Scientific Integrity in the Age of Generative AI

H

uman history is witnessing one of the most radical breaking points in the production and dissemination of knowledge. Historical turning points, such as the democratization of information with the invention of the printing press or the borderless reach of access with the expansion of the internet, remain merely preparatory stages compared to the transformation we are experiencing today. The process that began in late 2022 with the global accessibility of generative artificial intelligence technologies through Large Language Models has fundamentally shaken the concepts of authorship, originality, and scientific integrity, which are the values that academia has cherished for centuries.

Human intelligence is no longer alone in laboratories, libraries, and at desks; it is joined by a silicon-based partner whose capacity increases by the day, which never tires, and which can at times be dangerously convincing. While this new era comes with the promise of carrying the speed of scientific discovery beyond human imagination, it also leaves the scientific literature face-to-face with an irreversible risk of pollution.

The nature of the scientific production process is confronted with an epistemological dilemma. The processes of hypothesis formulation, literature review, and data processing, which traditionally took shape in a researcher’s mind, have begun to be delegated to algorithms that produce results in seconds. This situation has given rise to a new phenomenon in academia termed “cognitive substitution”. As researchers outsource literature reviews that once took months to AI tools or leave the task of coding entirely to autonomous systems, they face the risk of being deprived of the tacit knowledge inherent in the scientific process.

From a creative author to a curator or editor

When one AI tool performs the literature review of an article, another analyzes the data, and yet another handles the copyediting, the role of the human researcher evolves from a creative author to a curator or editor. This transformation brings with it profound philosophical questions that cannot be explained by efficiency alone. If the AI conceives the idea, processes the data, and writes the article, where does the human stand in this equation? This is the fundamental issue that leaves ethics committees, publishers, and university administrations across the world sleepless.

Looking at the global scale, it is evident that the reactions to this technological tsunami present a fragmented structure based on geopolitical and cultural priorities. The world has not yet been able to establish a common language regarding AI governance. While the European Union, with its Risk-Based approach, places safety and fundamental rights at the center, striving to maintain the protection of research data and transparency at the highest level, the United States exhibits a market-oriented and more flexible stance, wary of stifling innovation. China, on the other hand, with its state-driven strategy, views artificial intelligence as a tool for national development and strategic superiority, creating centralized datasets to increase academic productivity.

This fragmented structure leads to serious inconsistencies in international scientific collaborations. For example, when a European university and an American university conduct a joint project, the issue of which AI model the data will be uploaded to and how the privacy of this data will be ensured can turn into a difficult legal knot to untangle. Although international organizations such as UNESCO attempt to create a human-centered ethical compass, the speed of technological development is far higher than the reaction time of regulatory bodies.

Turning technology into an advantage

The world’s leading universities, after overcoming the initial shock, have begun to move away from a prohibitive mindset toward controlled integration strategies. Resource-rich institutions such as Harvard University have succeeded in turning technology into an advantage by providing their researchers and students with secure, closed-loop sandbox AI tools where data does not leak out. This approach aims to prevent intellectual property infringements that could occur when data is uploaded to publicly available models.

Oxford University follows a more liberal but responsibility-oriented policy, allowing the use of AI in idea development and data analysis, yet emphasizing that the human researcher is ultimately responsible for every output produced. Institutions located in the heart of Silicon Valley, such as Stanford University, have adopted a decentralized, flexible model, largely leaving the decision to instructors and project leads. The common denominator where all these different approaches intersect is the principle of transparency. The use of artificial intelligence must cease to be a secret and must be explicitly declared in the methodology section.

The academic ecosystem in Türkiye has not remained indifferent to this global transformation, clarifying its own roadmap particularly with steps taken in 2024 and 2025. The ethical guidelines published by the Council of Higher Education (YÖK) and TÜBİTAK of Türkiye have largely resolved the uncertainty by providing a binding framework for Turkish universities. The most critical point highlighted in these guidelines is that AI can only be used as a supportive tool, whereas higher-order cognitive processes such as hypothesis formulation or the interpretation of results must belong to the human mind. Turkish universities are updating their institutional policies in line with this framework.

While some universities position AI as a student-centered educational tool that does not reduce human interaction, others are ensuring integration into engineering education by establishing centers aimed at increasing technical capacity. At the same time, they are developing rigorous yet innovative procedures that are compatible with international standards regarding academic integrity and data privacy. Türkiye’s approach is based on establishing a culture of responsible and reliable use by pulling technology within ethical boundaries rather than rejecting it.

The dark side of technology

However, there is also a dark side to this bright technological picture. The academic publishing sector is under attack from fake articles produced by AI-powered paper mills. Scientific literature is at risk of being polluted by articles that are produced by AI, sound pleasant, but are empty or filled with entirely fabricated data. An anatomically impossible rat image (huge rat genitals scandal) recently published in a reputable journal and clearly produced by AI has exposed how vulnerable the peer-review system has become against this new threat.

Furthermore, the term “tortured phrases” which entered the literature as a result of some researchers using AI to bypass plagiarism software by replacing established scientific terms with strange synonyms, is disrupting the integrity of scientific language. Although publishers initially relied on AI detectors, they have been forced to change strategy due to these tools unfairly accusing non-native English-speaking authors (false positives) and failing to catch slightly modified AI texts. Their primary focus is now built upon authors presenting their raw data and transparently declaring their processes through various checklists and tools, rather than technological policing.

Despite all these risks, the momentum that artificial intelligence adds to the scientific discovery process is undeniable. Google DeepMind’s AlphaFold 3 project solving protein structures with atomic precision or the GNoME project discovering millions of new materials proves that AI is indispensable in deciphering data patterns that exceed the limits of the human mind. Research that would take months or even years with traditional methods can now be condensed into hours. It holds a massive potential that clears the path for humanity in many fields, from drug development to clean energy technologies.

In conclusion, the world of academia is walking a delicate balance between speed and reliability. Artificial intelligence has accelerated the production line of science but has strained quality control mechanisms. The scientists who will be successful in the future will not only be those who master their fields but also those who can manage these powerful tools most accurately from both an ethical and technical perspective. The quantitative production frenzy created by the “publish or perish” pressure reaches an unsustainable point when combined with AI.

The way out of this crisis lies not in banning technology, but in building a new scientific culture that centers on human responsibility, transparency, and verification. Science, regardless of how advanced the tools become, is essentially a human endeavor, and it will once again be the human conscience and intellect that have the final word in the effort to reach the truth.

(Originally published in Turkish by Kriter)