Background
Podcasts are widely used as a low-threshold tool, but their use is often passive, and depending on the context and the self-regulation skills of the recipients, there is a substantial potential for distraction while listening. Therefore, the mere availability of podcasts often has no significant effect on relevant learning outcomes. Self-assessments, such as in the form of exercises or "mock exams," become effective only when they a) are closely aligned with the learning objectives of the course and the available learning materials (e.g., podcasts, lecture slides), and b) are used regularly. Attempts to combine podcasts with self-assessments have not yet been made at the University of Bern.
Objective
The objective of this project is to use AI to generate self-assessment tasks from podcasts and other lecture materials and to evaluate their impact on students' learning outcomes. These self-assessment tasks will not only be made accessible to students but will also be reviewed by experts for accuracy, effectiveness, and efficiency. This ensures that the tasks are at an appropriate performance level. Ultimately, such AI-generated self-assessments should be used across different subjects and, with minor adjustments, be beneficial throughout the University of Bern.
Approach
Podcasts from a wide range of university courses will be automatically transcribed using speech software, to be used alongside other documents (e.g., slide sets, exam literature) as the basis for learning and self-assessment materials. In a second step, the questions developed by the LLM will be presented to the lecturers for review. Their expertise is needed to assess whether the questions adequately cover and test the expected knowledge from the lecture. In a third step, the efficiency and effectiveness of the LLM-generated tasks will be empirically evaluated by students' work. The following questions will be addressed: Which combination of digital teaching and learning methods is associated with better grades, greater satisfaction, better evaluations, and higher knowledge? Does the application of digital methods show long-lasting positive effects in a normal learning environment?
The insights gained and the developed tools will be made widely available to the entire University of Bern.
Project Information
Partner: Prof. Dr. Michael Schulte-Mecklenbeck (IMU)
Runtime: 05/2024 - 03/2027
Funded by: Digitalisation Committee of University of Bern (DigiK)
