1. Introduction

The advent of digital transformation in education resulted in a new era characterized by enhanced accessibility and new autonomy (^{Latifah, Budiyanto & Saputro, 2022}; ^{Nieuwoudt, 2020}). Students from different backgrounds now have the opportunity to attain their educational objectives without being restricted by geographical or time restrictions (^{Gerashchenko & Kovalev, 2021}). However, the change is not without its challenges (^{Adzima, 2021}; ^{Montenegro-Rueda, Luque-de La Rosa, Sánchez-Serrano & Fernández-Cerero, 2021}). The issue of integrity has become increasingly prominent with the widespread adoption of resources for educational instruction and assessment in universities (^{Comas-Forgas, Lancaster, Calvo-Sastre & Sureda-Negreet, 2021}; ^{Klijn, Mdaghri Alaoui & Vorsatz, 2022}).

The intersection of technology and education has long been a subject of significant scholarly inquiry (^{Hashim, 2018}; ^{Rahm, 2023}). Learning Management Systems (LMS), such as Moodle, have significantly transformed the learning process by providing individuals with a collaborative platform that facilitates new approaches to learning (^{David et al., 2022}; ^{Maslov, Nikou & Hansen, 2021}). However, with the increasing use of assessments, there has been a growing concern over the reliability of these platforms (^{Ababneh, Ahmed & Dedousis, 2022}; ^{Coghlan, Miller & Paterson, 2020}; ^{Duhaim, Al-mamory & Mahdi, 2022}; ^{Nguyen et al., 2022}).

The integration of sophisticated technology, especially artificial intelligence (AI), into educational settings has further complicated this discourse (^{Barrientos, Cuadros, Alba & Cruz, 2021}; ^{Benvenuti et al., 2023}). In the field of examinations and assignments, the ChatGPT-4, a sophisticated language model developed by OpenAI, distinguishes itself using its capacity to generate a narrative that exhibits human-like composition. This phenomenon presents both potential opportunities and ethical dilemmas (^{Susnjak, 2022}).

Large Language Models (LLMs) have the potential to serve as a supplemental resource for a variety of assessments and academic tasks (^{Bernabei, Colabianchi, Falegnami & Costantino, 2023}). Additionally, they can offer deeper perspectives on the learning process and customise evaluations according to students' particular knowledge, abilities, and experiences (^{Tsai, Ong & Chenet, 2023}). In general, LLMs have the capacity to offer efficient ways of evaluating learning, creating examinations, replying to questions, and assessing performance (^{Swiecki et al., 2022}).

The increased capabilities of ChatGPT may introduce new components to this discourse. The recognized impact of LLMs on enhancing education is well known (^{Essel et al., 2022}; ^{Naidu & Sevnarayan, 2023}; ^{Pérez, Daradoumis & Puig, 2020}). However, concerns are emerging regarding its potential for misuse, especially in the context of assessments (^{AlAfnan, Dishari, Jovic & Lomidze, 2023}; ^{Crawford, Cowling & Allen, 2023}; ^{Oravec, 2023}; ^{Yau & Chan, 2023}). The development of advanced models like ChatGPT raises significant concerns regarding the preservation of academic and exam integrity (^{Cotton, Cotton & Shipwayet, 2023}; ^{Currie, 2023}; ^{Eke, 2023}; ^{Elbanna & Armstrong, 2023}) due to their ability to produce content that bypasses conventional plagiarism detection methods (^{Misra & Chandwar, 2023}).

Studies about cheating on exams and assignments (^{Chirumamilla & Sindre, 2019}; ^{Holden, Norris & Kuhlmeier, 2021}; ^{Noorbehbahani, Mohammadi & Aminazadeh, 2022}; ^{Von Grunigen et al., 2018}) show that there are several ways to cheat, such as searching the Internet, joining social networks, pretending to be another student, and wireless devices like earphones and microphones. The possibility that ChatGPT-4 may produce human-like messages and exhibit critical and intellectual thinking skills raises the concern that this chatbot could be a new cheating method for exams and assignments (^{Susnjak, 2022}). In this direction, the research question was formed as follows: “Do Large Language Models based on Artificial Intelligence represent a threat to the integrity of exams and assignments in Higher Education?”

