The role and impact of artificial intelligence in the development and evaluation of doctoral theses: a qualitative analysis

In the quiet, years-long journey of earning a doctorate, a student is expected to do more than just learn what is already known; they must push the boundaries to create something entirely new. This process, known as a doctoral thesis, is the ultimate test of a scholar's ability to think independently, gather evidence, and construct original arguments. For decades, the tools of this trade were books, databases, and the slow, deliberate work of writing and editing by hand. But recently, a new kind of partner has entered the room: artificial intelligence. These are computer systems capable of generating text, summarizing vast amounts of information, and correcting grammar with startling speed. They are not merely spell-checkers; they are engines that can draft entire sections of a paper or suggest complex research paths. The question now facing universities is not whether these tools will be used, but how they change the very nature of learning and truth. If a machine helps write a thesis, who is the author? Does the student still understand the material, or have they simply become a manager of the machine's output? This uncertainty strikes at the heart of what it means to be a scholar.

A researcher set out to explore this shifting landscape by listening directly to the people living through it. They conducted a detailed study involving twenty individuals from three universities in Tunisia: ten doctoral students who were currently writing or had recently finished their theses, and ten university teachers who had supervised them. The researcher did not rely on surveys or statistics; instead, they held long, open conversations, asking these participants to describe their daily experiences with these new technologies. They wanted to understand the real-world mechanics of how AI is being woven into the fabric of academic work, from the first spark of an idea to the final evaluation of the finished document.

The students described a workflow that has been fundamentally altered by these tools. In the early stages of research, when a student must find relevant studies among thousands of possibilities, they now turn to AI assistants that can scan databases and summarize findings in seconds. One student noted that a tool designed for this purpose allowed them to identify key articles without spending hours manually browsing, a task that used to consume days. As the writing began, the role of the machine expanded. Students used these systems to organize their thoughts, generate outlines, and even draft initial paragraphs when they felt stuck. For those writing in a language that was not their native tongue, grammar-checking software became an essential partner, smoothing out sentences and ensuring a professional tone. In the final stages, before submission, students used detection software to check their own work, ensuring that they had not accidentally included text that looked too much like it was generated by a machine.

However, the researcher found that this reliance on technology created three distinct types of users. Some students used the tools strictly for technical tasks, like fixing grammar or translating a sentence, keeping the heavy lifting of thinking entirely to themselves. Others used the technology strategically, treating it as a sparring partner to challenge their ideas or find new angles, always checking the machine's work against their own knowledge. A third group, however, began to depend on the tools for the core of their thinking, delegating significant portions of the intellectual work to the algorithm. This shift raised a quiet alarm among the researcher: when a student lets the machine do the thinking, does the student still learn how to think?

The teachers interviewed offered a different, more cautious perspective. They acknowledged that the tools could make the writing process faster and the final text more polished. Yet, they expressed deep concern about the loss of the "human touch." One professor observed that work was beginning to look too similar, as if the unique voice of each student was being smoothed over by the same algorithmic hand. The teachers also faced a new and difficult challenge in their role as evaluators. They had to determine whether a thesis was the product of a student's mind or a machine's output. They tried using detection software to spot AI-generated text, but they found these tools to be unreliable. Sometimes, the software flagged a human-written paragraph as fake, and other times, it missed sophisticated AI contributions entirely. This uncertainty left them in a bind, unsure of how to judge the originality of a piece of work that might be a hybrid of human and machine.

The study revealed a tension between efficiency and integrity. On one side, the tools offered undeniable benefits: they saved time on repetitive tasks and allowed students to focus more on critical analysis. On the other side, there was a fear that this efficiency came at the cost of deep understanding. If a student uses an AI to summarize a complex theory, they might get the gist of it without ever grappling with the difficult details themselves. The researcher noted that this could lead to a "superficialization" of knowledge, where the thesis looks perfect on the surface but lacks the depth that comes from struggling with the material. Furthermore, the study highlighted an issue of fairness. Not all students had access to the same powerful versions of these tools. Some could afford premium subscriptions that offered better features, while others were limited to free versions, creating an uneven playing field where the quality of the final work might depend more on a student's budget than their intellect.

Ultimately, the researcher concluded that the relationship between students and their supervisors is changing. The teacher is no longer just a guide to knowledge but is becoming a mediator of technology, responsible for teaching students how to use these tools ethically and critically. The study suggests that universities cannot simply ban these tools, nor can they ignore them. Instead, they need to create clear rules that define what is acceptable use and what crosses the line into academic misconduct. They must train teachers to recognize the signs of AI dependence and to evaluate students based on their ability to use these tools wisely, rather than just on the final product. The goal is to find a balance where technology aids the creative process without replacing the human mind that must ultimately take responsibility for the ideas. The paper does not claim to have solved this problem, but it provides a clear map of the terrain, showing that the future of higher education depends on navigating these new waters with care, ensuring that the pursuit of knowledge remains a deeply human endeavor.