Investigating Generative AI for Personalized College Assignments Based on Student Interests
This paper presents Taklif.AI, a serverless generative AI platform that leverages Llama 3.3 and structured prompt engineering to create personalized college assignments based on students' unique interests and cultural contexts, demonstrating positive user reception while addressing engagement and ethical challenges in education.
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 classrooms around the world, teachers face a quiet but persistent struggle: how to make a single lesson feel personal to dozens of students who think, learn, and care about very different things. A student who loves astronomy might find a standard math problem about traffic lights boring, while a student obsessed with basketball might struggle to see the point of a generic coding exercise. For decades, educators have known that connecting schoolwork to a student's own passions can spark motivation and deepen understanding, but doing this for every single assignment in a large class is a task that would consume a teacher's entire life. The arrival of powerful computer programs capable of writing human-like text has opened a new door. These tools, known as large language models, can read a standard assignment and rewrite it to fit a specific interest, theoretically offering a way to give every student a unique version of the same lesson without requiring extra hours from the instructor.
A team of researchers from the Islamic University of Gaza and the Università del Salento in Italy set out to test whether this technology could actually work in a real college setting. They built a web platform called Taklif.AI, designed to take a standard assignment—such as a programming task or a mathematics problem—and automatically reshape it to align with a student's hobbies, like football, video games, or music. The system does not change the core lesson or the difficulty of the work; instead, it acts as a translator, taking the original requirements and wrapping them in a new story that speaks to the student's personal interests. To ensure the computer does not wander off-topic or produce unsafe content, the researchers programmed strict rules, or guardrails, that check both the student's input and the computer's output before anything is shown.
The researchers tested their system with 68 participants, including 65 undergraduate students and three educators. Each person used the platform to personalize one assignment of their choice. The results were encouraging. When asked if the personalized version felt more engaging than a standard one, 84 percent of the participants agreed or strongly agreed. The vast majority also found the system easy to use and felt that the original learning goals were preserved. The platform successfully generated examples such as a programming challenge about finding pairs of numbers, which was rewritten for a student interested in astronomy as a task about aligning telescope lenses to match starlight intensities. Another example transformed a set theory problem into a scenario about analyzing basketball teams and their shared players. In these instances, the mathematical logic and the rules of the problem remained exactly the same, but the context shifted to match the student's world.
However, the researchers are careful to note what their study has not yet proven. While the students reported that the assignments felt more interesting, the team did not measure whether this increased interest actually led to better test scores, deeper understanding, or fewer instances of academic dishonesty. The study was a preliminary check on whether people liked the idea and found the tool usable, not a final verdict on whether it improves learning outcomes. The system also has limits; it relies on students to honestly report their interests, and it cannot yet handle every type of assignment or subject with equal ease. The researchers acknowledge that highly abstract topics might be difficult to reframe without losing their essential nature, and they warn that too much storytelling could distract from the actual work.
The project represents a significant step in moving artificial intelligence from a tool that simply answers questions to one that helps design the learning experience itself. By combining a flexible computer architecture with careful safety checks, the team demonstrated that it is possible to generate unique, interest-based assignments at a scale that was previously impossible for human teachers to manage. The next phase of their work will involve a more rigorous comparison, pitting students who receive these personalized assignments against those who receive standard ones to see if the personalization truly changes how much they learn. For now, the study suggests that when technology is used to bridge the gap between a student's life and their schoolwork, the result is a classroom experience that feels less like a chore and more like a conversation.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.