. Staněk, P. Ivanová (2017), we are witnessing the disappearance of the ability for interactivity, attention, cognition, searching, exploration and the function of brain plasticity. In schools, we encounter questions of cyber-addiction, which manifests itself in the constant need to be present on internet and social networks, and is literally a disease of the present time.

Artificial Intelligence as an Integral Part of Transhumanism in Education

Although we have been recording the rise of artificial intelligence mainly in recent decades, as M. Gregor and M. Gregor (2014, p. 2) write, its beginnings can be traced back to the 1950s, when the first tests of artificial intelligence were conducted. In that period, the Turing criterion was proposed: "according to which a human was to communicate with a computer using artificial intelligence, so that the human would not know that the communication was with a computer." The improvement of artificial intelligence in the following years focused mainly on speech, text, image, voice message recognition and others. Currently, the field of computer learning and the development of computer communication programs capable of learning based on ongoing communication is at the forefront.

Further early efforts to promote artificial intelligence included, as J. Grudin and R. Jacques (2019) write, for example Eliza (1966) or Parry (1972), which imitated human beings and conversed via teletype and keyboard. The enduring interest in conversational software is reflected in the creation of software (so-called "chatbots") simulating natural human conversation, such as Alice (1995), Jabberwacky (1997), Cleverbot (2008), Xiaoice (2014), Hugging Face (2016) and others.

Current artificial intelligence systems can be classified based on their ability to solve certain tasks in an intelligent way. These are narrowly specialized artificial intelligences (so-called "narrow AI"), which are optimized for handling a specific task in one area, while failing in others. A more complex category is general artificial intelligence (so-called "general AI"), which can handle any intellectual task, aiming to be at the human level. Currently used artificial intelligence systems fall into the category of specialized AI systems, while the effort towards full development of general artificial intelligence is still a prediction for the future (Šantavý, P., 2023).

On the one hand, we can talk about the advantages that artificial intelligence undeniably brings to various spheres of everyday life. AI systems can be valuable helpers in expanding and re-evaluating thoughts and ideas; they represent a vast information base that offers summaries in a short time frame (Haase, J., Hanel, P., 2023).

Extensive use of AI systems is also recorded in the environment of school facilities; however, records of favorable effects of their use are found mainly in foreign periodicals. According to a survey conducted by the Arab Education Authority, more than 50% of teachers stated that they use tools and applications based on artificial intelligence during education. According to the findings of the mentioned study, in the near future traditional educational models will be supplemented by robots and AI components that will adapt to the individual needs of students. Furthermore, an increasing percentage of students and teachers will use customizable robots and take advantage of the benefits offered by AI systems (Al-Turki, S., 2019).

There may be advantages for teaching staff; according to N. Al-Johani (2020), teachers can reduce the time demands of bureaucratic work and more easily perform routine duties including administrative work, document coordination, analysis of trends in school education and responses to various questions and assignments. According to their findings, teachers spend 31% of their time on lesson planning, test evaluation and administrative duties. With the use of AI, teachers can spend more time with students and thus reduce the workload associated with manual duties such as administration, test scoring or assignment evaluation. As J. Bailey (2023) writes, although artificial intelligence can speed up the creation of lesson plans, speed should not compromise quality.

We also note the usefulness of AI systems in mutual communication between teachers and students; according to the findings of D. Leslie et al. (2021), "chatbots" can facilitate this communication with the aim of simplifying communication and ensuring free access 24 hours a day, 7 days a week.

At the same time, according to Z. Raykova et al. (2021), programs based on artificial intelligence can be a source of feedback for both the student and the teacher. AI systems have proven successful in online education, as they can monitor student performance and immediately alert the teacher to existing performance issues. Such AI systems create conditions for effectively improving the educational process through the timely introduction of relevant changes.

Another advantage is personalized AI learning, which can help students learn at their own pace and in the context of their individual needs. Artificial intelligence can also help improve student engagement and involvement in teaching by providing interactive and engaging learning experiences. Artificial intelligence can analyze large amounts of data and provide insights into student performance, enabling teachers to better understand their students and tailor instruction to them (Harry, A., 2023).

On the other hand, gradual technological development shows that the initial focus of AI has changed dramatically. Most robots are no longer conversational, but rather automatic systems that search the internet and monitor records accessible on social media platforms. The rise of social media has led to the spread of "social bots," such as "Twitterbot," which can gain access to a large audience by "following" people while impersonating a human person, "tweeting," forwarding and responding to simple messages (Grudin, J., Jacques, R., 2019).