The development of LLMs, such as ChatGPT-4, gives rise to notable concerns due to their remarkable ability to produce narratives that closely mimic human writing ability. The current literature has examined the potential misuse of AI in the context of evaluations. Nevertheless, a noticeable lack of understanding exists regarding the particular results of expanded LLMs in this specific domain, their capacities, and their potential to establish critical thinking standards.

This study aims to investigate the impact of ChatGPT-4 on exams in the field of tourism and the possibility of using it as a cheating tool in these exams. Also, this study aims to provide useful information for educators. Specifically, it tries to shed light on the many capabilities and possible disadvantages of ChatGPT-4 when used in exams. By making a contribution to the scope of the ongoing discourse, this study enhances the existing body of literature on this topic. This study examines the capabilities and potential challenges associated with Language Model Models (LLMs), with an emphasis on ChatGPT-4, within the academic domain. By engaging in this process, this study addresses the current knowledge gap and provides insights into these models' use and potential misapplication in the context of educational assessments. The findings of this study provide practical suggestions for educators and institutions. The recommendations aim to limit the misusing of LLMs. This study also advises on the ethical considerations of using LLMs within the educational environment.

2. Literature Review

Educational environments and challenges have been changing rapidly. These changes have fostered a shift toward active learning and creative thinking. Technology creates new learning settings and methodologies. Students’ capacity to succeed in different environments reflects their rising educational goals. Changing teachers’ perspectives and reorganising student-teacher interactions promote new educational methods (^{Petrolo et al., 2023}). These methods are digitalization and virtualization. They are used across various educational activities to enhance overall learning outcomes and expand the scope of educational resources (^{Hwang, 2018}).

The emergence of communication technologies has fostered the rise of distance learning (^{Yorkovsky & Levenberg, 2022}). COVID-19 accelerated this shift, prompting educational institutions to rapidly transition from traditional to online instruction, necessitating adjustments in scheduling, environment, and curriculum (^{Agasisti & Soncin, 2021}). This digital shift in education, popularized further by COVID-19, has made online education a prevalent global method, primarily facilitated through the Internet and supported by devices like smartphones and tablets (Asaqli, 2020; ^{Xie, Siau & Nah, 2020}). Online education offers flexibility, allowing students to access lessons and resources anytime, anywhere, saving time and costs associated with campus commutes (^{Mellieon, Jr & Robinson, 2021}). However, challenges exist. The absence of direct student-teacher interaction can complicate understanding, and the self-paced nature might diminish motivation. Connectivity issues and subpar instructional materials can impede learning, especially in rural areas (^{Ilynykh, 2021}).

Students may use some cheating methods during online exams and other assignments (^{Noorbehbahani, Mohammadi & Aminazadeh, 2022}). Students may search the Internet, join social networks, pretend to be another student, hide notes, and use mobile phones, scientific calculators, MP3 players with calculators, and wireless headphones and microphones. Also, cheating could include downloading papers and claiming them as one’s own, accessing test materials without authorization, and asking other students for answers. Teachers and students believe online examinations are easier to cheat. The biggest risks are students working together or using forbidden tools during the test (^{Holden et al., 2021}).

^{Lanier (2006}) reported that 41.1% of students admitted to cheating in online criminal justice and legal studies courses, compared to 21.3% in face-to-face courses. ^{King & Case (2014}) found that 15% of undergraduate business students admitted to cheating on exams, averaging 2.9 times each semester. ^{Watson & Sottile (2010}) noted that 32.7% of students across academic areas admitted to cheating online. ^{Dendir & Maxwell (2020}) highlighted a higher likelihood of cheating in online classes due to the absence of direct supervision and ease of accessing unapproved resources. However, not all online students cheat, and online proctoring can mitigate dishonesty risks. A ^{Wiley (2020)} survey of 800 global university educators found that 93% believed online exams offer more opportunities for cheating.

