Abstract: The integration of generative artificial intelligence into officer education represents a significant innovation that challenges doctrines, practices, and pedagogical strategies. At the same time, it demands a balance between the potential of generative artificial intelligence and the cadets’ autonomy to critically analyse and correlate the knowledge transmitted. In this exploratory-descriptive study, a questionnaire was administered to first-year cadets at the Portuguese Military Academy. These students were from the Portuguese Army Officer Courses and the Portuguese National Republican Guard.
Problem statement: What are first-year cadets’ perceptions of generative artificial intelligence as a classroom tool and a complement to learning methods at the Portuguese Military Academy?
So what: The results on cadets’ perceptions support the favourable integration of artificial intelligence in the classroom and as a complement to learning methods and techniques, with particular emphasis on active learning and brainstorming. However, for this integration to be beneficial, generative artificial intelligence must not replace the knowledge-transfer process or impair cadets’ ability to generate ideas and think independently. The findings underscore the importance of providing cadets with informed training in the use of generative artificial intelligence to ensure that it serves as a complementary tool rather than a replacement.
Introduction
The Portuguese Military Academy is aware of the opportunities, risks and challenges of Generative Artificial Intelligence (GenAI) and, as a result, a permanent directive on its use has been in place since December last year, with the aim of establishing rules and procedures for the autonomous use of tools in students’ education and training. The objective is to guide the use of GenAI in accordance with ethical principles, transparency, academic integrity, information reliability, respect for copyright, critical-thinking autonomy, and institutional security. As the Portuguese Military Academy is at the beginning of this implementation, studies on the use of GenAI underscore the need for learning strategies that harness GenAI’s capabilities without supplanting the development of individual skills and abilities — essential for future officers’ decision-making processes.
In this regard, the central theme of this article is the integration of GenAI into military training, with a focus on the Portuguese military education system in general, particularly the Portuguese Military Academy, and on first-year cadets’ perceptions of the use of GenAI. The study examines the use of tools such as ChatGPT functioning as classroom support instruments and complements to pedagogical methods and techniques, including active learning and brainstorming. This topic is particularly important due to GenAI’s potential to transform both military and civilian higher education. On the one hand, it offers opportunities such as study support, while on the other, it poses challenges, including the potential erosion of critical thinking and intellectual autonomy. In the military context, where the capacity for independent decision-making and complex problem-solving remains essential for future officers, the balance that preserves human reasoning constitutes a strategic and doctrinal issue.
Across studies, both opportunities and risks associated with GenAI integration are consistently reported. Identified opportunities include personalised learning experience,[1] learning support,[2] and learning efficiency.[3] Reported risks encompass reliability,[4] academic integrity,[5] excessive dependence and diminished critical thinking,[6] educational inequality,[7] and privacy and intellectual property concerns.[8]
In addition to this viewpoint, the literature underscores the necessity of striking a balance between maximising educational benefits from GenAI and minimising the risks of skills and capacity erosion to optimise student learning. This presents a challenge necessitating the development of regulatory policies, ethical frameworks, and institutional strategies, such as adopting alternative assessment and learning models.[9]
Studies focused on military higher education particularly emphasise this balanced perspective. For instance, Kelly and Smith’s study on professional military education proposes a balanced approach that neither completely rejects GenAI nor uncritically accepts it.[10] Their study identified several key factors in the total rejection approach: some instructors and leaders view ChatGPT primarily as a source of plagiarism and a threat to academic integrity, leading to blanket prohibitions on its use at any stage of assignments; denial of potential benefits by refusing to explore tools that could provide cognitive advantages; and hostile prohibitions that ultimately undermine the learning process by limiting data literacy and discouraging students rather than guiding them.[11]
In the uncritical acceptance approach, GenAI is portrayed as akin to the latest calculator or word processor — a tool that generates assignment ideas, serves as a conversational tutor, provides contextual background, frees time for creative thinking, synthesises sources, formulates theses, and constructs complete arguments. This perspective encourages students to substitute GenAI feedback for genuine reasoning, which can hinder the development of critical and creative thinking skills.[12] The Authors remind that knowledge is not only a product but also a process, and when GenAI is used to synthesise data, develop arguments and produce summaries, it eliminates the student’s engagement in, for example, analysing data and developing their own well-founded arguments.