Another disadvantage of artificial intelligence is precisely the fact, as P. Šantavý (2023) writes, that its system does something, but how and why it does so is not clear. This is the so-called "black box," which produces results but lacks solutions. At the same time, according to Y. Zhou, B. He and L. Sun (2024), with the development of large language models, it is increasingly difficult to distinguish whether a document was created with AI, as cases of misuse include the spread of misleading information, protection of intellectual property and prevention of academic plagiarism. According to a recent study, programmed text detectors, despite promising performance on unseen test data, may not cope with adversarial tactics such as paraphrasing.

While major concerns may relate to artificial intelligence surpassing human activity, more immediate concerns relate to the social and ethical implications of artificial intelligence – such as the misuse of personal data and the possibility that artificial intelligence may actually worsen existing inequalities rather than reduce them (Fengchun, M. et al., 2021). The bias of artificial intelligence algorithms also appears to be a problem, because AI systems derive their knowledge from the data on which they are trained; for example, if this data contains prejudices, the AI system can reinforce and strengthen them (data on student outcomes biased against a particular ethnic, gender or socio-economic group, AI may discriminate against students from that demographic group). Ethical questions also come to the fore, involving concerns about privacy protection, as generative artificial intelligence tools could lead to the storage and analysis of conversations and personal information. As well as questions of equality, since not all students have equal access to computer devices and the internet (Bailey, J., 2023).

However, artificial intelligence has also entered the world of education; as M. Fengchun et al. (2021) writes, the private sector is increasingly developing "intelligent," "adaptive" and "personalized" learning systems. Artificial intelligence has the potential to address some of the biggest challenges in contemporary education and innovate teaching and learning practices, but this rapid technological progress inevitably brings multiple risks and challenges that have far outpaced policy discussions and regulatory frameworks.

Currently, new artificial intelligence technologies are mentioned in schools mainly in negative connotations. The costs of AI systems can be high and complicated to implement and maintain, which can be a challenge (not only) for educational institutions already facing budget constraints. Institutions must carefully weigh the costs and benefits of implementing AI systems (Harry, A., 2023).

Threats related to AI in the environment of school institutions also directly affect the teaching profession itself. On the one hand, while AI targets helping teachers (in the context of reducing workload, providing personalized learning activities, administering attendance, providing continuous assistance to students), some schools in remote rural areas of China already use the so-called "Double Teacher Model," within which the replacement of some teacher tasks by an artificial intelligence system has been proposed. On the other hand, the implementation of AI in education reveals a fundamental misunderstanding of the basic social role of the teacher in the learning process. At the same time, AI systems interfere with the teacher-student relationship, weakening mutual connections. Automated Writing Evaluation (AWE) relieves teachers of grading, but it often represents a key opportunity for teachers to learn about their students' strategies and abilities (Fengchun, M. et al., 2021). As already indicated, in a certain way we can also speak of a reduction in social connection between teachers and students, as well as among students themselves. As S. Al-Mashaqba (2020) writes, AI helps students learn in the home environment, which can at the same time lead to insufficient communication with classmates. This can result in social isolation and a lack of interaction among peers, and thus over time we can speak of lower cohesion in the social sense as well.

One of the main disadvantages of using AI is the misuse of systems for "school cheating." As J. Bailey (2023) writes, "cheating" by students through artificial intelligence is currently one of the greatest threats to society. Improper use of AI (creating homework, completing tests, and others) can jeopardize the natural learning process of students and the very integrity of schools. Excessive reliance on technology and "over-dependence" could hinder genuine learning, especially the development of critical thinking.

Currently, many AI systems focused on text creation are coming to the forefront, promising answers, finding inspiration and higher productivity, such as Gemini, ChatGPT, ScioBot, DeepAI, Deeply AI, Editee and others.

One of the most well-known in public awareness is the Chatbot Generative Pre-trained Transformer (ChatGPT), a type of artificial intelligence software that simulates conversation with users. The algorithms used by this chatbot are designed to understand natural language and provide relevant responses, either pre-written or created by artificial intelligence. In order to better understand user needs and fully satisfy them, ChatGPT is constantly being improved through machine learning, natural language processing and various approaches. Specifically, it can write a short text on a given topic; obtain information about a topic of interest; compose an email or message with specific content addressed to a specific person; correct the form of a text or change its wording, or solve problems. In other words, in conversation you can ask anything and receive a quick and sufficient written response like from a human (King, M. R., 2022).

