CoChat
Sign Up
Cited by 41Online Learning and AnalyticsEthics and Social Impacts of AIAI and HR Technologies

Should learning analytics models include sensitive attributes? Explaining the why

Deho, Oscar Blessed, Joksimovic, Srecko, Li, Jiuyong, Zhan, Chen, Liu, Jixue, Liu, Lin·2023·IEEE Transactions on Learning Technologies
DOI 10.1109/tlt.2022.3226474
Open this paper in CoChat — free account

Abstract

Many educational institutions are using predictive models to leverage actionable insights using student data and drive student success. A common task has been predicting students at risk of dropping out for the necessary interventions to be made. However, issues of discrimination by these predictive models based on protected attributes of students have recently been raised. An important question that is constantly asked is: should the protected attributes be excluded from the learning analytics (LA) models in order to ensure fairness? In this article, we aimed at answering questions that if we exclude the protected attributes from the LA models, does the exclusion ensure fairness as itsupposedlyshould? Does the exclusion affect the performance of the LA model? If so, why? We found answers to these questions and went further to explain why. We built machine learning models and performed empirical evaluations using a three-year dropout data for a particular program in a large Australian university. We found that excluding or including the protected attributes had marginal effect on predictive performance and fairness. Perhaps not surprisingly, our findings suggest that the effect of including or excluding protected attributes is a function of how they relate with the prediction outcome. More specifically, if a protected attribute is correlated with the target label and proves to be an important feature, then their inclusion or exclusion would have effect on the performance and…

Integrity signals

Registry

Crossref · 10.1109/tlt.2022.3226474

Cites

54 of 54 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • AI systems may replicate social and cultural inequalities embedded in training data, influencing assessment outcomes and learning opportunities.

    “Empirical studies show that AI systems may replicate social and cultural inequalities embedded in training data, influencing assessment outcomes and learning opportunities (Baker & Hawn, 2022;Barbierato et al., 2022;Deho et al., 2023;Jiang & Pardos, 2021).”

    From Academic Integrity and Students’ Ethical Use of ChatGPT in Higher Education (2026) · cites this paper as Deho et al., 2023

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Deho, O. B., Joksimovic, S., Li, J., Zhan, C., Liu, J., & Liu, L.
Affiliations
  • University of South Australia (AU)
  • UniSA Education Futures, University of South Australia, Adelaide, SA, Australia
ORCID
Venue
IEEE Transactions on Learning Technologies
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Published
2023-08
Volume
16
Issue
4
Pages
560-572
ISSN
  • 1939-1382 (ISSN-L)
  • 2372-0050
DOI
10.1109/tlt.2022.3226474
OpenAlex
W4312816903
Cited by
35
References
54
Citations by year
2026: 5 · 2025: 9 · 2024: 15 · 2023: 6
FWCI
5.80
Type
journal article
Language
en
Open access
closed
Licence
Locations
  • IEEE Transactions on Learning Technologies (publishedVersion)
  • UniSA Research Outputs Repository (University of South Australia) (submittedVersion)
  • Adelaide Research & Scholarship (AR&S) (University of Adelaide) (submittedVersion)
Funders
  • Australian Research Council (DP200101210)
  • Postgraduate Research Scholarship of University of South Australia
Topics
  • Physical Sciences › Computer Science › Computer Science Applications › Online Learning and Analytics
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › AI and HR Technologies
Keywords
learning analytics, protected characteristics, algorithmic fairness, algorithmic discrimination, student dropout prediction, sensitive attributes, model performance, university students

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review