The intermediate approach, on the other hand, suggests that GenAI be used to complement human thought while restricting its use to instances where it does not substitute for human thought.[13]
Accordingly, the proposal outlines several guidelines, including redesigning assessments to enhance student engagement, incorporating specific clauses on GenAI use in course syllabi to prevent overreliance, and providing faculty training on GenAI’s capabilities and limitations.[14]
In their study, Watson and Romic advocate incorporating GenAI — particularly ChatGPT — into education. They emphasise that this integration must be done responsibly to harness its potential for inclusion while mitigating the risks of exclusion.[15] They argue that integrating GenAI necessitates contextual, practice-oriented research involving multiple stakeholders, supported by strategic educational policies that foster technical, critical, and ethical skills. This approach allows for informed, creative, and equitable use of ChatGPT by students and teachers alike.[16] The authors further recognise that the core challenge lies in integrating ChatGPT into educational programs in a practical and ethical manner while preserving the integrity of the entire educational system. To ensure the success of GenAI integration, it is essential to approach it thoughtfully and allocate sufficient time for meaningful human interaction with the technology.[17]
In the field of military higher education, studies consistently identify the potential of GenAI across three key categories: personalised learning,[18] simulations[19] and educational efficiency.[20] The following examples are the most relevant to this study: enhanced student motivation and engagement;[21] development of digital environment skills;[22] learning process flexibility;[23] and learning support, particularly the value of brainstorming for writing tasks.[24]
It is imperative to acknowledge the potential implications of technological dependence on military capabilities. This dependence poses a significant risk of eroding critical thinking, independent decision-making, and problem-solving skills.[25] Privacy and data security concerns[26] have been raised. There has been a noticeable decline in teacher-student relationships, resulting in fewer interactions,[27] and the potential to undermine discipline, character, and group dynamics, which are critical in military contexts and lastly, difficulties in adapting traditional curricula to GenAI dynamics.[28]
Methodology
This study employed an exploratory-descriptive research design to examine first-year cadets’ perceptions of generative artificial intelligence (GenAI) as a classroom tool and as a complement to pedagogical methods at the Portuguese Military Academy. As no formal GenAI training is currently included in the first-year curriculum, the study provided an opportunity to investigate how cadets independently engage with tools such as ChatGPT and how they perceive their educational value.
Participants
The target population consisted of 103 first-year cadets enrolled at the Portuguese Military Academy during the first semester of the 2025–2026 academic year. Participants were drawn from five study programmes: Military Sciences (Security), Military Sciences (Administration), Military Engineering, Military Mechanical Engineering, and Military Electrical Engineering. These cadets represent future officers of both the Portuguese Army and the National Republican Guard, providing a broad perspective on officer education at the beginning of military higher education.
Educational Context
All participants attended the same curricular unit, which incorporated a variety of teaching methods. These included the expository method, based primarily on lectures and presentations; the interrogative method, centred on question-and-answer interactions; and the active method, which employed techniques such as brainstorming, think-pair-share, case studies, and role-play. This common educational experience provided a consistent basis for evaluating cadets’ perceptions of GenAI across different pedagogical approaches.
Instrument Development
Data were collected through a structured questionnaire developed using Google Forms. The instrument was designed to assess cadets’ perceptions of GenAI’s role in facilitating content comprehension, idea generation, critical thinking, and support for different pedagogical methods and techniques.
A pilot test was conducted with 43 cadets (41.75 per cent of the target population). Feedback from this process was used to improve clarity, reduce redundancy, and increase the instrument’s overall usability. As a result, the questionnaire was refined from twenty to twelve questions, reducing completion time while maintaining alignment with the study’s objectives.
The final questionnaire consisted of two sections. The first included five multiple-choice questions examining patterns of GenAI use. The second comprised seven items measured on a seven-point Likert scale ranging from 1 (“Strongly disagree”) to 7 (“Strongly agree”). These items explored perceptions of GenAI as a facilitator, complement, or potential obstacle to learning and cognitive development.