Discussions about the possible replacement of human-generated texts with texts generated by artificial intelligence were sparked in the media and professional circles after the introduction of chatbots with generative artificial intelligence, such as Chat-GPT. The basis of this idea are examples of AI bots that compose specific texts in a certain language and deceive people into thinking it is human-created content. As a result of these cases, a wave of opposition arose against chatbots, with some even calling for legislation to ban their use (CNBC, 2023).

Since chatbots use sophisticated methods such as natural language processing and machine learning to create content that resembles human writing, it can be challenging to determine whether a contribution was created by a chatbot or a human. Therefore, identifying the author is more difficult and requires critical assessment (Else, H., 2023). Several signs, such as lack of nuance, style or originality, could however indicate that the work was created by a chatbot, and could be recognized by skeptical reviewers or other applications verifying AI outputs. Additionally, it can be inaccurate and contain errors that would not be found in a human-written document. A higher number of grammatical and structural errors may more likely indicate that the document was created by a human, but the reverse should not be assumed. Finally, it is unlikely that a chatbot could create a document on a highly specialized topic, as it would need a thorough understanding of the subject and the ability to produce scientific analyses and conclusions (Gao, C. A. et al., 2023).

ChatGPT is "trained" on a large amount of text data obtained from the internet, which allows it to machine-learn a complex language model, individual phrases and relationships between them. The produced text is unreliable when it comes to facts, but more valuable when fiction and random combinations are required (in poetry, creation of dialogic texts and the like). The current advanced version GPT-4 presents more complex results than the previous version (OpenAI, 2023).

In the school environment, ScioBot (https://sciobot.cz/) is often mentioned, which promises complete preparation of a teaching unit in "3 minutes" (clearly defined goals, activities for students, worksheets, tips for inspiring resources, quiz and test tasks for students, and the like). The basic version is free, only registration is required. However, as R. Čapek (2024) writes, despite some advantages (multiple ideas for teaching and activities) that the mentioned AI system offers, it is rather a "helper - muddler" that does not teach teaching methods and contains many shortcomings, such as an insufficient and limited repertoire of methods, does not categorize assessment methods, places the teacher at the main center of events (lecture, discussion), does not differentiate difficulty, does not include peer learning in teaching, and uses overly specialized terminology.

In Slovakia and the Czech Republic, Editee/Deeply AI (https://editee.com/blog/update-je-live-aktualizovali-jsme-deeply-ai-prvni-umelou-inteligenci-v-cesku-a-na-slovensku) is promoted, which states on its platforms that it is "the first artificial intelligence in the Czech Republic and Slovakia," while its updated version offers new text AI templates (sentence simplification, prompts for marketing and advertising), language and image model updates and others.

Editee, formerly known as Deeply, has ambitious plans for the Czech and Slovak markets. Its goal is, alongside text and image creation, to soon include sound generation from text or films with avatars, as well as offer an interface for editing generated photographs. Among other things, the aim is to have everything needed for creating a marketing package (written texts, photographs, videos, social media posts, etc.) through a single application (Marková, J., 2024).

AI systems are also entering the creative world, not only through text creation, but also through the creation of images, music, videos or animations. An alarming situation is represented by the AI system, as L. Payne (2024) writes, that serves to create media content, specifically various images, videos, audio recordings (e.g., main functions include inserting a face into an existing video and others). "Deepfake" is the latest evidence that created content may not be truthful, and thus cannot be blindly trusted. The term was created by combining the English terms "deep learning" – machine learning of AI and "fake" – content is not real. As S. Vinc (2024) warns, these AI systems are easily misused. A continuation of "Deepfake" is the faster version called "Reface," also known as "Doublicat," which is a mobile application that allows inserting a face into animations or videos within seconds. The user of this application risks misuse of photographs due to them being sent to foreign servers. We need not remind that the face also serves as a biometric data point, and the user has no guarantee that it will not be misused by someone.

Text produced by artificial intelligence can be detected through various AI software that determines, in a percentage range, whether the text was written by a human or an AI system. These include, for example, Turnitin, GPTZero, ZeroGPT, OpenAI, Copyleaks, Scribbr, Originality.ai, Content at Scale, Sapling, Winston AI and others. GTPZero (https://www.zerogpt.com/) is widely used, and its main advantages are: each sentence written by AI is highlighted, multiple files can be uploaded for cross-checking, it is trained in all languages, and it automatically generates reports in .pdf format for each detection. It also offers several functions such as a paraphraser, summarizer, grammar check, translation and writing assistant.

