AI-Supported Question Generation for Scientific Reading
MSc Thesis Project · Educational Technology · Saarland University. As part of my MSc in Educational Technology, I investigated how AI-generated questions can support scientific reading comprehension and cognitive engagement.
My role
Research Design · Learning Design · Educational Evaluation
Value created
Within the limits of the thesis study, the results suggested that AI-generated questions can support scientific reading when they are connected to clear learning objectives, cognitive complexity, and thoughtful evaluation.
Case story
Overview
As part of my MSc in Educational Technology at Saarland University, I investigated how AI-generated questions can support scientific reading comprehension and cognitive engagement. The study used OwlMentor, an AI-supported learning prototype, to explore how question generation can scaffold active reading, self-explanation, and reflection.
Learning Problem
Scientific reading requires more than moving through a text. Learners need to identify central ideas, connect concepts, monitor understanding, and notice where their comprehension is incomplete.
Research Question
The thesis examined how AI-supported question generation might support scientific reading comprehension, self-explanation, reflection, and cognitive engagement during reading.
My Role
My role covered research design, learning design, educational problem framing, question-generation concepts, question type design across cognitive complexity levels, contribution to question formats based on learning objectives, suggested evaluation approaches including learning evaluation techniques and rubric-based AI evaluation concepts, study evaluation, data analysis, interpretation of findings, and discussion of educational implications.
Research Prototype: OwlMentor
OwlMentor was used as the research prototype for this thesis study. The core software implementation was developed by my supervisor / under academic supervision. My work focused on the educational design, research methodology, question framework, evaluation strategy, and analysis of learning outcomes.
Question Design Framework
I contributed to designing question types and complexity levels to support different learning objectives and cognitive processes. The question framework treated questions as prompts for comprehension, self-explanation, reflection, and deeper engagement rather than as simple quiz items.
Study Design
The study design connected the educational problem, generated question types, learner interaction, and evaluation strategy. Evaluation thinking included learning evaluation approaches and rubric-based concepts for assessing AI-generated question quality.
Key Findings
Within the limits of the sample, the thesis study indicated that AI-generated questions may support scientific reading when they are aligned with learning objectives, cognitive complexity, and meaningful reflection. The results suggested educational promise, while also showing the need for careful evaluation and interpretation.
Limitations
The findings should be interpreted cautiously because they come from a thesis study with a limited context and sample. The work should not be read as evidence of a launched product or published research.
What I Learned
I learned that AI-supported learning design depends on more than generating good outputs. It requires clear learning objectives, thoughtful question framing, evaluation criteria, and an understanding of how learners engage with questions during real reading tasks.
Thesis Materials
MSc thesis and thesis presentation materials can be linked here when the PDFs are available. They should be presented as thesis materials, not publications.
Process artifacts
Evidence of the thinking behind the work
Research framing
Scientific reading can look complete even when learners still struggle to explain key ideas, connect concepts, or identify gaps in their understanding.
Workflow sketch
The thesis used OwlMentor, an AI-supported learning prototype, to explore how generated questions could scaffold active reading, self-explanation, and reflection. The core software implementation was developed by my supervisor / under academic supervision.
Design decision
The project helped me understand how educational value depends not only on AI generation quality, but on learning objectives, question framing, evaluation criteria, and how learners use questions while reading.
Materials & Links
Open the project evidence
Reports, posters, demos, code, and related links are collected here so visitors can inspect the work behind the case study.