ThinkBox: Integrating generative artificial intelligence into graduate studies in administration: impacts, ethical dilemmas, and methodological paths Open Access

The rapid integration of artificial intelligence (AI), particularly Generative AI (GenAI), into graduate programs in business administration is transforming teaching and research. Tools such as ChatGPT, Gemini, Claude and other text-generation models now assist with tasks once entirely manual, from literature reviews to statistical analyses. This progress raises profound ethical and methodological dilemmas: How can academic integrity and originality be preserved when algorithms contribute to the production of knowledge? Graduate programs worldwide are debating policies for the responsible use of GenAI, but the pace of discussion varies across contexts. Globally, the adoption of AGI in higher education is expanding rapidly, with a 43% increase in applications between 2018 and 2022 (Zawacki-Richter, Marín, Bond and Gouverneur, 2019). Since the release of ChatGPT, growth has accelerated further. Universities are launching ambitious initiatives, such as the University of Eindhoven’s Institute for AI Systems, which includes 50 new chairs (Zawacki-Richter et al., 2019) – highlighting the relevance of GenAI. However, institutional guidelines remain underdeveloped in Brazil. Agencies such as CAPES and CNPq are yet to establish national frameworks, leaving researchers reliant on policies set by global technology companies (Sampaio, Sabattini and Limongi, 2024).

Our essay critically examines the impacts, dilemmas and methodological pathways for responsibly integrating GenAI into graduate programs in business administration. We discuss the decline of the traditional paradigm of scientific methodology under the “datafication” of science, analyze GenAI’s epistemological potential in relation to ethical principles, explore its practical impacts and risks (such as algorithmic authorship, bias and overdependence) and assess new ethical responsibilities in academic advising. We also evaluated GenAI’s benefits and limitations by comparing global initiatives in regulation and cooperation (the EU AI Act, UNESCO, OECD and the Stanford AI Index) with the Brazilian context. We argue that methodological re-education, together with faculty training, equitable access to technology and the preservation of critical autonomy, is essential for integrating GenAI responsibly into graduate research.

For decades, methodological training in business administration followed a rigid, linear model: formulating a problem, reviewing the literature, defining hypotheses, selecting methods, collecting data, analyzing the results and drawing conclusions. Although useful for beginners, this paradigm shows apparent limitations in the era of big data and GenAI. Currently, datafication enables inductive and iterative approaches that break from linear thinking. Machine learning and data mining can detect previously unimaginable patterns, automatically generate hypotheses and accelerate the discovery process (Heidt, 2025). As Williamson, Bayne and Shay (2020) note, “the availability of big data, the rise of data science and the development of deep learning algorithms have established new modes of quantitative knowledge production and decision-making.” Methods are becoming less sequential and more adaptive, informed by real-time data-driven insights from platforms such as SciSpace and AnswerThis.

Clinging to rigid models risks overlooking the research opportunities. Instead of beginning with fixed hypotheses, researchers can explore open datasets with GenAI, identifying trends before formulating theories and reversing traditional steps. Ethical concerns, previously relegated to conclusions, now require attention during the research design phase given GenAI’s involvement in issues such as bias, privacy and reproducibility. Knox (2020) highlighted that advances in GenAI for education often follow nonlinear trajectories, challenging conventional sequences of scientific progress.

The traditional paradigm is limited by the lack of reflexivity. Zawacki-Richter et al. (2019) observed through a systematic review that research on GenAI in higher education predominantly emphasizes accuracy and performance, frequently overlooking considerations of risk and pedagogical frameworks. These findings highlight the importance of adopting a more adaptable, reflexive and multi-dimensional methodological approach. Such a paradigm should combine iterative experimentation with ethical considerations from the outset, embrace interdisciplinarity and combine classical methods with computational analysis. In short, the “recipe-book” approach gives way to a hybrid methodological ecosystem where data science, digital humanities and critical epistemology coexist. Graduate programs must update their curricula or risk-producing researchers who are disconnected from contemporary practices.

Rather than undermining academic rigor, GenAI can serve as an epistemological ally when used in accordance with ethical principles. Current GenAI tools already support research: language models accelerate text drafting, writing assistants improve style and grammar, algorithms scan vast bibliographic databases, some with more than 200 million studies analyzed (SciSpace, 2025) and intelligent systems automate systematic reviews. These technologies expand cognitive capacity, allowing researchers to focus on interpretation and implications rather than on repetitive tasks. Buriak et al. (2023) noted that many scholars have already employed GenAI as a “personal proofreader,” enhancing clarity before submission. For non-native English-speaking graduate students, GenAI mitigates linguistic barriers, helping to ensure that intellectual contributions are not overshadowed by writing limitations.