A new risk has emerged with using AI in exams and assignments: Using chatbots. Chatbots are software applications that communicate with people through text and speech on a certain topic or area. Chatbots offer many educational and training purposes. Recent advances in this subject indicate that communication with technology, whether through natural language or speech, is more feasible due to technical improvements and people’s growing familiarity with engaging digital entities (^{Smutny & Schreiberova, 2020}). Chatbots may be used for exams and assessments (Kooli, 2023); students can ask chatbots any exam question and answer that question (^{Dias, Kamdi, Gharat & Chudhari, 2019}). One of the most developed chatbots is ChatGPT. ^{Susnjak (2022}) mentioned that ChatGPT demonstrates the ability to manifest critical thinking abilities and produce writing that closely resembles reality with less input. Consequently, this poses a possible risk to the authenticity of examinations, especially in higher education environments where these assessments are more common.

The advent of generative AI models, such as ChatGPT is an important advance in the constantly changing field of higher education. Although these tools offer new and innovative methods for teaching and learning, they create concerns about academic integrity, especially in the context of assignments and exams (^{Perera & Lankathilaka, 2023}). The availability of ChatGPT enables those with limited knowledge of technology to make use of its capabilities, potentially bypassing the challenges associated with authentic learning. These models have the capability to rapidly generate a large amount of information, which presents challenges for monitoring during assessments. Moreover, their capacity to produce apparently original content presents a challenge for traditional plagiarism detection methods. The excessive dependence on AI has the potential to hinder the development of critical thinking skills while also presenting the hidden danger of AI models promoting or generating deceptive academic material. Educators often encounter challenges in distinguishing real student effort from content produced by artificial intelligence, which presents a complex and intricate situation (Gimpel et al., 2023).

LLMs can be used in distance learning or online exams, but they can also pose a threat to the integrity of assignments. The integration of AI into the field of higher education has caused a new era in ways of teaching (^{Carvalho et al., 2022}). One of the major advances in this field is ChatGPT. This adaptable language model has the potential to significantly improve the academic experience for students by offering a variety of uses. ChatGPT can serve as a tool for students to improve the effectiveness of their literature search techniques when completing assignments. Individuals can get selected collections of significant and up-to-date academic papers by entering specific questions, improving their understanding of their selected subject matter (^{Wagner, Lukyanenko & Paré, 2022}). However, students might use the advantages provided by ChatGPT in a negative manner. After completing the literature review, students can choose to use ChatGPT as a means to produce a written copy of their academic assignments. The accessibility of ChatGPT could promote excessive reliance among students on the model for solutions, thereby decreasing their preference towards self-sufficient critical thinking and problem-solving (^{Jungherr, 2023}; ^{Putra, Rangka, Aminah & Aditama, 2023}).

In the case that a significant number of students choose ChatGPT as a means of requesting assistance, it is possible that they might get standardised or similar responses in answer to their assignment questions. The procedure of homogeneity is an obstacle for educators in terms of detecting individual understanding and can potentially decrease the variety of diverse viewpoints, which is a fundamental aspect of academic discussion. According to ^{Hill (2023}), to prevent this, educators can provide an environment that encourages students to critically evaluate and correct potential biases and similarities present in the outputs produced by chatbots.

There is concern over the potential effects of ChatGPT on the academic integrity of assignments. The discussion and worry regarding their performance on many exams and assignments has generated significant attention and concern (^{Hill, 2023}). Some studies have been conducted on how ChatGPT performs in examinations in higher education institutions and other fields. According to a recent study in the medical domain (^{Gilson et al., 2022}), ChatGPT demonstrated performance equal to that of a third-year medical student who successfully passed a medical question-answer examination. In another medical study conducted by ^{Kung et al. (2023}), ChatGPT showed a performance of over 60%, meeting the established standard. ^{Lin et al. (2023}) found that ChatGPT exhibited superior performance compared to both ophthalmology residents and practising ophthalmologists.