Data Collection
The final questionnaire was administered in person on 15 January 2026 to all 103 first-year cadets. Participation was voluntary, anonymous, and based on informed consent. Respondents completed the questionnaire independently and submitted their responses electronically. The data were subsequently aggregated for analysis.
Data Analysis
The analysis focused on identifying patterns in cadets’ perceptions of GenAI across different educational contexts. Responses were analysed descriptively, with particular attention to perceived usefulness for content comprehension, idea generation, and critical thinking, as well as perceptions of GenAI’s relationship to expository, interrogative, and active learning methods.
Ethical Considerations
Participants were informed of the purpose of the study, the voluntary nature of participation, and the measures adopted to ensure anonymity and confidentiality. All respondents provided informed consent prior to participation. Data were analysed in aggregate form and used exclusively for academic research purposes.
Presentation of Result
This chapter presents the data obtained from the questionnaire, subdivided into six categories: most-used artificial intelligence tools; learning methods; learning techniques; GenAI as a facilitating element; GenAI as a complementary element; and GenAI as a hindering element.
Specifically, ChatGPT (67.96%) has the highest adoption rate among cadets, followed by Bing/Microsoft Copilot (12.62%), which is used five times less frequently. This fact indicates that ChatGPT’s marked predominance positions it as the baseline tool for GenAI integration into the expository, interrogative, and active methods examined in subsequent sections.
The following section presents the results obtained regarding the question of whether the use of GenAI complements learning methods, namely active, interrogative, and expository. The results indicate that, in most cadets’ perception, GenAI complements the identified learning methods. While most cadets view GenAI as complementary, agreement declines as the order changes. Specifically, the expository method has an agreement rate of 55.34%, followed by the interrogative method at 54.43%, and finally, the active method at 42.72%. This hierarchy indicates a higher level of integration of GenAI in direct content-transmission contexts than in practical approaches that require greater autonomy.
However, the findings underscore the potential risks associated with technological dependency among cadets, regardless of the method utilised. This fact supports the view that GenAI should serve as a facilitator rather than a conditioner of intellectual autonomy.
Additionally, although the active method shows the lowest percentage of “slightly agree” responses, it stands out among the three methods as the one most complemented by GenAI. This apparent contradiction may indicate technology’s role in enhancing cadets’ thinking without replacing it, or reflect recognition of its long-term strategic value. This inclination towards the active approach is further validated by the selection of interaction techniques, such as brainstorming, which will be discussed in the subsequent point.
Regarding whether the use of generative artificial intelligence complements learning techniques, the results are as follows. Concerning techniques that most encourage thinking when using GenAI, 33.98% of cadets again highlighted brainstorming. In comparison, 22.33% selected the case study, 19.42% chose the think-pair-share, 11.65% indicated the question-and-answer technique, 9.71% preferred the oral presentation, and 2.91% mentioned role-playing. In the classroom, the main techniques identified by cadets are brainstorming, case study, and think-pair-share. The selection of these three techniques directly correlates with the cadets’ preference for the active method. In particular, brainstorming emerges as the most frequently identified technique among cadets as effective for supporting content comprehension (22.33%), idea generation (46.60%), and thinking (33.98%).
An analysis reveals that, for content comprehension tasks, cadets’ choices are balanced across the active method (via brainstorming, case study, and think-pair-share), the interrogative method (via the question-answer technique), and the expository method (associated with PowerPoint). Conversely, cadets prefer techniques associated with the active method when engaging in cognitive tasks that require idea generation and critical thinking. These findings suggest that cadets view GenAI as a more effective tool for content comprehension than for idea generation or thinking.