As a result of the rapid development of artificial intelligence, a large number of texts have been produced by these systems. These machine-generated texts lack the "humanness" required in the creation of any text. The effort to "humanize" text has spurred the creation of many applications or programs that "humanize" AI-generated text, such as WriteHuman, StealthWriter, Humanize AI, AISEO, AIHumanize, Humbot, Chatbot, RewriterPro, Undetectable AI, Bypass GPT, HIX Bypass and others.

These programs are designed to convert information generated by artificial intelligence into a form naturally similar to human output. Among other things, the programs first analyze the original text and then rephrase it, preserving its original content, using sophisticated natural language processing algorithms, thereby creating new content that recipients may find more convincing. In this context, W. Ali (2024) writes that text humanization, i.e., reworking text content through artificial intelligence into human-like form, is an invaluable tool for marketing, academic, and technical purposes. At the same time, a significant advantage is that it provides real-time results, as well as grammatical analysis and preservation of emotional context. Wide language support and privacy protection features further facilitate the use of the tool in an international environment.

A text that has been humanized in an educational context can be misused for unethical purposes, especially if it is used to conceal the fact that the text was generated by AI. Pupils and students can rewrite texts and scholarly articles using "humbots," thereby feigning the originality of the text. This includes, for example, creating text content using artificial intelligence, thereby avoiding plagiarism checks and AI content detection (e.g., GPTZero), but such conduct violates academic integrity. One text can be used by multiple students, resulting in multiple sharing of the same work by a single creator. This may also be associated with the formulation of the same incorrect claims or answers in class, along with distortion of information in original texts that may have been informational in nature.

W. Ali (2024) specifies the key functions of an AI-to-human text converter; for clarity of findings we have created Table 1. These include, for example, preservation of emotional context – one of the unique features of an AI-to-human text program is the ability to preserve the affective context of the original text. It adapts to preserve the intended message regardless of whether the content is formal, informal or sentimental. An advantage is also grammatical analysis and text reformulation – the tool reformulates sentences while preserving their original context and meaning; to ensure the content contains no errors, it also performs comprehensive grammar checks, as well as extended language support – the system can be used on a global scale because it supports many languages. Security and privacy protection are also essential; the AI-to-human text converter uses state-of-the-art encryption techniques to protect data during conversion, placing great emphasis on user privacy. Users need not worry about data breaches or illegal access during content conversion. And finally, real-time results – users receive immediate feedback from the artificial intelligence system.

Table 1 Key functions of an AI-to-human text converter

(Own processing)

The humanizer WriteHuman (https://writehuman.ai/) has a large user base, stating on its platforms: "elevate AI to human perfection, protect your privacy online and maintain anonymity, WriteHuman is truly the most powerful AI detection removal tool and AI humanizer that helps make your text undetectable by AI content detectors." This software is paid, but only minimal amounts, which is undeniably a "lure" of the said system, i.e., $9 per month for the basic version and $36 per month for the full version with unlimited requests for text humanization. Subsequently, AI Humanizer.ai (https://aihumanizer.ai/), which states on its platforms that "it instantly humanizes text to natural and human, is trained on millions of datasets of human-written content, uses sophisticated algorithmic technology that analyzes and replicates known patterns of human writing to perfectly convert AI text to human." Humanize artificial intelligence text in over 50+ original languages, including the Slovak language. At the same time, it is alarming that it offers a free version for one month and then only $9.99 per month for the basic version.

Another widely used humanizer is Humbot (https://humbot.ai/), promising 100% success, specifically "that it is a powerful AI humanizer that can convert AI to human text that is 100% undetectable and plagiarism-free." It compares text humanization with texts produced by the already mentioned ChatGPT and the newer version GPT4. Like the previous humanizers, it helps remove AI detection in over 50 languages. At the same time, it emphasizes the grammatical or orthographic level of the text, which it produces without errors or limits. Monthly use is available in a basic unpaid package, which certainly motivates many users to choose this AI system.

Similarly, Semihuman (https://www.semihuman.ai/), which writes on its platforms about complete freedom in content creation. As stated on their website, users can completely bypass AI detectors, while also increasing the quality of their own content creation, giving their texts a "human touch." In the academic sphere, it promises content similar to "human," rich in keywords, ranking high in search engines, relevant academic outputs and minimizing plagiarism risks. This AI system is free in the basic package for one month and then monthly in the pro version for only €6.75, which is certainly a negative finding. At the same time, concerns are also raised by the development of this system, which is overseen by a team of research and marketing experts who work with the resources of their own language model. This model undergoes training where experts thoroughly analyze a range of content including materials generated by humans and AI. As they state on their platforms, their text content is "quality, consistent and authentic, capable of fooling even the most advanced AI detectors."