A study in the domain of law was undertaken by (^{Choi, Hickman, Monahan & Schwarczet, 2023}). ChatGPT was evaluated in four distinct final examinations at the University of Minnesota. ChatGPT answered over 95 multiple-choice questions and 12 essay questions. ChatGPT’s scores were sufficient to meet the passing requirements.

^{Geerling, Mateer, Wooten & Damodaran, 2023} evaluated ChatGPT’s performance on a well-recognised economic knowledge assessment used in the United States. This assessment primarily focuses on assessing individuals’ comprehension at the principle level. ChatGPT achieved an accuracy rate of 63.3% in responding to microeconomics questions, while it demonstrated an accuracy rate of 86.7% in addressing macroeconomics queries.

In a study about tourism courses, the usage of ChatGPT has emerged as a significant concern concerning the trustworthiness and dependability of examinations (^{Göktaş, 2023}). The system exhibited weaknesses in accuracy and verification stages when responding to assignment questions since it offered inaccurate answers and referenced non-existent sources. Nevertheless, the system exhibited acceptable performance in multiple-choice questions, correctly answering 16 out of 20 questions. This outcome serves as a confirmation of its competence in assessing knowledge regarding multiple-choice tourism-related topics.

^{Newton & Xiromeriti (2023}) examined 41 research papers about the performance of ChatGPT. These studies together used a sample of 46204 multiple-choice questions (MCQs) derived from 97 distinct question sets. The researchers discovered that versions of ChatGPT using GPT-4 had a high success rate in multiple-choice question (MCQ) examinations. The collection of 41 studies encompasses various academic disciplines, including medical, dentistry, computer science, life support, business, law, economics, mathematics, parasitology, physics, chemistry, social work, engineering, anatomy, and accounting. Based on the available empirical evidence, it can be said that ChatGPT exhibits considerable potential as an assessment tool in several academic disciplines, assignments and exams. The significant challenge in academic evaluations is characterised by its consistent performance, adaptability with various question forms, and domain-specific proficiencies.

According to ^{Eke (2023}), generative AI systems like ChatGPT pose a real threat to academic and exam integrity because they can make students more likely to give the system their writing projects. This could lead to copying and other forms of cheating in education. ChatGPT can write logical and interesting text, but it does not fully understand the complexities of human language and conversation, which means that the system could sometimes come up with the wrong information, give damaging directions or biased information, or have limited knowledge because of how it was trained.

The utilization of ChatGPT may have negative impacts on academic and exam integrity due to its potential to enable deceptive practices and cases of plagiarism. ChatGPT possesses the ability to produce replies to questions that display a resemblance to human writing, hence presenting a difficulty in distinguishing between text created by the machine and that generated by a human. The use of ChatGPT by students to generate homework or test responses might lead to instances of academic dishonesty (^{Cotton, Cotton & Shipway, 2023}).

This study considers the most advanced ChatGPT model, ChatGPT-4. GPT-4, which was made available to the public in March 2023, is capable of accepting visual queries, can manage 32,000 tokens at the same time, and has 100 trillion parameters (^{Barrot, 2023}), whereas chatGPT-3.5 has 175 billion (^{Zaitsu & Jin, 2023}). This study seeks to assess the cognitive abilities of ChatGPT-4, a recently constructed AI agent, as well as its capacity to generate human-like language in tourism exams. There has been a growing apprehension over the potential misuse of ChatGPT-4, as it has become more possible to use its capabilities for the purpose of cheating on examinations, trespassing upon the principles of academic honesty.