Cadets generally perceive that using GenAI in class facilitates content understanding, idea generation, and critical thinking. However, notable variations exist across these dimensions. The intensity of this perception varies according to task complexity. Content comprehension shows the highest intensity (91.25%), indicating GenAI’s primary role as a support tool. Idea generation decreases to 81.54%. Thinking, the dimension with the least support (64.08%), exhibits the highest neutrality (15.53%) and disagreement (20.38%). These findings are reinforced by responses to a question about cadets’ perceptions of using GenAI when their knowledge of the content is still general. In this situation, cadets recognise that GenAI facilitates comprehension, serving as a valuable starting point when they are unfamiliar with the content. In generating ideas, cadets’ perception of utility decreased (75.73%), which may indicate reduced effectiveness without a solid knowledge base. Regarding thinking, this element remains consistent at 67.96%. This is the area where cadets consider GenAI least useful.
The data indicate a decline in perceived utility: GenAI has been found to be highly effective in content comprehension and idea generation, though its reliability is less optimal for more complex tasks. GenAI is also perceived as more valid in early learning phases, when knowledge remains generic.
As cadets progress from content comprehension to autonomous thinking, neutral and disagreeing responses increase, demonstrating their awareness that GenAI serves as a facilitator but not a substitute for reasoning. Cadets also maintain a degree of scepticism regarding GenAI’s role in more complex cognitive processes, such as thinking. Most cadets recognise that the use of GenAI in class complements content understanding, idea generation, and critical thinking. These findings are reinforced by responses to a question about GenAI use when knowledge of content remains general. The findings indicate that GenAI’s effectiveness is optimal for content comprehension and minimal for critical thinking, validating the perception that cadets feel more secure using GenAI to understand transmitted material than to support their own reasoning. The data also suggest that cadets hesitate to consider GenAI as a complement or useful tool for idea generation when they lack mastery of the topic.
In both situations, the percentage of cadets who disagree that GenAI serves as a complement remains consistent: 3.88% for content, 7.77% for idea generation, and 11.65% for thinking. These values indicate that a group of cadets maintains reservations about GenAI’s complementarity, regardless of knowledge level. Cadets generally recognise that the use of GenAI in the classroom does not hinder content understanding, idea generation, or thinking. However, this perception varies with task complexity and knowledge mastery. Only 11.65% of cadets believe that GenAI negatively impacts content comprehension, while the vast majority (79.69%) disagree that it poses a significant obstacle. However, 20.38% of cadets view GenAI as a limitation on idea generation, and 25.24% consider it an obstacle to developing independent thinking.
Moreover, in class, when knowledge is still general, 19.41% of cadets see GenAI as a potential hindrance to understanding of the taught content; 20.38% perceive it as a limitation on idea generation; and 27.18% consider it an obstacle to autonomous thinking when the subject lacks mastery. Neutral responses increased significantly (20.39%).
Although the overall view remains positive, cadets are aware that GenAI may pose an obstacle to intellectual autonomy, particularly when used as a substitute for reasoning in phases of generic knowledge.
Analysis and Discussion of Results
The results show that the effectiveness of GenAI is inversely proportional to the task’s intellectual depth. On the one hand, cadets recognise GenAI as a useful tool for simple tasks such as understanding content, but on the other hand, uncertainty increases as intellectual demands become more complex, such as thinking. The transition from “certainty” in the comprehension phase to “uncertainty” in the critical thinking phase indicates that cadets do not yet recognise GenAI as a useful tool for complex reasoning.
It is observed that the use of GenAI at the Military Academy, whilst it may serve as a facilitator of general knowledge, excessive use may erode critical thinking, as corroborated by studies by authors such as Batista and Bouguettaya. By facilitating access to final products, such as summaries or immediate answers, GenAI becomes a shortcut that avoids intellectual discomfort through discovery, reflection, and discussion.
Beyond the recognised utility, cadets perceive that GenAI may constitute an obstacle to intellectual autonomy and cognitive independence. Although a high percentage of cadets rate the tool as useful and clearly reject the idea that it is useless, this high level of acceptance may signal a vulnerability to uncritical acceptance and technological dependency. In the military context, this dependency is particularly alarming, as the potential deterioration of autonomous thinking may compromise the development of essential decision-making capacities. Thus, the tool that facilitates initial understanding of content may, if used without context, end up conditioning or reducing intellectual depth, thereby becoming a hindrance to the development of independent thinking.