Bypass AI Detection (https://bypassai.ai/) is also popular, supporting slightly fewer languages (30+) than previous systems. It likewise promises a reliable choice for bypassing AI text detection, but among other things, they state on their website that it is broadly applicable for students themselves, literally writing "achieve academic success without stress." Almost identically to the previous humanizers, the enhanced version can be used for only $9.99 per month.

"Humanization" of Text in the Institutional Environment

The creation of text, activities, summaries, images or videos through artificial intelligence systems can be used in practically every subject, whether by the teacher or the student. However, if we are to discuss the humanization of AI-generated text, the risks particularly concern the humanities subjects, whose foundation is text itself and working with it.

This applies, for example, to language subjects, specifically creative writing, where students create texts that imitate real dialogues or stories. When translating literary works and articles or when preparing various essays, letters, motivational texts, stories or reports. In history, they can create stories or diaries from the perspective of historical figures and analyze historical sources. Through humanization, students can make complex historical documents accessible to classmates. In science and natural science subjects, they can directly experiment with AI tools and participate in creating chatbots. In addition to the aforementioned, they can rework specialized and scientific texts into more accessible forms along with creating projects, presentations and specific texts.

In lower secondary education, according to the Innovated State Educational Program in the area of language and communication, students can use AI tools for external and internal characteristics of a person; for orders, invitations, posters and advertisements; for a structured CV; for writing announcements and messages or an outline of a read passage (Štátny pedagogický ústav, 2014a). From the area of language and communication, Slovak language and literature, for grammar schools and secondary vocational schools and for years 5–8 of eight-year grammar schools, a student can use AI tools for interpreting a literary work; creating required texts of various stylistic forms; creating texts of various genres and finally creating a poetic text, short story, script or other shorter literary text (Štátny pedagogický ústav, 2014b).

For our investigation, we selected topics from the Slovak language school-leaving essay examinations for the school year 2022/2023, namely In Winter in the Countryside (Artistic Description) and from the school year 2021/2022 on the topic Where Is Our Home Really? (Reflective Essay). In the creation and summarization of text, we worked in partial parts of the investigation: the first part – text generation, the second part – text detection, the third part – text humanization. We used the following AI systems: ChatGPT (text creation) and ZeroGTP (AI detection). For text humanization, we used two AI systems, specifically Humbot and Bypass AI Detection. We chose these programs based on their accessibility (mostly free in the basic version).

In the first step, we used artificial intelligence to generate a text on the topic In Winter in the Countryside (Artistic Description), and the result is shown in Figure 1. The theme of In Winter in the Countryside is an idyllic description of winter in a rural environment, with AI attempting to capture the beauty and peaceful atmosphere of winter nature. The text is lyrical and seeks to evoke a feeling of silence, harmony and majesty of nature. It is an artistically oriented, descriptive and poetic style using various linguistic devices. Rich metaphors, personifications, visualizations, color contrasts and sensory descriptions are used. The text is divided into thematically separate parts such as introduction, body and conclusion. AI captured the atmosphere calmly, gently melancholically, also capturing not only the physical beauty of winter, but also its symbolic meaning — a time of peace and recognition of nature's beauty.

Figure 1 ChatGPT and artistic description on the topic In Winter in the Countryside

(Own processing)

Subsequently, we inserted the text into the GTPZero system, which generates a percentage match of text created "artificially" based on the use of AI systems. From our findings, in the context of the data in Figure 2, the text we examined was evaluated as follows: 18% written by a human (Human), 1% is a combination of human and artificial intelligence (Mixed) and 81% was created by artificial intelligence.

Figure 2 GTPZero program evaluation of the artistic description

(Own processing)

The GPTZero tool marked the text as 81% created by AI, thus the text is evaluated as generated or modified by artificial intelligence. It has certain stylistic elements that are very common in generated texts today, such as the recurring figurative expression "Snow, like a white carpet" or "trees, bare and seemingly dead." Artificial intelligence can be very good at creating texts rich in imagery, but they often seem less concrete and more abstract. The overly uniform tone of the text is consistent and does not exceed a certain framework of originality. It focuses on feelings and images without any hint of a particular, unique perspective or original reflection typical of a human. Formulations or images that are recurring and popular in the data database also appear, for example, "Snow like a white carpet" can be considered a cliché. Analysis using the GPTZero tool may have its limitations; interpretation of results may depend on the algorithm used and its ability to identify patterns in the text.