3. Methodology

This study presents an explanation of the technique employed to assess the critical and higher-order thinking abilities of ChatGPT. First, ChatGPT-4 was asked to create critical thinking questions for undergraduate students in “tourism marketing, tourism management, tourism economics, tourist guidance, and gastronomy”. The questions should include a scenario. In the second stage, ChatGPT-4 was asked to answer these questions. Finally, ChatGPT-4 was asked to evaluate the answers critically. The examples of the prompts entered for the three phases are shown in Table 1.

ChatGPT was evaluated regarding its potential for formulating and answering questions and thinking critically about these answers in tourism courses. The answers were examined according to the critical thinking standards of ^{Paul (2005}) (Table 2).

Critical thinking is an idea that ties together what teachers and students need to know. It lets us help organizations change for the better. When critical thinking is known in the right way, it helps people internalize a deep knowledge of it. It is a skill that people need to be able to examine, evaluate, and understand knowledge. It is a process of actively and skillfully conceptualizing, analyzing, synthesizing, and evaluating information to reach a conclusion or solve a problem (^{Paul, 2005}). Figure 1 presents the possible critical thinking standards of ChatGPT-4 .

Figure 1 Possible critical thinking standards of ChatGPT-4 concerning tourism exams/assignments 

According to ^{Paul (2005}), the standards of critical thinking and intellectuality are below:

Clarity: It means understanding without ambiguity or confusion. It means a statement is unambiguous. Because ambiguity makes it hard to judge a statement’s accuracy or relevance, clarity is a “gateway” criterion. Elaboration, illustration, and exemplification questions can improve comprehension (^{Paul & Elder, 2013}).

Accuracy: The lack of faults, blunders, or distortions and the state of being truthful and accurate are intellectual standards of accuracy. It signifies truthfulness. Thinking is always accurate, so it is safe to assume a statement’s accuracy hasn’t been fully assessed until its authenticity is validated. One might ask questions about facts and sources to strengthen their reasoning.

Precision: The ability to be precise is an intelligence criterion. It means a comment is clear and precise. Unless one specifies a statement, one may not fully understand it. One can ask particular questions to strengthen reasoning.

Relevance: It is an intellectual criterion that requires a connection to the topic. It must be significant and logically related to the issue. It means a comment is relevant to the situation. Thoughts can always stray from the work, question, problem, or issue at hand; thus, it is advisable to assume that individuals have not properly examined their thinking until they have considered all relevant topics, thoughts, and facts. Ask questions to make sure your thoughts are relevant.

Depth: It is considering all aspects of a situation, concept, or subject. It means considering all factors. A statement or idea examines a problem’s beliefs, implications, and impacts. Thinking can uncover what is underlying or work on the surface. Asking questions about a problem’s assumptions, meanings, and consequences helps deepen thinking.

Breadth: Academic breadth requires openness and multi-perspective thinking. It means a statement or thought process recognizes a problem’s complexity and considers other viewpoints. Open-mindedness demands multi-perspective reasoning. Assessing how much thought it takes to understand a stream of thought is essential.

Logic: Logical things do not conflict with each other. It is founded on rationality. A remark or concept makes sense and does not conflict with itself. Finding out how rational thought might be beneficial. Asking questions to examine the consistency, coherence, and veracity of reasoning and proof will help think rationally.

Significance: It is the intellectual norm of being significant, impactful, or meaningful. It signifies that a remark or notion is significant, meaningful, and has significant implications. It is good to assess a thought’s relevance to the situation. Questions about the most essential facts, ideas, and thoughts about a topic and what they mean or lead to are significant.

Fairness: Intellectual fairness excludes bias, dishonesty, partiality, self-interest, lying, and injustice. It means a statement or thought is fair, unbiased, and evaluates all important points of view without considering the speaker’s feelings or interests. Asking questions encouraging people to explore alternative viewpoints and treat everyone equally, regardless of their feelings or goals, can help them think more fairly.

4. Results

This section presents the evaluation of the question generation and critical thinking skills of ChatGPT. The answers to each question are examined separately.