In light of this scenario, the pedagogical response should reflect a restructuring of teaching practices to integrate GenAI through active methods, such as brainstorming, ensuring that the cadet remains actively engaged in the process. The central objective should be to ensure that the technology enhances human thinking, strengthening informed decision-making processes rather than replacing or weakening them.
Conclusions
Integration of GenAI into the training of future officers at the Portuguese Military Academy necessitates an “intermediate approach” that avoids both the absolute rejection and the uncritical acceptance of these tools, as advocated by Kelly and Smith. First-year cadets recognise ChatGPT as an effective facilitator of content comprehension and a strategic ally in active methods, particularly for brainstorming. However, the results signal progressive uncertainty and the risk of technological dependence as cognitive task complexity increases. Therefore, effectively integrating GenAI into the Portuguese Military Academy’s education system requires rethinking pedagogical practices to ensure that technology enhances individual capacities rather than substituting autonomous reasoning. For the Portuguese Army and National Republican Guard, the strategic challenge lies in supporting educational efficiency without compromising critical thinking and independent decision-making. These are irreplaceable human competencies for command in complex scenarios. The permanent directive on artificial intelligence has been adopted by the Portuguese Military Academy, as well as a workshop series “Innovate to teach and research” for professors, which aims to strengthen faculty knowledge of the use of AI and innovative pedagogical methodologies in military teaching and research, and has been running from December 2025 to March 2026. Other measures are equally relevant. Two stand out. The first issue pertains to the adoption of the active method, founded on its techniques, for content transmission, idea development, and thinking, in conjunction with the redesign of the assessment process to ensure its active nature. As Watson and Romic suggested, the second involves a seminar cycle on GenAI for cadets to foster critical thinking over uncritical acceptance and reduce apparent technological dependence, thereby strengthening each Cadet’s individual autonomy for future evidence-based decisions.
Limitation
As an exploratory-descriptive study, further comprehensive follow-up research is recommended, such as surveying instructors from comparable classes who use these techniques. This would facilitate cross-referencing to analyse cadets’ perceptions of GenAI as a facilitator, alongside instructors’ observed performance perceptions.
Additionally, further refinement of the quantitative analysis, particularly by examining potential correlations among variables, would enhance the study’s value. For instance, it would be valuable to ascertain whether there is a statistically significant relationship between the frequency of GenAI use and levels of scepticism about thinking.
Another point that could be explored through a qualitative study using focus group techniques is whether uncertainty levels are related to a lack of knowledge about GenAI (ChatGPT) or to a conscious awareness of its true risks.
Ethics Statement
The author affirms that they utilised the artificial intelligence tools Perplexity Pro, NotebookLM and DeepL during the preparation of this article. These tools assisted with language revision, text reformulation, and enhancing clarity. Subsequently, the author conducted a thorough review and editing of all content and undertook a critical analysis of the entire manuscript. The final version is the sole responsibility of the author, who must ensure its integrity and accuracy.
[1] Julija Mironova, Viktoria Riiascochenko, Andrey Bondarenko, Remigijus Kinderis, and Olga Verdenhofa, “Generative Tools of AI in Education,” in New Trends in Contemporary Economics, Business and Management (Vilnius: Vilnius Tech, 2024), 362–67; and Sirine Bouguettaya, Francesco Pupo, Min Chen, and Giancarlo Fortino, “A Meta-Survey of Generative AI in Education: Trends, Challenges and Research Directions,” Big Data and Cognitive Computing 9, no. 3 (2025): 1–25.
[2] Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25.
[3] João Batista, Ana Mesquita, and Gonçalo Carnaz, “Generative AI and Higher Education: Trends, Challenges, and Future Directions from a Systematic Literature Review,” Information 15, no. 11 (2024): 676, https://doi.org/10.3390/info15110676; and Mironova et al., “Generative Tools of AI in Education,” 362–67.
[4] Mironova et al., “Generative Tools of AI in Education,” 362–67; Maureen Ebben and Julien Murphy, “Theorizing the Future of Generative AI in Education,” Learning, Media and Technology (2025); and Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25.
[5] Batista, Mesquita, and Carnaz, “Generative AI and Higher Education”; and Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25.