Subsequently, we inserted the AI-generated text into the "Humbot" program, which "humanizes" the text. The results in Figure 3 show that the language in the text paraphrased by the humanizer is still poetic, but less "refined" than in the original text. It uses less natural word combinations e.g., "shadows with elongation." This text seems somewhat ambiguous and is sometimes grammatically less correct. Some images are changed or described in less detail, which can reduce the emotional power of the text. Instead of a fluent narrative, there are several fragments that appear insufficiently coherent. It remains faithful to the theme, but the descriptions are less precise. Instead of a nuanced expression of contrasts between nature and the human world, only fragmentary images appear in the text without deeper emotional impact. Images and metaphors "Snow with a white carpet covers" – a similar image, but with imprecision in formulation including a grammatical error. The statement "Nestling up to the gentle contours" is less clear and engaging than "creating gentle contours."

The paraphrased text is less fluent, sometimes stylistically and grammatically imprecise, containing less powerful images. It is a rather more literal and less dynamic version of the first text, losing some of the subtle details and artistic impression; it is also less stylistically precise, which may be a result of simplification or direct reformulation.

Figure 3 Humbot program humanizing text

(Own processing)

In the last part, we re-inserted the reformulated text into the GTPZero program; the results in Figure 4 describe the difference: the text was constructed by a human at 37%, again 1% is a combination of human and artificial intelligence and 62% was created by artificial intelligence. In the examined example of the text on the topic In Winter in the Countryside, humanization improved the result by 19%.

Figure 4 GTPZero program evaluating the humanized artistic description text

(Own processing)

After humanizing the text on the topic In Winter in the Countryside, the GPTZero system evaluated that the probability that this text was written by a human is only 37%. This result indicates that the text exhibits only some characteristics that are more typical of generated content. Specifically, these are recurring phrases and occasional imprecision in fluency and connection of ideas. Although the text still exhibits certain artistic and descriptive elements that could indicate "humanness," it also informs us of its shortcomings.

The second example presented has the same sequence of steps, however we used a reflective essay on the topic Where Is Our Home Really?. From Figure 5, it is evident that the text is written in a very universal manner, with frequent repetitions and without specific personal experiences or strong emotional elements that would more realistically capture the deeper meaning of the theme of home. Although AI is capable of creating texts that are cultivated or logical, it still tends to avoid the subjective depth that a human could embed in the text. A technical error in the text may also indicate mixed content.

Figure 5 ChatGPT reflective essay on the topic Where Is Our Home Really?

(Own processing)

The GTPZero program evaluated the following (Figure 6): 16% written by a human (Human), 1% is a combination of human and artificial intelligence (Mixed) and 83% was created by artificial intelligence. In analyzing this text with an emphasis on it being written by AI, we can point to several critical points that could be characteristic of generated content. There is an absence of deeper personal reflection. Although the text touches on strong and universal themes such as safety, love and the place where we feel at home, it lacks a unique perspective. "Home is more than four walls" is a sentence appearing in various forms in many texts about home. It does not bring a new perspective, but rather a summary that could come from a database of existing texts. These common phrases and generalized statements may appear as outputs that AI generates based on analysis of large amounts of available data, but they do not allow for truly original thinking that a human could offer.

A problem may also be the repetition of ideas: "It is a space where we feel safe, where we can be ourselves" versus "These are memories tied to a place, but also a feeling of security and love." – in both cases, it is an expression of the same idea: home is a place of safety and love. The text may thus appear as a summary of common ideas that a human would likely express in other, more original forms. Although AI offers correct and template-like answers on the theme of home, it lacks human depth, sensitivity and the ability to express a personal perspective on this topic.

Figure 6 GTPZero program evaluation of the reflective essay

(Own processing)

We again subjected the created text to "humanization," using Bypass AI Detection; the detailed results are shown in Figure 7. The AI texts (the originally generated text as well as the humanized text) address the same concept of home, but their language, structure and some details differ. The humanized text uses fewer poetic images and is somewhat more concrete. It uses formulations such as "four-walled barrier piece" and "terrain that allowed making mistakes," which are visually and emotionally stronger, but more pragmatic and less poetic. The expression "Home is not a belief" is also unusual and appears ambiguous. Nevertheless, although the ideas are similar to those in the first text, they are expressed in a different way, using more direct and concrete expressions. The first text focuses on beauty, imagery and emotional relationships, while the humanized one is more analytical, focusing on concepts and philosophical questions.