When the use of GenAI reduces linguistic barriers, the less visible risk of “data colonization” models trained primarily on Anglophone corpora tends to standardize styles, ontologies and relevance criteria, rendering local problems, sources and practices invisible. In graduate student training, this can shift researchers’ attention toward global agendas devoid of context, impoverishing the capacity to formulate questions anchored in regional realities. Mitigation involves curricular and methodological decisions, including building and prioritizing local corpora (such as Brazilian Portuguese, regional journals and databases), guiding prompts that require coverage of Latin American literature, demanding disclosure of the linguistic coverage of the models used and establishing human checks of contextual relevance before incorporating evidence into manuscripts. In parallel, data governance and open science policies must address asymmetries in the extraction and circulation of academic data, lest they reproduce, in the scientific field, power dynamics already described in decolonial AI critique and data coloniality theory (Mohamed, Png and Isaac, 2020; Couldry and Mejias, 2019).

However, the epistemological value of GenAI depends on its robust ethical frameworks. Floridi and Cowls (2022) identify five principles recurring across ethical codes: beneficence, non-maleficence, autonomy, fairness and explainability. In practice, GenAI should promote research and societal benefits, avoid harm, support researchers’ autonomy, reduce bias and ensure transparency. Researchers must document how GenAI tools are used (e.g. acknowledging GPT-generated drafts reviewed manually), ensuring accountability and methodological clarity.

Based on these principles, GenAI can enhance knowledge production. For instance, algorithms can identify theoretical trends across thousands of articles; however, researchers must explain the selection and analysis criteria. GenAI-powered tools can also expose linguistic or geographic biases, enabling more inclusive scholarship. Significantly, ethical use strengthens critical thinking rather than replacing it. Suggested arguments must be evaluated against theory and evidence, and algorithmic outputs should be scrutinized for validity. Floridi and Cowls (2022) emphasize that transparency is key to maintaining human–machine dialog. Ultimately, GenAI should be viewed as a partner that expands creativity and precision, without replacing human judgment and responsibility.

The presence of GenAI in graduate education has produced tangible benefits and risks. Students use writing assistants to refine their manuscripts, apply machine learning in Python or R to analyze large datasets and consult chatbots as real-time tutors. Tasks, such as transcription and reference formatting, are increasingly being automated. However, these efficiencies are accompanied by dilemmas that challenge the academic norms.

One major debate concerns algorithmic authorship (Stokel-Walker, 2023). In 2023, several papers listed ChatGPT as a coauthor (Thorp, 2023), sparking backlash. Editors and scholars have emphasized that GenAI lacks responsibility, consent and intellectual contributions. Journals such as Science and Nature quickly clarified policies: GenAI can assist but cannot be listed as an author. This episode raises broader questions: If students use GenAI extensively, are they still authors or curators of machine-generated text? Institutions are beginning to draft codes of conduct to address these issues.

Bias is another important issue. GenAI systems replicate the biases present in the training data. Even advanced models, such as GPT-4, exhibit implicit biases, for example, associating women less with STEM fields or preferring men for leadership roles (Stanford University, 2025). These biases can distort research outcomes unless carefully mitigated. Strategies such as algorithmic audits and balanced datasets are recommended; however, they require specialized technical expertise and time.

Concerns about plagiarism and integrity are growing. GenAI can generate essays in minutes, raising questions regarding their originality. Hallucinations, fabricated references or inaccurate claims further undermine reliability (Buriak et al., 2023). Overreliance risks are turning GenAI into an “intellectual crutch,” discouraging critical thinking (Balasubramanian, Ye and Xu, 2022). Preventing this requires cultivating discernment and critical autonomy among the students. In short, GenAI offers significant benefits but also introduces dilemmas related to authorship, bias, plagiarism and overreliance on technology. The challenge is to avoid both rejection and uncritical adoption instead of fostering responsible, transparent and critical use.

The role of advisors in graduate programs has been redefined. Traditionally, advisors have enforced methodological rigor and supervised compliance with academic standards. Now, they must also act as mentors, guiding students through a hybrid human–GenAI ecosystem (Schmidgall et al., 2025). This shift required a new ethical framework.

