1) The document proposes a personalized remedial recommendation system for SQL programming practice that provides recommendations on learning content and explanations for the recommendations.
2) A study was conducted with students in a database management course to evaluate the persuasiveness and effects of the recommendations with and without explanations.
3) The results found that recommendations affected students' selection of learning content and that explanations helped students critically judge recommendation quality but did not significantly affect learning outcomes. Limitations and future work are discussed.
Personalized Remedial Recommendations for SQL Programming Practice System (Jordan Barria-Pineda in APCSE20 at UMAP20)
1. Personalized Remedial Recommendations for
SQL Programming Practice System
Jordan Barría-Pineda ( @j_barria), Kamil
Akhuseyinoglu ( @cskamil1), Kerttu Pollari-Malmi,
Teemu Sirkiä, Lauri Malmi, Peter Brusilovsky
Adaptive and PersonalizedComputer Science
Education (APCSE) workshop @ UMAP’20
1
6. Exploring the Need for Explainable Artificial Intelligence
(XAI) in Intelligent Tutoring Systems (ITS) (2019)Vanessa
Putnam and Cristina Conati at IUI’ 19
Complementing educational recommender systems
with open learner models (2020) Solmaz Abdi et. al. at
LAK’ 20
2 Related work APCSE
Workshop
Improving Student-System InteractionThrough Data-
driven Explanations of Hierarchical Reinforcement
Learning Induced Pedagogical Policies. (2020) Guojing
Zhou et. al. at UMAP’20
6
Overall positive students’ attitude towards
explanations (why?, how?) for hints in ITS (specially
when disagreeing with the system)
Explaining the tutor’s decisions to students through
data- driven explanations could improve the
student-system interaction in terms of students’
engagement and autonomy.
Some students stated feeling of unfairness when
given recommendations through OLM
“explanations”
8. Jordan Barria-Pineda and Peter Brusilovsky. 2019. Making Educational RecommendationsTransparent through a Fine-GrainedOpen
Learner Model. In IUI’19Workshops.
APCSE
Workshop
8
Mastery Grids
with
transparent
recommenda-
tions
3 Proposed approach
9. 3 Proposed approach
Learning Content Recommendations
APCSE
Workshop
9
Roya Hosseini, I-Han Hsiao, Julio Guerra, and Peter Brusilovsky. 2015. What Should I Do Next? Adaptive Sequencing in the Context
of Open Social Student Modeling. In Design for Teaching and Learning in a Networked World, Gráinne Conole, Tomaž Klobučar,
Christoph Rensing, Johannes Konert, and Elise Lavoué (Eds.). Springer International Publishing, Cham, 155–168.
23. 6 Discussion / Conclusions APCSE
Workshop
Remedial recommendations affect learning content selection behavior
of students
Explanations can help students to add their own judgment on the top of
recommendation to make more balanced decisions about working or not on an
open recommended problem (rather than rushing to anything that is
recommended)
23
With full explanations (i.e., more information about the generation), the
recommendations’ quality was more likely to be critically judged
+
This help could be useful in online learning contexts where either student modeling
or recommendation approaches are not highly accurate.
24. 7 Limitations / Future Work
Students accessed practice content at the
end of their course, a chance for
recommendations to be less than perfect
was relatively high. Marginally significant effects
24
Replicating the study in a setup where students would use the system
throughout the whole term
Exploring explanations to explain other information (e.g. input data) and
recommendations focused on other learning goals (e.g. reinforcing existing
knowledge)
APCSE
Workshop
25. Thanks for your attention!
If you have any question please ask me or
email me to jab464@pitt.edu
In case you are teaching Java, Python or SQL classes we are open for
collaborations!
@j_barria
Editor's Notes
Educational Recommender Systems (EdRecSys) are different in nature from conventional Recommender Systems (RecSys) -mostly related to e-commerce- as the main goal of EdRecSys is supporting students learning’ instead of maximizing users’ satisfaction from consuming the recommended items. Thus, research on transparency for traditional RecSys is hard to transfer from e-commerce contexts to educational scenarios, as the level of knowledge of the end-user (i.e. the student) is crucial for generating and evaluating the impact of the recommendations on students’ learning.
A classroom experiment was run where we implemented a Mastery Grids version in an introductory Java course with a rule-based recommendation engine designed for supporting students’ knowledge expansion [1] (see Fig. 1). The group that received explanations exhibited more persuasive recommendations in terms of adopting the suggested material as part of their learning process.
College students
2X2 design
These results, combined, make us hypothesize that the inclusion of individual explanatory elements can lead students to think a little bit more about the appropriateness of the recommendations, especially when they have the meta-cognitive ability to do this (more likely to happen at the end of the course where students have a higher degree of self-awareness about their own knowledge).
This shows that students overall persistence on problems, once they start working on them, is lower when they only had access to partial explana- tions (opposite phenomenon discovered for access rate).
No significant differences in system’s satisfaction.
Rec activities highlighted with stars were on average more attractive
As the data show, the learners exhibited lower conversion rates for recommended problems than for non-recommended ones when any type of explanations (i.e., visual or textual) were provided. This result suggests that both explanation types enabled students to better add their own judgment on the top of recommendation and make a more balanced decision about working or not on an open recommended problem rather than rushing to anything that is recommended.
when the students accessed practice content at the end of their course, a chance for recommendations to be less than perfect was relatively high since the system was not able to track knowledge gained by the students through most of their work in class