Figure 7 Bypass AI Detection program humanizing text

(Own processing)

In the last part, we again inserted the reformulated content into the GTPZero program; the results describe a larger difference, the text was written by a human at 99%, 0% is a combination of human and artificial intelligence and 1% was created by artificial intelligence.

In the examined example of the text on the topic Where Is Our Home Really?, humanization improved the result by 83%. After humanization and detailed evaluation of the text, the GPTZero system assessed that the text has a 99% probability of having been written by a human. The result should indicate that the text exhibits a very high degree of naturalness and emotional depth characteristic of human writing. The use of imagery, sensitive thoughts and a personal approach can contribute to an authentic and original character; these elements can also reduce the likelihood that the text could be generated by artificial intelligence, which has not yet reached this level of emotional and stylistic sophistication.

In both cases, the humanizer exhibited a higher word count than the original text, with possible typos, grammatical errors and incorrect word order. The imperfection was influenced by the use of the Slovak language, which is not yet sufficiently refined.

Figure 8 GTPZero program evaluating the humanized reflective essay text

Conclusion

Artificial intelligence and its development bring many challenges, opportunities and threats into our lives. Within education, while it has the potential to "improve" or "innovate," on the other hand it can alarmingly affect the development of students' critical thinking, their study activities and ultimately also the role of the teacher or the very integrity of the institutional environment. An acceptable path is awareness, prevention and education, focused not only on students and educators, but also continuously in the context of social awareness.

Awareness and preventive modules can be supported by mutual cooperation; as D. O. Negoiță and M. M. A. Popescu (2023) write, currently it is necessary to pursue collaboration between educators and AI developers, while effective integration of AI into educational practices remains a shared challenge and opportunity.

It is also necessary to develop digital literacy among students and educators. Key is understanding how to effectively and safely use artificial intelligence tools. Teachers must be equipped with the latest and updated skills and resources to responsibly guide students in working within the digital environment and to create an environment where technology enhances learning without overshadowing the human elements of education. A common way of misusing artificial intelligence in classrooms is using AI to write essays and homework; therefore it is important that teachers are equipped with tools and ways to deal with plagiarism and AI forgery in the classroom while maintaining academic integrity. Furthermore, the adoption of artificial intelligence in teaching requires a commitment to lifelong learning. Teachers must continuously update their skills to keep pace with technological progress. This includes not only technical training, but also developing an understanding of the ethical and pedagogical factors of using AI in the classroom. Creating a community of educators who share their insights, challenges and successes can provide invaluable support. Collaboration platforms, professional networks and discussion forums in education offer teachers a space for exchanging ideas and strategies for effectively using artificial intelligence in teaching (Deshpande, A. S. et al., 2024).

The concept of education within AI offers an innovative approach to addressing special needs of students through artificial intelligence and improves understanding of how AI can identify and support students (Yang, S. J. H., 2021). The use of AI, intelligent sensors and devices for self-regulated learning highlights the potential for improving communication between students and teachers (Zawacki-Richter, O. et al., 2019); deeper integration of modern technologies with educational theories, bridging the gap between AI capabilities and the goals of modern education remains a priority (Chiu, T. K. F. et al., 2022).

In the context of education, we randomly selected a primary school in the Žilina district (https://zskamennaporuba.edupage.org/a/podpora-pedagogom), which states on its social platforms in the "support for educators" section – opportunities for teacher training in the area of information and communication technologies. Employees of the primary school, as well as the professional public, can read information about artificial intelligence in education and its systems, digital technologies in teaching or cyberbullying. These AI systems and applications are presented with a brief explanation, supplemented by hyperlinks to specific websites. Each section is complemented by "questions for reflection," such as "How to properly set up time management for teaching staff," "How should the school prepare children for the new phenomenon of the time?", "What impact will AI have on a child's psyche?" and others.

It is more than certain, however, that the role of the teacher and the education system will change under the influence of AI development. The modern teacher will have to reflect technological progress in society; it will be necessary to build new competencies in connection with the effective mastery of AI and its implementation into the education system (Fengchun, M. et al., 2021). We concur with the ideas of J. Haase and P. Hanel (2023), that the more our lives become automated, the more important the human role and creative abilities become.

Authors: PaedDr. Beáta, Pošteková, PhD. Department of Pedagogical Studies, Faculty of Humanities, University of Žilina PaedDr. Libuša, Gužíková, PhD. Department of Pedagogy, Faculty of Education, Constantine the Philosopher University

References

AI Humanizer.ai. 2024. Available at: https://aihumanizer.ai/ [cit. 2024-10-21].