5. Discussion

There has been a focus (^{Susnjak, 2022}) on the value of critical thinking in both academic and practical contexts. The ability to objectively examine information holds significant relevance, especially in the contemporary world, where the Internet facilitates the rapid dissemination of incorrect information. Chatbots have the capacity to bring about a crucial change in various domains, such as the scientific realm, by expediting the distribution of knowledge. ChatGPT can respond to questions, provide informative cases, and join discussions with individuals seeking knowledge.

ChatGPT can be utilized for evaluation objectives, such as evaluating the difficulty level in generating assessments. According to ^{Ventayen (2023}), the utilization of AI in academic settings raises ethical considerations concerning academic and exam dishonesty issues. ChatGPT can write text that is nearly indistinguishable from content written by humans. This can make it challenging for tools like Turnitin to find plagiarism. Concerns have been raised that students might use ChatGPT to cheat on tests or assignments (^{Skavronskaya, Hadinejad & Cotterell, 2023}). Figure 2 shows the potential features that ChatGPT-4 can provide in tourism exams or assignments. The use of ChatGPT as a means of academic dishonesty may occur when students overly depend on it to provide solutions rather than employing it as a supplementary resource that enhances their learning experience.

Figure 2 Using ChatGPT-4 for tourism exams and assignments; Source: Created by the author. 

ChatGPT has several significant strengths. It uses an advanced natural language model to produce logical and coherent answers. Additionally, it has the capacity to enhance its own performance over time. ChatGPT exhibits some drawbacks, a deficiency in deep understanding, and challenges in assessing the quality of its generated responses. The difficulties faced by the field of education include a deficiency in threatening the principles of academic integrity and enabling the widespread occurrence of plagiarism. The findings presented in a study conducted by ^{Farrokhnia, Banihashem, Noroozi & Wals (2023}) support this study's results.

According to the results of this study, ChatGPT-4 demonstrates notable capabilities in several intellectual criteria; nonetheless, there are certain aspects that could benefit from further improvement to better accord with academic norms and expectations (Figure 3). This view is consistent with the worries expressed by scholars in the wider academic field about the potential risks associated with too much reliance on AI for important tasks in the absence of human supervision (^{Cotton et al., 2023}; ^{Currie, 2023}; ^{Eke, 2023}).

The ethical use of ChatGPT in examinations depends on how it is integrated into the assessment process and how students use it (^{Elbanna & Armstrong, 2023}). Educators need to ensure that students have a comprehensive understanding of the limitations associated with ChatGPT and that its use is closely supervised to reduce cases of academic dishonesty. ChatGPT-4, an AI language model, can potentially facilitate academic dishonesty by enabling cheating during examinations or assignments. Several studies (^{Ali, Barhom, Marino & Duggalet, 2023}; ^{Atlas, 2023}; ^{Currie, 2023}; ^{Geerling et al., 2023}) support this view. The utilization of ChatGPT-4 can serve both positive and negative objectives, and students who employ it without permission or in unethical ways are breaking the principles of academic honesty. OpenAI, the organization responsible for the development of ChatGPT-4, is currently engaged in advancing technology aimed at facilitating the identification of text produced by ChatGPT-4.

Figure 3 ChatGPT-4 as a tool for tourism exams and assignments; Source: Created by the author. 

Consequently, educators are advised to be proactive and ethical when utilizing these tools. In summary, although ChatGPT-4 possesses the potential to facilitate academic dishonesty during examinations and assignments, educational professionals and institutions should implement proactive methods to counteract this issue and uphold the principles of academic integrity.

6. Conclusion

The rapid growth of AI, particularly in sophisticated language models, such as ChatGPT-4, has caused intense discussion regarding its implications for several fields, including tourism education. The aim of this study within the field of tourism education was to assess the capacity of ChatGPT-4 to generate and assess queries that require critical thinking across several subdisciplines of tourism. The present study uses the critical thinking standards suggested by ^{Paul (2005}) to conduct a thorough assessment of the model's usefulness.

