Generative AI Use and Academic Integrity Behaviour among Law Students in Bangladesh
This study of Bangladeshi law students reveals a significant gap between their high endorsement of verifying AI-generated legal materials and their lower willingness to disclose AI assistance, suggesting a need for distinct educational workflows to address verification and transparency separately.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the modern classroom, a new tool has arrived that can write essays, explain complex ideas, and draft legal arguments in seconds. This technology, known as generative artificial intelligence, offers students a powerful shortcut, but it brings a difficult question: how do we keep learning honest when a machine can do so much of the work? Educators and researchers are trying to understand two specific rules that students must follow to stay honest. The first is verification, which means checking that the facts and laws the computer gives you are actually true and exist in the real world. The second is disclosure, which means telling the teacher, "I used a computer to help me write this." While these two rules often get lumped together as "academic integrity," they are actually very different tasks. One is about finding the truth, and the other is about being transparent. Understanding whether students treat these two duties the same way is crucial for designing fair tests and teaching methods in law schools and beyond.
A team of researchers in Bangladesh set out to see how law students in their country handle these two distinct rules. They gathered responses from 245 students currently studying law at a university in Dhaka. The researchers asked these students to rate their agreement with statements about their own habits and their sense of duty. They wanted to know if students felt a strong need to check the computer's work, if they felt a strong need to admit when they used it, and if their feelings about these duties matched what they actually said they did. The study focused on a specific group of students to ensure the answers reflected the mindset of those training to become lawyers, a profession where a single made-up fact can have serious consequences.
The results revealed a clear and surprising split in how students view these two obligations. When asked about their actual habits, 83.3 percent of the students said they check the legal sources and facts provided by the artificial intelligence to make sure they are correct. However, only 65.3 percent said they actually disclose or admit when they have used the tool to help them. This gap of 18 percentage points suggests that students see checking the work as a much more important part of their job than admitting they used help. The researchers found that this pattern held true even when they asked about what students should do rather than what they did do. In terms of duty, 82.4 percent felt responsible for verifying the information, while only 66.9 percent felt responsible for disclosing the assistance. This means the difference is not just about students failing to do what they promise; it is a deeper difference in how they value these two specific actions.
The study also looked at whether students who felt a strong sense of duty actually followed through with their actions. Surprisingly, the data did not show a large gap between what students said they ought to do and what they said they actually did. For both checking facts and admitting help, the difference between the "should" and the "did" was small and statistically uncertain. This finding challenges a common assumption that students simply know the rules but choose to ignore them. Instead, the evidence suggests that for these students, the act of checking a legal source feels like a natural, necessary step in their work, while the act of declaring they used a computer feels like a separate, perhaps optional, communication task. It is as if they view the computer as a research assistant that must be vetted, but not necessarily as a co-author that must be named.
The researchers were careful to note that they could not prove these students were telling the absolute truth about their behavior, nor could they determine exactly how many were engaging in academic misconduct. The study relied on what the students reported about themselves, and the data came from a specific group that was not randomly selected from all law students in the country. Furthermore, the researchers could not verify the original source of the digital records they analyzed, so they treated the findings as a snapshot of reported attitudes rather than a confirmed record of behavior. Despite these limits, the pattern was consistent across different ways of analyzing the data. Even when the researchers looked at students who used the technology frequently or those who had received training, the gap between checking and admitting remained.
This study suggests that for law students in Bangladesh, the concept of "integrity" is not a single block of rules but a collection of different habits. They are highly motivated to ensure the legal facts they use are accurate, likely because they understand that a lawyer cannot rely on a fake case. However, they are less consistent about the requirement to be transparent about using the tool itself. The researchers propose that schools should not treat these two issues as the same problem. Instead of a single policy on "using AI," educators might need to teach two separate workflows: one for rigorously checking the authority of legal sources, and another for clearly documenting how technology was used in the drafting process. By separating these tasks, teachers can help students understand that being honest about their process is just as vital as being accurate in their facts, even if the students currently feel differently about the two.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.