AL-JOHANI, Nourah. 2020. The impact of artificial intelligence on education in the Arab world. In The International Journal of Modern Techniques in Education. Vol. 15, 2020, no. 5.

AL-MASHAQBA, Samer. 2020. The impact of artificial intelligence on the outcomes of students' learning in the Jordanian Universities. In The Journal of Education and Practice. Vol. 11, 2020, no. 1.

AL-TURKI, Sami. 2019. The impact of artificial intelligence on education in the Kingdom of Saudi Arabia. In The International Journal of Modern Technologies in Learning. Vol. 14, 2019, no. 13.

ALI, Waseem. 2024. AI to Human Text [online]. Scotland : University of the West of Scotland, 2024. Available at: https://www.researchgate.net/publication/383022660_AI_to_Human_Text [cit. 2024-11-19].

BAILEY, Jessica. 2023. AI in Education [online]. Available at: https://www.aei.org/articles/ai-in-education/ [cit. 2024-11-22].

BEŇO, Miloš – ŠIMČÁKOVÁ, Ľubica – HERICH, Ján. 2006. Rodičia o súčasnej škole. Bratislava : Ústav informácií a prognóz školstva, 2006. 314 p. ISBN 80-7098-441-4.

Bypass AI Detection. 2024. Available at: https://bypassai.ai/ [cit. 2024-11-19].

CNBC. 2023. Italy Became the First Western Country to Ban ChatGPT. Here's What Other Countries are Doing [online]. Available at: https://www.cnbc.com/2023/04/04/italy-has-banned-chatgpt-heres-what-other-countries-are-doing.html [cit. 2024-11-26].

FEŘTEK, Tomáš. 2015. Co je nového vo vzdělávání. Praha : Nová Beseda, 2015. 100 p. ISBN 978-80-906089-2-4.

KREJČÍ, Martin. 2019. DigiDetox: Ako dosiahnuť digitálny minimalizmus. Bratislava : IKAR, 2019. 217 p. ISBN 978-80-551-7685-7.

NEGOIȚĂ, Ovidiu Daniel – POPESCU, Andreea Maria Mihaela. 2023. THE USE OF ARTIFICIAL INTELLIGENCE IN EDUCATION. In International Conference of Management and Industrial Engineering [online]. Vol. 11, 2023. ISSN 2344-0937. Available at: https://doi.org/10.56177/11icmie2023.43 [cit. 2024-12-12].

PORUBSKÝ, Štefan – KOSOVÁ, Beáta – VANČÍKOVÁ, Katarína – BABIAKOVÁ, Silvia. 2013. Premeny spoločnosti a perspektívy školy. Banská Bystrica : Belianum, 2013. 124 p. ISBN 978-80-557-0590-3.

PRŮCHA, Jan. 2017. Moderní pedagogika. Praha : Portál, 2017. 488 p. ISBN 978-80-262-12287.

STANĚK, Peter – IVANOVÁ, Petra. 2017. Spoločnosť 5.0 – ekonomika budúcnosti. Bratislava : Wolters Kluwer, 2017. 176 p. ISBN 978-80-8168-978-8.

ŠANTAVÝ, Petr. 2023. Umelá inteligencia – dobrý sluha a zlý pán? Ľudia, stroje, spoločnosť... Etika, hodnoty, budúcnosť... Bratislava : Univerzita Komenského, 2023. 319 p. ISBN 978-80-88696-91-9.

ŠÍP, Radim. 2020. Proč školství a jeho aktéři selhávají: Kognitivní krajiny a nacionalismus. Brno : Masarykova univerzita, 2020. 471 p. ISBN 978-80-210-9377-5.

ŠTECH, Stanislav. 2016. Velká regrese: Proměny vzdělávání mezi hlasy politiků, expertů a rodičů. In STROUHAL, Milan – ŠTECH, Stanislav. Vzdělání a dnešek: Pedagogické, filosofické, historické a sociální perspektivy. Praha : Univerzita Karlova, 2016, p. 179–201. ISBN 978-80-246-3558-3.

YANG, Stephen J. H. 2021. Precision Education – A New Challenge for AI in Education. In Journal of Educational Technology & Society. Vol. 24, 2021, no. 1.

ZAWACKI-RICHTER, Olaf et al. 2019. Systematic review of research on artificial intelligence applications in higher education – where are the educators?. In International Journal of Educational Technology in Higher Education. Vol. 16, 2019, no. 1.