When evaluating the performance of ChatGPT-4 in relation to the critical thinking standards established by ^{Paul (2005}), many significant findings come to light. ChatGPT-4 effectively conveys concepts precisely, demonstrating a limited degree of complexity. Although the model’s accuracy is often reliable, it occasionally relies on hypothetical circumstances that may lack practical significance. While ChatGPT-4 provides accurate replies, there is room for improvement in terms of enriching the depth of its content. The text primarily emphasises the importance of relevance, with intermittent deviations. Although many responses demonstrate depth, certain situations necessitate a more nuanced examination.

The model demonstrates breadth by incorporating many opinions, although there is potential for greater diversity. The logical coherence exhibited by ChatGPT-4 is a notable feature, as it guarantees a systematic and organized development of ideas. The examination of the key concepts or the significance of the subject matter has been partially accomplished, while a more comprehensive investigation into wider implications may prove beneficial. Finally, with regard to fairness, the model tries to balance its replies; however, a more comprehensive depiction of all viewpoints would improve its dedication to this criterion.

ChatGPT-4 model shows competence in various aspects of critical thinking, including clarity, accuracy, precision, relevance, depth, breadth, logic, and fairness. However, there are still areas that could benefit from further improvement. The expertise shown by artificial intelligence-based LLM, such as ChatGPT-4, may be posed as a reasonable concern regarding the preservation of the authenticity of examinations and academic tasks within the realm of higher education. Students may misuse their capacity to provide clear, logical, and pertinent feedback that showcases critical thinking to improve the quality of their academic assignments artificially. Students' potential use of AI for completing assignments or providing exam responses can hinder real-world education and the development of critical thinking skills and compromise the fundamental objectives of educational assessments.

The results of this study are significant for education. According to the results, it is highly likely that the use of ChatGPT in exams and assignments for marketing, management, economics, tour guiding, and gastronomy, which are important areas of tourism education, will have significant effects. Consequently, the advent of AI technologies, such as ChatGPT-4 can offer advantages and potential risks for tourism exams and assignments. According to the author of this study, it becomes apparent that although AI models have achieved notable advancements in natural language processing, they are not devoid of errors.

Theoretical Implications

This study intends to fill the gap in the existing literature by examining the ethical implications and issues associated with the integration of AI in the field of education, specifically in relation to the comparison between conventional cheating tactics and AI-related risks.

The study highlights an important shift in our understanding of academic honesty within today’s environment. The growing sophistication of AI and other digital technologies, demonstrated by ChatGPT-4, presents complex problems that pose a challenge to traditional understandings of cheating and academic integrity. The process of evolution in academia calls for researchers to seriously examine the fundamental principles that define academic integrity in the modern era. Moreover, the study investigates the thinking skills of artificial intelligence, employing ChatGPT-4’s effectiveness within the field of tourism as a norm for evaluation. The study provides valuable perspectives on the domain-specific capabilities of artificial intelligence and offers indications of its wider potential and limitations within several academic disciplines. Furthermore, the findings of the study about the competence of AI in producing and addressing questions have the potential to bring about an important change in the design of assessments. In future examinations, there is a potential shift towards emphasizing the assessment of advanced intellectual skills, such as critical thinking, innovation, and applied competencies, as opposed to simply assessing the ability to memorize and recollect information.

7. Recommendations

ChatGPT-4 has demonstrated considerable achievement in fulfilling the criteria of intellectual thinking standards and exhibiting writing talents that closely resemble those of humans. The circumstance mentioned above presents a potential challenge to administering examinations in the context of tourism education. This section presents a set of recommendations for educators to reduce this threat.

^{Coghlan et al. (2020}) highlight that implementing online examination supervision systems has effectively reduced cases of academic dishonesty and unethical behaviour. These systems specifically address behaviours such as copying test answers from peers or online sources and the unauthorized use of cheating technologies. Certain universities have offered reasons for utilizing these systems, whereas others have decided to cease or decline their implementation. The author warns that despite the implementation of these regulations, there is a possibility of certain students engaging in academic dishonesty.