[6] Batista, Mesquita, and Carnaz, “Generative AI and Higher Education,” 676.
[7] Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25; and Ebben and Murphy, “Theorizing the Future of Generative AI in Education.”
[8] Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25; and Ebben and Murphy, “Theorizing the Future of Generative AI in Education.”
[9] Mironova et al., “Generative Tools of AI in Education,” 362–67; Batista, Mesquita, and Carnaz, “Generative AI and Higher Education”; and Bouguettaya et al., “A Meta-Survey of Generative AI in Education,” 1–25.
[10] Patrick Kelly and Hannah Smith, “How to Think about Integrating Generative AI in Professional Military Education,” Military Review (Online Exclusive, 2024): 1–8.
[11] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[12] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[13] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[14] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[15] Steven Watson and Jonathan Romic, “ChatGPT and the Entangled Evolution of Society, Education, and Technology: A System Theory Perspective,” European Educational Research Journal 24, no. 2 (2025): 205–22.
[16] Watson and Romic, “ChatGPT and the Entangled Evolution of Society, Education, and Technology,” 205–22.
[17] Watson and Romic, “ChatGPT and the Entangled Evolution of Society, Education, and Technology,” 205–22.
[18] Rupali Gaikwad and Aparna Choudhary, “Integrating Artificial Intelligence into Military Education: Opportunities, Challenges, and Future Directions,” Journal of East-West Thought 5, no. 1 (2025): 1317–30.
[19] Gaikwad and Choudhary, “Integrating Artificial Intelligence into Military Education,” 1317–30; and Andryi Bestyuk and Serhii Pokhnatiuk, “Integration of Artificial Intelligence into Higher Military Education as a Factor in Increasing the Efficiency of Professional Training,” Scientific Bulletin of Mukachevo State University. Series ‘Pedagogy and Psychology’ 11, no. 2 (2025): 60–71.
[20] Bestyuk and Pokhnatiuk, “Integration of Artificial Intelligence into Higher Military Education,” 60–71; and Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[21] Ali Mohamed Ali Alnaqbi and Azlina Md. Yassin, “Current Status, Challenges and Strategies of Artificial Intelligence and E-Learning in the UAE Military Education System,” International Journal of Sustainable Construction Engineering and Technology 12, no. 2 (2021): 140–51, https://doi.org/10.30880/ijscet.2021.12.03.034; Bestyuk and Pokhnatiuk, “Integration of Artificial Intelligence into Higher Military Education,” 60–71; and N. Mastorakis, S. Kallou, and D. Papachristos, “The Impact of an AI Course on Hellenic Naval Academy Students’ Attitudes towards AI,” International Journal of Education and Information Technologies 20 (2026): 18–24.
[22] Alnaqbi and Yassin, “Current Status, Challenges and Strategies of Artificial Intelligence and E-Learning,” 140–51; and Bestyuk and Pokhnatiuk, “Integration of Artificial Intelligence into Higher Military Education,” 60–71.
[23] Novruzova Zohrab, “Opportunities, Risks, and Development Paths for Integrating Artificial Intelligence into the Educational Process of a Military University,” System Analysis, Modeling and Optimization 1 (2025): 89–98.
[24] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[25] Gaikwad and Choudhary, “Integrating Artificial Intelligence into Military Education,” 1317–30; Bestyuk and Pokhnatiuk, “Integration of Artificial Intelligence into Higher Military Education,” 60–71; and Zohrab, “Opportunities, Risks, and Development Paths for Integrating Artificial Intelligence,” 89–98.
[26] Gaikwad and Choudhary, “Integrating Artificial Intelligence into Military Education,” 1317–30; Mastorakis, Kallou, and Papachristos, “The Impact of an AI Course on Hellenic Naval Academy Students’ Attitudes towards AI,” 18–24; and Zohrab, “Opportunities, Risks, and Development Paths for Integrating Artificial Intelligence,” 89–98.
[27] Alnaqbi and Yassin, “Current Status, Challenges and Strategies of Artificial Intelligence and E-Learning,” 140–51; and Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
[28] Kelly and Smith, “How to Think about Integrating Generative AI,” 1–8.
