Early Prediction for Targeted Interventions in Educational Games
Event details
| Date | 24.08.2021 |
| Hour | 11:00 › 13:00 |
| Speaker | Lucas Ramirez |
| Category | Conferences - Seminars |
EDIC candidacy exam
exam president: Prof. Pierre Dillenbourg
thesis advisor: Prof. Tanja Käser
co-examiner: Prof. Robert West
Abstract
The open-nature and heterogeneity of open-ended
learning environments (OELEs) makes processing their interaction
data particularly challenging. Learner modeling methods
developed for a specific OELE often do not transfer to other
OELEs easily, if at all. This diminishes the generalizability of
data processing techniques and reproducibility of experimental
results. To tackle this issue, we propose to define an intermediate
representation for student traces derived from interaction data
that could be used universally among a well-defined subset
of OELEs. Data processing algorithms could then leverage
this shared representation on interaction data originating from
different OELEs without methodological changes. We hope our
contribution can lower the effort required to create and evaluate
novel digital spaces and learner modeling techniques.
Background papers
exam president: Prof. Pierre Dillenbourg
thesis advisor: Prof. Tanja Käser
co-examiner: Prof. Robert West
Abstract
The open-nature and heterogeneity of open-ended
learning environments (OELEs) makes processing their interaction
data particularly challenging. Learner modeling methods
developed for a specific OELE often do not transfer to other
OELEs easily, if at all. This diminishes the generalizability of
data processing techniques and reproducibility of experimental
results. To tackle this issue, we propose to define an intermediate
representation for student traces derived from interaction data
that could be used universally among a well-defined subset
of OELEs. Data processing algorithms could then leverage
this shared representation on interaction data originating from
different OELEs without methodological changes. We hope our
contribution can lower the effort required to create and evaluate
novel digital spaces and learner modeling techniques.
Background papers
- Exploring networks of problem-solving interactions (https://dl.acm.org/doi/abs/10.1145/2723576.2723630)
- Integrating Model-Driven and Data-Driven Techniques for Analyzing Learning Behaviors in Open-Ended Learning Environments (https://ieeexplore.ieee.org/abstract/document/7368935)
- Mapping Python Programs to Vectors using Recursive Neural Encodings (https://educationaldatamining.org/EDM2021/virtual/static/pdf/EDM21_paper_J499.pdf)
Practical information
- General public
- Free
Contact
- edic@epfl.ch