Advisors must encourage the critical and responsible use of GenAI rather than banning it or allowing it to be used uncritically. For example, if students use chatbots for writing assistance, advisors should ensure that the content is verified, revised and appropriately attributed. Buriak et al. (2023) emphasized that GenAI lacks analytical judgment, making human interpretation essential. Advisors must also teach AI literacy, including verifying references, understanding dataset limitations and reporting on the use of GenAI in a transparent manner.

Institutional codes of conduct are of significant importance. Although international universities have started issuing guidelines, Brazil remains in its early stages (Sampaio et al., 2024). These guidelines emphasize methodological transparency, AI integrity and continuous researcher training. Advisors should take the lead in adopting and promoting such practices. In addition, faculty training is essential. Many advisors educated before GenAI’s emergence might not have been familiar with the latest technology. Workshops and training programs can equip students with the skills to guide them effectively (Zhai, 2024). Without proper preparation, there is a higher risk of misguided bans or overly permissive approaches.

Finally, classical academic values should be reinforced in new ways. Honesty now extends to disclosing the use of GenAI, and autonomy refers to resisting intellectual dependence on algorithms. Advisors and students should engage in a collaborative learning process, learn from one another and cultivate critical thinking skills and independence. The ethical integration of GenAI will depend heavily on advisors’ preparedness and engagement.

The main benefits of GenAI in graduate education include:

1. Accelerated literature reviews: GenAI can quickly synthesize large volumes of research, providing an overview of key debates (SciSpace, 2025);

2. Novel hypothesis generation: By detecting hidden patterns, GenAI suggests hypotheses that researchers can overlook;

3. Bias and inconsistency detection: GenAI helps identify sample imbalances and logical errors;

4. Operational efficiency: Tasks such as transcription, translation and reference formatting were automated and

5. Personalized learning: GenAI functions as a tutor, offering tailored explanations and feedback.

The principal risks include:

1. Shallowness: Over-reliance may lead to formulaic, superficial analyses;

2. Inequality: Limited access to advanced tools risks widening gaps between institutions and regions;

3. Dependency: Students may fail to develop independent skills (Balasubramanian et al., 2022);

4. Integrity concerns: Risks include plagiarism, fabricated content and compromised data privacy and

5. Bias and opacity: Pretrained models often carry biases and lack transparency, thereby undermining reproducibility.

Thus, GenAI should be used for repetitive, large-scale tasks under human supervision, whereas human judgment must guide originality, interpretation and ethical decisions. Optimal integration strikes a balance between efficiency and intellectual depth.

The integration of GenAI into graduate education is a global issue. The EU has taken the lead with the AI Act (2024), imposing strict rules for high-risk systems, such as those used in education. The U.S. has pursued fragmented initiatives, including the AI Bill of Rights (2022) and NIST frameworks, favoring innovation along with self-regulation. At the multilateral level, UNESCO’s Recommendation on the Ethics of AI (2021), adopted by 193 member states, emphasizes transparency, fairness, accountability and human oversight. The OECD’s 2019 AI Principles, as well as more recent convergent frameworks from the EU, OECD, UN and AU, emphasize transparency and trustworthiness (Stanford AI Index, 2025). Initiatives such as the GPAI and G7 dialogs foster further alignment.

In Brazil, progress has been slow. The 2021 National AI Strategy set broad goals, but regulatory frameworks remain under debate in the Congress. Universities have begun issuing their own guidelines, and events at the USP and UNESP have reflected growing awareness. However, there is a lack of national coordination. Brazil should align with international frameworks and strengthen its cooperation through joint research and open-access AI initiatives to reduce inequalities.

The integration of AI into graduate education in business administration is inevitable and transformative. This signals the decline of linear and traditional methodologies and the rise of data-driven, interdisciplinary approaches. However, this also requires methodological re-education, ethical guidance and institutional adaptation.

Successful integration depends on striking a balance between benefits and risks: leveraging efficiency while safeguarding originality, equity and integrity. Ethical principles must guide the use of GenAI, and transparent disclosure should become standard practice. Equitable access is critical in preventing the widening of inequalities, and institutional support for open access platforms can play a significant role.

Future research should address the unresolved question of how long-term AI use affects cognitive development. How should authorship be attributed to human–machine collaboration? What new forms of assessment are required when students use AI? Will the massive adoption of science make it more innovative or conformist? Addressing these issues is essential for shaping the future of academia.

Ultimately, AI should be viewed not as a substitute for human intelligence but as a partner. Graduate programs must model responsible practices, fostering creativity, curiosity and ethical responsibility. If aligned with human judgment and reflection, AI can make graduate education more inclusive, agile and profoundly human-like.