There is a diverse range of regulations that can be employed to detect and reduce cases of academic dishonesty. The regulations include various measures, such as identity verification, keyboard recognition, video proctoring, browser control, test duration limitations, question and answer randomization, the creation of cheat-resistant questions, and reducing the significance of the exam in the overall course marking. Offline or in-person proctored examinations may present a more effective obstacle against cheating than online exams (^{Noorbehbahani et al., 2022}).

The findings of a study indicate that the utilization of online proctoring has the potential to serve as an efficacious strategy in addressing instances of academic dishonesty within the context of online courses. Nevertheless, the research also underscores the necessity for continuous improvement in all aspects of distance education, including the level of instruction and knowledge acquisition, active participation, fairness, inclusivity, and moral behaviour. Hence, higher education institutions that provide online courses must consider the incorporation of online proctoring as a fundamental component of a holistic strategy aimed at maintaining academic and exam integrity within online learning settings (^{Dendir & Maxwell, 2020}). Potential solutions for this problem include advancing innovative plagiarism detection mechanisms with enhanced capabilities in distinguishing AI-generated content (^{Skavronskaya et al., 2023}).

In general, various methods can be implemented to prevent the utilization of ChatGPT-4 or similar tools for academic dishonesty in examinations or assignments, particularly in the domain of tourism education. The use of online exam supervision or proctoring tools may reduce academic dishonesty. During exams, these systems can monitor students in real time and look for suspicious behaviour, like moving away from the exam tab or using another device. Using identity verification makes sure that the student who signed up for the test is the one who is taking it. Keyboard recognition can determine if someone is using external tools or software by looking for trends in how they write. The advancement of new plagiarism detection techniques capable of discerning AI-generated text is crucial to upholding online exam and assignment integrity. Offline or in-person supervised examinations can more efficiently reduce cheating than the web methods. The use of a restrictive time limit during examinations can effectively limit the opportunity for potential dishonesty. The use of randomization in the assignment of questions ensures that every student is presented with an individual set of questions, hence increasing the difficulty of sharing answers. Designing questions that require the utilization of critical thinking, practical application, and analytical skills may deter students from seeking direct responses from AI systems. In conclusion, a combination of technology solutions, improvements to exam design, and a focus on ethical behaviour can achieve the effective prevention of risks associated with tools like ChatGPT-4 in exams and assignments.

Studies should be conducted across various academic disciplines to investigate the critical thinking capabilities and human-like writing abilities of ChatGPT-4. It is essential to determine the abilities or limitations of this chatbot in many academic fields and to conduct a comprehensive assessment covering various question formats, such as multiple-choice and open-ended questions. This evaluation will enable an in-depth evaluation of its performance in an examination setting.

Limitations and directions for further studies

The primary emphasis of this study focuses on the performance evaluation of ChatGPT-4 concerning intellectual criteria. The model’s range may not be comprehensive regarding its possible uses or potential misapplications across various academic and non-academic contexts. The findings and recommendations could apply particularly to the context of tourism education. It is uncertain whether these findings may be readily generalised to other disciplines or fields of study without conducting more studies. The discussion seems to be centred around hypothetical scenarios and evaluations rather than empirical studies that involve actual students, educators, or examination settings. The methodology, technologies, and issues associated with online education are subject to continual change. The material may not completely include all the details or predict future difficulties within online examinations.

Further studies should be conducted across various academic disciplines to investigate the critical thinking capabilities and human-like writing abilities of ChatGPT-4. It is essential to determine the abilities or limitations of this chatbot in many academic fields. Furthermore, it is essential to subject ChatGPT-4 to a comprehensive assessment covering various question formats, such as multiple-choice and open-ended questions. This evaluation will enable an in-depth evaluation of its performance in an examination setting.

Declaration of competing interest

None