Overview of the position
- Basic and applied research
- Standard full-time program (4 years) – funding provided with a scholarship; no tuition fees
- In-person in Brno, Czech Republic – read why it's a great student city
- At Masaryk University, Faculty of Informatics
- Accredited program with a world-wide recognized Doctoral degree
- Learn about the university's Ph.D. admissions
- Learn about the faculty's Ph.D. program
- Learn about the lab's approach to Ph.D. studies
- Deadline for preparing all application steps: 21 April 2025
- Final deadline for the application submission: 21 May 2025
- Enrollment to the Ph.D. program: 1 September 2025
Research description
Motivation
- Various types of algorithmic biases influence the behavior of machine learning (ML) models, leading to inaccuracies or even unfairness in their applications.
- It is essential to study properties of biases to better understand them, and, as a result, inform stakeholders and decision-makers about how to mitigate the issues.
Goals
- Systematically evaluate ML models for learning analytics, specifically:
- Explore data biases and data augmentation in real-world datasets.
- Investigate method biases within ML models.
- Analyze the ML model sensitivity across various practical use cases.
Methods and tasks
- Experimentally manipulate the ML model training data, inputs, and parameters to observe changes in model properties.
- Conduct power analysis with authentic datasets from exercises in cybersecurity (and possibly other domains).
- Evaluate the model behavior within practical applications for hands-on training in cybersecurity (and possibly other domains).
Examples of related publications
- V. Švábenský et al., 2025: Evaluating the Impact of Data Augmentation on Predictive Model Performance
- O. Deho et al., 2024: When the Past != The Future: Assessing the Impact of Dataset Drift on the Fairness of Learning Analytics Models
- J. Čechák, 2024: Exploring Statistical Biases in Educational Data
- S. Barocas et al., 2023: Fairness and Machine Learning: Limitations and Opportunities
Target publication venues
- International conferences: LAK, EDM, AIED, SIGCSE, ITiCSE, ICER…
- International journals: JLA, IJAIED, BJET, Computers and Education…
Context and people
- Supervisor: Pavel Čeleda, a well-established researcher (89 Scopus-indexed research publications with 3600+ citations), who has extensive experience with advising doctoral students (8 defended, 2 ongoing).
- Co-supervisor: Valdemar Švábenský (Ph.D. 2022), who published 39 research papers in the field and has 3 years of international research experience across several countries, mainly USA and Japan.
- Thanks to our active collaboration with international researchers, there is potential to complete internships at universities abroad (see below).
- The research is conducted in connection to the recent grant project “Algorithmic Biases in Machine Learning Models in Education” (funded by the Czech Science Foundation).
Selected international collaborators
- University of Pennsylvania, USA
- Michigan State University, USA
- Kyushu University, Japan
- University of South Australia, Australia
Requirements
- Master's degree in a related discipline (e.g., computer science).
- IT skills: Python, ML, Git, and LaTeX.
- Some level of practical experience, not only theoretical knowledge, is required.
- Basic proficiency should be sufficient to apply.
- Some level of practical experience, not only theoretical knowledge, is required.
- English language skills: at least upper-intermediate level (B2) or higher.
If you have any questions or wish to apply
- Email Valdemar Švábenský:
- Subject: “Ph.D. position in learning analytics”.
- Attach your CV/résumé (PDF file, max. 2 pages, including 2–3 reference contacts).
- Applicants will be expected to complete a practical assignment on the topic to demonstrate the required skills.
- The assignment will be specified upon application.
- This requirement may be exceptionally waived if the applicant already has outstanding prior track record related to the topic (e.g., own open-source software projects, research experience, and ranked publications).