Our recommendations on Open Science practices in educational data mining/learning analytics, focusing on: * Ways to prevent comparison-hacking * When preregistration not appropriate ++ Context, containerization, deidentification learninganalytics.upenn.edu/ryanbaker/open… Stephen Hutt Christopher A Brooks

Confusion and Frustration are each complex, and manifest in a variety of ways. Some forms of Confusion and Frustration are highly related. Our theoretical preprint on what we term the confrustion constellation. osf.io/preprints/edar… w/ Elizabeth Cloude

Analysis of use of ASSISTments A/B tests & data: A/B testing used by USA-based researchers Open data used for studying knowledge tracing algorithms in China; used for wide range of purposes in other countries EDM: where 1st-time researchers publish learninganalytics.upenn.edu/ryanbaker/scie…

I'm pleased to announce the 8th edition of our Massive Online Open Textbook *Big Data and Education* (the MOOC will update in a few weeks). learninganalytics.upenn.edu/MOOT/bigdataed… There's an 80% new chapter on large language models, and other updates as well. The world is moving fast!

A multi-dimensional framework for gender explains more variance in student math knowledge than the traditional binary framework. learninganalytics.upenn.edu/ryanbaker/AIED…

Ordered network analysis in intro Comp Sci; differences in novice & more experienced students' patterns involving help-seeking and response to unsuccessful unit tests learninganalytics.upenn.edu/ryanbaker/ONA_…

Blog post describing our partnership with Top Parent to use learning analytics and learning engineering to improve learning for children in low-income families in India centralsquarefoundation.org/articles/top-p…

Investigating algorithmic bias in grade prediction based on student regional background in the Philippines; no evidence for bias found. learninganalytics.upenn.edu/ryanbaker/2405… w/ Didith Rodrigo

What cultural values do GPT-4o, 4, 3.5, 3 express? Using World Values Survey questions, we find GPT consistently aligns with English-speaking countries/Protestant Europe. We show that Cultural Prompting improves alignment. arxiv.org/abs/2311.14096 Yan Tao Olga Viberg Baker EDM Lab 배박사 연구소

Initial work studying the impact of parent-facing videos in Top Parent, a learning platform serving economically underpriviliged families in India: If a parent watches more educational videos, the student makes more progress in the app. learninganalytics.upenn.edu/ryanbaker/pare…

GPT hints in intro programming embedded into the compiler improves affect and leads to better performance even after hints are taken away: a second study replicating our earlier results. learninganalytics.upenn.edu/ryanbaker/ITIC…

A jupyter notebook for rescaling detector confidences. For use when a set of detectors are scaled/centered differently due to the use of resampling/SMOTE/etc. Fits a set of constants. github.com/pcla-code/simp…

Jadud's error quotient is better at predicting CS1 exam grades than repeated error density or error counts. learninganalytics.upenn.edu/ryanbaker/2404… w/ Valdemar Švábenský, Maciej Pankiewicz, Joyce Zhang, Elizabeth Cloude, Eric Fouh

Usage and performance data from nursing learning platforms can predict formative assessment and (less well) licensure exam. learninganalytics.upenn.edu/ryanbaker/L@S-…

International Educational Data Mining Society 2024 Test of Time Award goes to our 2012 paper on sensor-free affect detection learninganalytics.upenn.edu/ryanbaker/EDM%…

Very honored to be profiled here. It's been an amazing journey so far, being able to learn from 517 collaborators (and so many more friends and colleagues and mentors!). We all lift each other up, advance science, & make a better world together. the-learning-agency.com/the-cutting-ed…

Have a bold vision for transforming K-12 learning using advanced tech? Renaissance Philanthropy is seeking breakthrough approaches where technological advancement meets educational need to guide future investments & catalyze impact. RFI responses due by Aug. 16! form.fillout.com/t/boBPHLEzQWus

First white paper on JeepyTA, our GPT-powered TA at Penn. Gets students responses at all hours, generally seen as being as good as a human TA (but not for SRL or motivation). Coming soon: evidence that its feedback leads to better final assignments. learninganalytics.upenn.edu/ryanbaker/Jeep…

We are publicly releasing all the resources from our IES-funded digital training certificate "Data Science Methods for Digital Learning Platforms" learninganalytics.upenn.edu/IES/digital_tr… w/ Seiyon M Lee Anthony Botelho and several great instructors

Hi friends and colleagues! Please follow me on LinkedIn from now on.